# MULTIPLE IDENTITIES MANAGEMENT

EDITED BY: Clara Kulich, Soledad de Lemus, Natasza Kosakowska-Berezecka and Fabio Lorenzi-Cioldi PUBLISHED IN: Frontiers in Psychology

#### *Frontiers Copyright Statement*

*© Copyright 2007-2018 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

*The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.*

*Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.*

*Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.*

*As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.*

> *All copyright, and all rights therein, are protected by national and international copyright laws.*

*The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use.*

ISSN 1664-8714 ISBN 978-2-88945-429-7 DOI 10.3389/978-2-88945-429-7

# About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

# Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

# Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

# What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# **MULTIPLE IDENTITIES MANAGEMENT**

## Topic Editors:

**Clara Kulich,** University of Geneva, Switzerland **Soledad de Lemus,** University of Granada, Spain **Natasza Kosakowska-Berezecka,** University of Gdan'sk, Poland **Fabio Lorenzi-Cioldi,** University of Geneva, Switzerland

Image: Vasilius/Shutterstock.com

In this ebook, a collection of 18 papers presents empirical research, as well as novel theoretical considerations, on how multiple identities are being managed by the individuals holding them. The papers draw on theories from social psychology in the context of the social identity approach. The first chapter presents eight papers on different types of multiple identity configurations in a variety of contexts, and the costs and benefits of these configurations for the individual (e.g., well-being). The second chapter gives insights on how conflict between multiple identities is managed by individuals. And the final chapter analyses how multiple identities impact intragroup and intergroup relations.

**Citation:** Kulich, C., de Lemus, S., Kosakowska-Berezecka, N., Lorenzi-Cioldi, F., eds. (2018). Multiple Identities Management. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-429-7

# Table of Contents

*05 Editorial: Multiple Identities Management: Effects on (of) Identification, Attitudes, Behavior and Well-Being* Clara Kulich, Soledad de Lemus, Natasza Kosakowska-Berezecka and Fabio Lorenzi-Cioldi **1. Multiple Identity Configurations – Benefit or Cost?** *09 Conceptualizing the Dynamics between Bicultural Identification and Personal Social Networks* Lydia Repke and Verónica Benet-Martínez *28 Self-consistency in Bicultural Persons: Dialectical Self-beliefs Mediate the Relation between Identity Integration and Self-consistency* Rui Zhang, Kimberly A. Noels, Richard N. Lalonde and S. J. Salas *42 Counter-Stereotypes and Feminism Promote Leadership Aspirations in Highly Identified Women* Carola Leicht, Małgorzata A. Gocłowska, Jolien A. Van Breen, Soledad de Lemus and Georgina Randsley de Moura *54 Multiple Social Identities Enhance Health Post-Retirement Because They Are a Basis for Giving Social Support* Niklas K. Steffens, Jolanda Jetten, Catherine Haslam, Tegan Cruwys and S. Alexander Haslam *65 The More (Social Group Memberships), the Merrier: Is This the Case for Asians?* Melissa X.-L. Chang, Jolanda Jetten, Tegan Cruwys, Catherine Haslam and Nurul Praharso *82 Multiple Group Membership and Well-Being: Is There Always Strength in Numbers?* Anders L. Sønderlund, Thomas A. Morton and Michelle K. Ryan *102 Women (Do Not) Belong Here: Gender-Work Identity Conflict among Female Police Officers* Jenny Veldman, Loes Meeussen, Colette Van Laar and Karen Phalet *114 Stereotype Threat and Perceptions of Family-Friendly Policies among Female Employees*

Courtney von Hippel, Elise K. Kalokerinos and Hannes Zacher

# **2. How to Solve Multiple Identity Conflict**

*124 Similarly Torn, Differentially Shorn? The Experience and Management of Conflict between Multiple Roles, Relationships, and Social Categories*

Janelle M. Jones and Michaela Hynie

*140 Does Identity Incompatibility Lead to Disidentification? Internal Motivation to Be a Group Member Acts As Buffer for Sojourners from Independent Cultures, Whereas External Motivation Acts As Buffer for Sojourners from Interdependent Cultures*

Christina Matschke and Jennifer Fehr

*150 Combining Gender, Work, and Family Identities: The Cross-Over and Spill-Over of Gender Norms into Young Adults' Work and Family Aspirations* Loes Meeussen, Jenny Veldman and Colette Van Laar

# **3. Multiple Identities, Intragroup and Intergroup Relations**

*161 The Role of Indian Caste Identity and Caste Inconsistent Norms on Status Representation*

Sindhuja Sankaran, Maciek Sekerdej and Ulrich von Hecker

*175 "I Want, Therefore I Am" – Anticipated Upward Mobility Reduces Ingroup Concern*

Marion Chipeaux, Clara Kulich, Vincenzo Iacoviello and Fabio Lorenzi-Cioldi


Adrienne Pereira, Eva G. T. Green and Emilio Paolo Visintin


Aharon Levy, Tamar Saguy, Eran Halperin and Martijn van Zomeren

# Editorial: Multiple Identities Management: Effects on (of) Identification, Attitudes, Behavior and Well-Being

Clara Kulich<sup>1</sup> \*, Soledad de Lemus <sup>2</sup> , Natasza Kosakowska-Berezecka<sup>3</sup> and Fabio Lorenzi-Cioldi <sup>1</sup>

<sup>1</sup> Social Psychology, Section of Psychology, Faculty of Psychology and Educational Sciences, University of Geneva, Geneva, Switzerland, <sup>2</sup> Department of Psychology, Centro de Investigación Mente, Cerebro y Comportamiento, University of Granada, Granada, Spain, <sup>3</sup> Division of Cross-Cultural Psychology and Psychology of Gender, Institute of Psychology, Faculty of Social Sciences, University of Gdansk, Gdansk, Poland

Keywords: multiple identities, social identification, dissonance, coping, conflict, well-being

#### **Editorial on the Research Topic**

#### **Multiple Identities Management: Effects on (of) Identification, Attitudes, Behavior and Well-Being**

Individuals belong to gender, ethnic, or national groups; they can be categorized depending on their religious beliefs, or the activities they are involved in such as professions, political groups, etc. These multiple social identities differ considerably in the way they are acquired (e.g., inherited or achieved through accomplishments), in their relative stability or malleability, and in the value which they assign to the individual (e.g., low vs. high social status).

Inherited identities (based on gender, skin color, class background, etc.), and in some cases achieved ones (e.g., through migration or professional mobility), cannot be voluntarily dismissed. Although inherited and achieved social identities tend to correspond in their value and content over time (Bourdieu, 1979; Ridgeway and Erickson, 2000), identities can differ in status and value, and create distressing experiences which call for a coping strategy in order to increase identity fit (e.g., Deaux and Greenwood, 2013; Turner-Zwinkels et al., 2015). However, there is also evidence that multiple identities may provide a pathway to gain social support and positively influence individuals' well-being (Walter et al., 2015).

When individuals face conflicting identities in terms of status or value, they use different coping strategies. Individuals may attempt to discard one of the identities, use in turn one or the other, or integrate or fuse both identities (e.g., Roccas and Brewer, 2002; Deaux, 2008; Shields, 2008; Berry and Sabatier, 2011). Identities may also change or be newly developed in particular contexts (e.g., politicized, opinion-based, or solidarity-based groups; McGarty et al., 2009). The type of strategy is also likely to have an impact on well-being, as it has the potential to reduce dissonance and distress (e.g., Sampson, 1969; Jetten et al., 2012). Research shows that upward (Derks et al., 2011; Kulich et al., 2015) and downward (Jetten et al., 2015) individual mobility impinge on attitudes and support for other ingroup members. Moreover, conditions such as situational threats to a social identity (e.g., Kosakowska-Berezecka et al., 2016), as well as the type of integration culture (e.g., colorblindness vs. multiculturalism) moderate whether multiple identities produce positive or negative outcomes (Wilton et al., 2015).

In this Research Topic, we focus on contexts in which two or more social categories are simultaneously salient. Eighteen papers present empirical research, as well as novel theoretical considerations, to understand how such multiple identities are being managed by the individuals

#### Edited and reviewed by:

Anat Bardi, Royal Holloway, University of London, United Kingdom

> \*Correspondence: Clara Kulich Clara.Kulich@unige.ch

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

Received: 16 November 2017 Accepted: 12 December 2017 Published: 22 December 2017

#### Citation:

Kulich C, de Lemus S, Kosakowska-Berezecka N and Lorenzi-Cioldi F (2017) Editorial: Multiple Identities Management: Effects on (of) Identification, Attitudes, Behavior and Well-Being. Front. Psychol. 8:2258. doi: 10.3389/fpsyg.2017.02258 holding them. We were particularly interested in the positive and negative outcomes they bring for individuals' well-being, and in the importance of multiple group memberships for intergroup relations. In the following, we provide a brief discussion of, and structure for, the main themes included in this Research Topic.

# MULTIPLE IDENTITY CONFIGURATIONS—BENEFIT OR COST?

Identity configurations take many different forms depending on contextual factors. Looking at the antecedents of different identity configurations, Repke and Benet-Martínez highlight that the structure of an immigrant's social network and interconnection of same ethnicity (rather than the number of coethnic and host individuals in the networks) play a significant role in predicting whether individuals develop coexisting cultural identifications, conflicting cultural identifications, or a mixture of the two.

Belonging to a number of groups may be beneficial, as group membership enhances individuals' well-being and other positive outcomes. Zhang et al. show across diverse groups of bicultural Canadians that having an integrated bicultural identity is related to being more consistent across roles and more congruent and less ambiguous about self-evaluations. Thus, both content of heritage culture and the dynamic process of integrating cultural identities affect self-consistency among biculturals. Levels of identification with multiple groups affect the group members' performances or preferences. Leicht et al. show that women are more likely to manifest leadership aspirations in a work scenario when they indicate simultaneous high identifications with women and with feminism, but only in a context where gender counter-stereotypes are made salient (as compared to a stereotypical context). Moreover, cross-sectional work by Steffens et al. suggests that multiple social identities are positively related with better health and increased well-being in retirement because they allow individuals to both give and receive social support. However, these positive outcomes are often contingent on specific conditions. In this way, Chang et al.'s studies and meta-analysis propose that, although for Europeans multiple group memberships have a number of positive outcomes, for Asians they do not convey better well-being because receiving social support is perceived as burdening others. It is thus uncomfortable for Asians to draw from such psychological resources. In accordance, Sønderlund et al. argue, based on correlational studies, that specific features of the groups, such as their social value and visibility to others, need to be considered beyond the number of identities. Indeed, low value of highly visible identities makes them more of a burden than a benefit for individuals' well-being.

Multiple group memberships may further engender dissonance and threat and thus be troubling for the individual. Identities that differ in value and content are particularly common among members of inherited low status groups in work situations. For instance, Veldman et al. find that female police officers' experience of being gender-dissimilar from the work-team members is associated with gender-work identity conflict. As a consequence, female police officers identify less with their team, leading to negative work and health outcomes such as lower work satisfaction and motivation, higher burnout, and turnover intentions. von Hippel et al. further show that the incompatibility of female gender and work identities impacts on attitudes toward family-friendly policies. They demonstrate that women suffering from stereotype threat perceive such policies as having negative career consequences—although paradoxically they are still more inclined to use them.

# HOW TO SOLVE MULTIPLE IDENTITY CONFLICT

When identity conflict occurs, different identity strategies serve to reduce this inconsistency. Jones and Hynie focus on the experience and management of conflict between different types of multiple identities: roles (e.g., being a student), relational (being a friend), or social identities (nationality). They suggest four strategies of identity management: reconciliation, retreat, realignment, and reflection.

Matschke and Fehr analyze the impact of incompatibilities between an established social identity and a potential new social identity in an acculturation context. They suggest that lack of compatibility between cultural identities leads to higher disidentification with the receiving society. Depending on cultural self-construal (individualistic or collectivistic), individuals make use of different resources (related to either intrinsic or extrinsic motivation) to fight the negative effect of incompatibility on the social identity.

And finally, Meeussen et al. focus on managing potentially conflicting work-family identities. They point to the importance of perceived gender norms in achieving compatibility between gender, family, and work. Their findings show that gender identity influences women's and men's work-family conflict resolution aspirations through social norms. Thus, changing these norms could allow both men and women to combine their multiple identities in a more effective and self-enhancing way.

# MULTIPLE IDENTITIES, INTRAGROUP, AND INTERGROUP RELATIONS

Multiple identities can compromise interpersonal relations by creating conflict within the group. In this line, Sankaran et al. present an experimental study in the context of the Indian caste system, where social mobility is strongly restricted. They find evidence for the black sheep effect: When an ingroup member (as compared to an outgroup member) threatens the high-castes' social identity by violating an ingroup norm (morality), high caste identity increases in concert with the derogation of the ingroup perpetrator. In parallel, Chipeaux et al. investigate, with correlational data, moves from a disadvantaged ingroup to a higher status outgroup. They observe a decrease in ingroup concern among mobile individuals and among those who anticipate moving up the hierarchy, compared to nonmobile individuals. These results suggest that multiple identities with differing statuses lead to identity conflict, resulting in the discount of the low-status ingroup. Finally, Chiou and Mercado analyze whether biculturals have fixed or shifting loyalties toward their home/host culture. Their findings reveal that the degree to which biculturals manifest loyalty to each of the respective cultures is easily influenced by a priming procedure, demonstrating that bicultural loyalties are rather malleable.

Further, multiple identities can contribute to a more nuanced understanding of intergroup relationships. For instance, van Breen et al. show that a multiple identity approach to gender allows predicting women's attitudes toward gender stereotypes as well as collective action. They propose that women's attitudes are regulated by two dimensions of gender identity: identification with women and with feminists (politicized identity).

Indeed, interacting social identities play an important role in mobilizing resources for activism. Pereira et al. analyze the "sedating" effect of positive intergroup contact for Roma minority activism. They demonstrate that positive contact reduces Roma's ethnic identification and activism but only among low national identifiers. Thus, national identification can buffer against negative effects of contact on collective action. Similarly, in the context of Ukraine ban protests of 2014, Chayinska et al. show that national and politicized identities (online and street protest groups) predict support for collective action if protest is perceived as legitimate and politicized identities as compatible.

Finally, Levy et al. focus, in a theoretical contribution, on how the salience of multiple group memberships impacts on within-group and between-groups relations. They define gateway groups as individuals who simultaneously belong to two groups who are in conflict and propose that such gateway groups operate as both bridges and barriers in intergroup relations, thus improving or worsening intergroup relations. They conclude that the way in which gateway groups build their multiple identities and shift across categories play an important role in reducing group stereotypes and fostering tolerance toward outgroups.

# CONCLUSIONS

The aim of this Research Topic was to extend a growing body of research on multiple identities by focusing on how people

# REFERENCES


negotiate conflicting social identities and the consequences of such negotiations. The present collection of research findings investigated the moderating role of group status, group visibility, dissimilarity from others, diversity cultures, and types of identity configurations on the ways individuals handle their multiple identities. They further investigated the particularities of psychological constructs such as social identification, ingroup and inter-group attitudes and behaviors (e.g., collective action and social mobility), and well-being in the context of multiple identity configurations of different kinds (e.g., gender, professional, and politicized identities). Future research should focus on a systematic analysis which extends the correlational level in order to better understand the specific contexts and the underlying processes that determine whether multiple identities are beneficial or detrimental for the individual's well-being. Moreover, research should examine how social perceptions of multiple identities impact the process of coping and identity management in individuals who are holding these identities. This implies bringing together the literature on the perspective of individuals holding multiple identities, presented here, and the literature on attitudes toward individuals with multiple identities.

# AUTHOR CONTRIBUTIONS

CK, SdL, NK-B, and FL-C all drafted the Research Topic proposal, took on editorial tasks, and participated in writing and commenting the editorial.

# FUNDING

CK and FL-C were supported by the Swiss National Science Foundation Project funding grant N◦ 100014\_149197/1. SdL was supported by Grant n◦ PSI2016-79971-P from the Spanish Ministry of Science and Technology (AEI/FEDER, UE).

# ACKNOWLEDGMENTS

We thank all authors and reviewers for their contributions to this project.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Kulich, de Lemus, Kosakowska-Berezecka and Lorenzi-Cioldi. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Conceptualizing the Dynamics between Bicultural Identification and Personal Social Networks

Lydia Repke<sup>1</sup> \* and Verónica Benet-Martínez1,2

<sup>1</sup> Department of Political and Social Sciences, Universitat Pompeu Fabra, Barcelona, Spain, <sup>2</sup> ICREA (Catalan Institution for Research and Advanced Studies), Barcelona, Spain

An adequate understanding of the acculturation processes affecting immigrants and their descendants involves ascertaining the dynamic interplay between the way these individuals manage their multiple (and sometimes conflictual) cultural value systems and identifications and possible changes in their social networks. To fill this gap, the present research examines how key acculturation variables (e.g., strength of ethnic/host cultural identifications, bicultural identity integration or BII) relate to the composition and structure of bicultural individuals' personal social networks. In Study 1, we relied on a generationally and culturally diverse community sample of 123 Latinos residing in the US. Participants nominated eight individuals (i.e., alters) from their habitual social networks and across two relational domains: friendships and colleagues. Results indicated that the interconnection of same ethnicity alters across different relationship domains is linked to cultural identifications, while the amount of coethnic and host individuals in the network is not. In particular, higher interconnection between Latino friends and colleagues was linked to lower levels of U.S. identification. Conversely, the interconnection of non-Latino friends and colleagues was associated with lower levels of Latino identification. This pattern of results suggests that the relational context for each type of cultural identification works in a subtractive and inverse manner. Further, time spent in the US was linked to both Latino and U.S. cultural identifications, but this relationship was moderated by the level of BII. Specifically, the association between time in the US and strength of both cultural identities was stronger for individuals reporting low levels of BII. Taking the findings from Study 1 as departure point, Study 2 used an agent-based model data simulation approach to explore the dynamic ways in which the content and the structure of an immigrant's social network might matter over time in predicting three possible identity patterns: coexisting cultural identifications, conflicting cultural identifications, and a mixture of the two. These simulations allowed us to detect network constellations, which lead to identification or disidentification with both cultures. We showed that distinct patterns of social relations do not lead to identity outcomes in a deterministic fashion, but that often many different outcomes are probable.

#### Edited by:

Natasza Kosakowska-Berezecka, University of Gdansk, Poland ´

#### Reviewed by:

Peter Bevington Smith, University of Sussex, UK Gabriel Horenczyk, Hebrew University of Jerusalem, Israel

> \*Correspondence: Lydia Repke lydia.repke@upf.edu

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

> Received: 28 October 2016 Accepted: 14 March 2017 Published: 31 March 2017

#### Citation:

Repke L and Benet-Martínez V (2017) Conceptualizing the Dynamics between Bicultural Identification and Personal Social Networks. Front. Psychol. 8:469. doi: 10.3389/fpsyg.2017.00469

Keywords: social networks, acculturation, immigration, cultural identification, complex contagions

# INTRODUCTION

Allport's (1954) theory on intergroup contact states that, under certain conditions, contact between members of minority and majority groups will not only reduce prejudice and conflict, but will also improve interethnic attitudes (Pettigrew and Tropp, 2000; Binder et al., 2009). Nowadays, various scholars agree that one prerequisite for immigrants' successful and peaceful integration into their host society is that they develop social networks which include host culture contacts in central positions, as these contacts provide access to critically important social and informational resources (Smith, 2013; Damstra and Tillie, 2016). These host nationals may improve the immigrant's acculturation potential by helping with the acquisition of culturally appropriate skills and by providing exposure to new norms and value systems (Ward and Kennedy, 1993; Kim, 2001; Smith, 2005, 2013; Jasinskaja-Lahti et al., 2006). However, contact with coethnic individuals (living in the country of origin and in the country of destination) is beneficial as well. Coethnic friends and relatives living back home may give social support, safeguard the immigrant's ethnic identity and skills, and even encourage adjustment to the new society (Lebon, 1983; Smith, 1999; Schultz, 2001). Similarly, coethnics in the country of destination may give important information and access to resources related to adapting to the host society (e.g., where and how to find a job), reducing the immigrant's costs and risks in the country of settlement (Liu, 2013). Having said this, a social network comprised of too many coethnic individuals might be a burden to the immigrant's acculturation potential, as the immigrant may feel pressured to hold on to habits or customs from the country of origin and may also lose an opportunity to learn and practice the host culture behaviors and norms (Luo and Wiseman, 2000). Ultimately, these processes may depend on the available social network opportunities, how much ethnic and host cultures objectively differ from each other (i.e., how much new cultural learning is called for), and whether the individual internalizes the differences as reflecting cultural conflict and incompatibility (Searle and Ward, 1990; Benet-Martínez and Haritatos, 2005).

Even though patterns such as low levels of identification with the host society and scarce friendships with host individuals are widely recognized in the literature, their interrelation is still open to question. Leszczensky (2013), for example, finds only a spurious relationship between degree of national identification and share of host national friends. Given the importance of social networks for integration and acculturation, it is surprising that hardly any study has examined how key acculturation variables (e.g., ethnic and host cultural identifications, bicultural identity integration, BII) relate to the composition (who is in the network) and structure (how are the network members connected) of immigrants' personal social networks. Up to now, only a few sociological studies have attempted to do so, but did not include psychological measures (e.g., Lubbers et al., 2007; Vacca et al., 2016; but see also Mok et al., 2007). In particular, the relational perspective offered by the social network approach is suited perfectly for the acculturation and immigration context, as it captures intercultural contact in a way that goes beyond the commonly used self-reports. Most psychological research, including acculturation studies, focuses almost exclusively on individual-level characteristics (e.g., self-reported values and behaviors) in an effort to mirror what happens inside of people's minds. But human behavior is also shaped by what happens between people's minds. In this paper, we study how individuals' cultural identities are influenced by their relational contacts, and the interactions that these contacts have between each other (Brown and Zagefka, 2011; Postmes et al., 2015).

We hypothesize that, in order for immigrants and their descendants to develop and strengthen their cultural identifications, repeated contact with culture-specific, attituderelevant information (such as communication styles, cultural activities, gender roles, etc.) from individuals representing different roles is needed. The rationale behind this is the idea of complex contagions, which attests that certain social behaviors may only be changed after having had multiple contact with a variety of sources (e.g., as this adds credibility to the information received) (Centola and Macy, 2007; Centola, 2010). Behavioral changes then would be reflected in a change in cultural identification. More particularly, we argue that it is the interconnection of same ethnicity contacts belonging to different relationship domains (i.e., friendship versus work) that predicts the strength of individuals' multiple cultural identifications.

In two separate studies, we explored possible relationships between key acculturation variables and personal social networks of immigrants and their descendants. In Study 1, we derived predictions for ethnic and host culture identifications from the idea of complex social contagion, and tested them using survey and network data collected from a community sample of 123 Latino-American biculturals residing in the US. In Study 2, using an agent-based model (ABM), we simulated data on the basis of the findings from Study 1 and explored whether and how the content and the structure of a bicultural individual's social network matters over time in negotiating coexisting cultural identifications, conflicting cultural identifications, and a mixture of the two (e.g., being conflicting with regards to one life domain, but coexisting in another one).

We believe that our contribution to the study of multiple identities management in the acculturation context is twofold. From a scientific point of view, we will shed light on the unexplored possible interdependence between the micro-level represented by individuals' self-reported acculturation processes (e.g., strength of cultural identifications and degree of conflict the individual feels between different cultural orientations) and the meso-level represented by these individuals' habitual personal social networks. The fact that, in Study 2, we adopt a process-oriented rather than an outcome-oriented approach allows us to understand better how changes in the network may influence the identity negotiation process of immigrants and their descendants. From a societal point of view, our studies are informative in that they could be used to assist policy-makers involved with the integration of immigrants and other cultural minorities. In particular, the models examined in Study 2 could be used to identify social environments (i.e., specific network constellations) that are beneficial for fostering harmonious multicultural identities, and those that could lead to the development of risky patterns of cultural disidentification

or radicalization. The paper is structured in the following way: First, we present our theoretical framework and our predictions. Then, we describe our two studies and their results. Finally, we summarize our main findings and offer some suggestions for future research in the discussion.

# THEORETICAL FRAMEWORK AND PREDICTIONS

# Acculturation, Cultural Identification, and BII

When moving to a new country, immigrants and their descendants often experience radical changes in their social and cultural contexts. The resulting acculturation processes may be described as psychological and behavioral changes that occur due to intercultural contact (Gibson, 2001; Sam and Berry, 2010). These processes oftentimes involve managing multiple, and sometimes conflictual, cultural value systems and identifications, and they also lead to changes in individuals' social networks. These changes may include the creation of new relationships, the dissolution of old ones, or simply the diminishment or the consolidation of existing connections.

Generally, cultural identification can be understood as the sense of belonging to a cultural group (Nguyen and Benet-Martínez, 2007). In particular, long-term immigrants and their descendants may feel attached to not only one, but several cultures. As people who have been exposed to and who have internalized at least two sets of cultural meaning systems (e.g., beliefs, values, behaviors, languages), these individuals may be described as bicultural or multicultural (Hong et al., 2000; Nguyen and Benet-Martínez, 2007). Multicultural individuals have the capacity to acquire and use several cultural frames, even when these may be conflictual. Experimental research has shown that, depending on the available contextual cues, different cultural frames become salient, and that identification with a cultural group can shift accordingly (Hong et al., 2000, 2003; Verkuyten and Pouliasi, 2006; Verkuyten and Yildiz, 2007). This suggests that cultural identification is rather dynamic, and as such, neither primordial nor predefined, and thus it can undergo change (Lubbers et al., 2007).

Although it has been shown that acculturating individuals prefer the integration mode (i.e., being involved with both the ethnic and host cultures) (Benet-Martínez et al., 2002), these individuals may vary in how much they integrate their different cultural orientations and identities into a coherent sense of self (Huynh et al., 2011). Some biculturals may internalize cultural differences as reflecting conflict and incompatibility, while others may view their cultural orientations as compatible and even blendable. The construct of BII captures these differences and has become a central focus of empirical research on biculturalism (Benet-Martínez et al., 2002; Benet-Martínez and Haritatos, 2005). People high on BII view their two cultural identities as compatible and feel part of a combined (sometimes third) culture, whereas individuals low on BII consider their cultural identities as conflictual and dissociated from one another. The validity of BII as a psychologically meaningful construct has been wellestablished over the past decade, with research pointing to a wide variety of benefits associated with higher levels of integration (for reviews see Huynh et al., 2011; Benet-Martínez, in press).

# Personal Social Networks: Relational Domains and Social Contagion

According to the social network analysis framework, social networks consist of nodes and ties. Nodes are actors (e.g., individuals, groups, organizations), and their ties are connections (i.e., social relations) between them. While sociocentric network studies typically focus on complete networks, personal social network studies take the perspective of one particular actor. This focal node is the respondent in the study and is referred to as ego, which is why these studies are also called egocentric network studies. The members of ego's network are called alters.

In Study 1, we were particularly interested in two relational domains that personal social networks commonly entail and that cross-cultural research often highlights (Pouliasi and Verkuyten, 2007): (1) close friends (excluding family members) and (2) classmates, co-workers or colleagues (who are not friends)<sup>1</sup> . Even though the interpretation of the term friend is culture and language specific (Scheuch, 1968; Fischer, 1982), people have more or less an understanding of what a friend is. Generally speaking, individuals tend to choose their friends freely from the social contexts available to them. They are not born into a circle of friends, like they are born into a family whose members are to a wide extent given. As such, people have some influence on the composition of their friendship network. Further, individuals might influence the structure of their friendship network by introducing friends from different areas of life to each other or by keeping them intentionally separate. Nevertheless, ego's close friends tend to be engaged in each others' social lives, whereas ego's acquaintances are likely to not be involved with one another (Granovetter, 1983).

Colleagues, on the other hand, are often more given than selected freely. In some cases, people might have some influence on who becomes a colleague, but, normally, they cannot choose them as they wish. For instance, who becomes a colleague depends on who applies for a job, whether a particular candidate matches the job description, gets selected to fill in the position and, then, also accepts the offer. However, individuals may indirectly determine who their colleagues are by specifically deciding to work in an environment that is ethnic homogenous versus heterogeneous, or mainly coethnic versus non-coethnic. Yet, they are more restricted in choosing their colleagues than in selecting their friends. Further, people's influence on how their colleagues are connected among each other might be quite limited as well, as the structure is often given by the company's internal organization. In contrast, connections between friends and colleagues are usually not imposed by some third party and do not occur as naturally as maybe among friends, which leaves more freedom for ego to actively initiate relationships between alters of different relational domains (**Table 1**).

<sup>1</sup>From now on, we will refer to the second group in shorthand as 'colleagues' to avoid repetition.


Previous research has shown that friendships between immigrants and natives are positively linked to identification with host culture (e.g., Phinney et al., 2006; Leszczensky, 2013), and that friendships among coethnics are positively associated with higher levels of ethnic identification (e.g., Phinney et al., 2001; Ono, 2006). Yet, these studies did not actually measure social networks, but rather relied on self-reported number of friendships or frequency of contact (for exceptional examples see Lubbers et al., 2007; Mok et al., 2007). Social desirability and other types of biases might influence these responses. People may lie consciously about their social interactions with others, or may be influenced by memory biases and wishful thinking. In contrast, social network data does not rely on people's self-assessment of their social lives, and instead maps onto actual contact between people. In this way, the network data collection mode is a more implicit and less obtrusive approach, and yields less danger of being actively manipulated by the respondent (Molina et al., 2014). Ergo, in the context of studying acculturation, the network approach more adequately grasps real-life situations of intercultural contact, while also measuring directly and more objectively with whom an individual interacts. Relying on real network data, we argue that the interconnection between friends and colleagues of the same ethnicity is a stronger predictor for cultural identification than the mere amount of alters belonging to a particular ethnic group. Our assumption is that receiving attitude-relevant information from individuals representing different roles (i.e., friend, colleague) strengthens the effect of attitude formation in the context of migration.

There are different theories about how behaviors spread via social contact in social networks. The most famous one is Granovetter's (1973) seminal theory on the strength of weak ties (SWT). He defines the strength of a tie as a "combination of the amount of time, the emotional intensity, the intimacy [. . .], and reciprocal services" (p. 1361). Usually, the concept of tie strength is measured by how well ego knows the network members or how close ego is to the alters (Marsden and Campbell, 1984). Duration of relationship, frequency of contact, and relationship categories are often used as proxies although, empirically, they are not necessarily correlated with tie strength. For example, family members do not need to have strong ties among each other, although in many cases immediate family members probably do.

The first premise of Granovetter's theory states that the stronger the tie between two people A and B, the more likely it is that their social worlds overlap. So if A and B have a strong tie, and A and C have a strong tie, then, the likelihood for B and C to have at least a weak tie is increased (so-called transitivity). More concretely, one can expect that if A and B are good friends, and A and C are good friends, at some point in time, A will present B and C to each other (e.g., at a birthday party) and they might become also friends.

The second premise of his theory introduces the logic of bridging ties. This type of tie connects a person A to a person Z, who is not linked to A's other contacts (**Figure 1**). This person Z may provide A with information that is different from what is already communicated in A's other groups. This is because Z's social world does not overlap with the social worlds of A's other contacts, and, hence, is likely to be a distinct social environment with access to different information. In this sense, a bridging tie is seen as a "potential source of novel information" (Borgatti and Halgin, 2011, p. 1171). In an egocentric or personal social network that is reflected by having more separate groups (also referred to as structural holes; Burt, 1992) that would lead ego to possibly get more non-redundant, novel information at any given time (e.g., on the availability of a job offer).

In conclusion, classical network theory argues that less connected networks with many weak ties diffuse novel information, such as behavioral norms and values, faster and more effectively than networks with highly clustered ties. In this view, spread of behavior is understood as a simple contagion via social contact in the network. For example, simple contact with information relating to a score on a volleyball match or the time of a concert might be enough to inform an individual. In this regard, contact with one source is sufficient to change the behavior of one person. Hence, an immigrant's personal social network with many weak ties would facilitate the efficient and fast spreading of culture-related behaviors and norms.

However, this may not always be the case and may depend on what it is that is being diffused. Especially for "costly, risky, or controversial" behaviors "independent affirmation or reinforcement from multiple sources" might be required (Centola and Macy, 2007, p. 703). The contagion then is not simple anymore but complex, because the individual needs to have contact to at least two different sources before credibility is assigned to the received information and a change in behavior is initiated. Especially in the acculturation context, we argue that contact with a single host culture individual is not enough for an immigrant to change host cultural identification. Instead, the immigrant may need repeated contact to several host nationals before a change in cultural identification may be activated. Similarly, a single contact to only one coethnic individual may not be enough to trigger such a change either, but repeated contacted to different coethnic individuals might be. Receiving the same information through repeated contact with different people is more likely in highly clustered networks. Thus, an immigrant's personal social network with many redundant (i.e., strong) ties fulfills the structural conditions to provide the social affirmation and reinforcement mechanisms that are necessary for adapting a change in cultural identification.

The underlying network dynamic of social contagion is influence, which refers to the fact that individuals change their attitudes and behaviors in reaction to their network members. The complementing network dynamic is selection. It describes the process in which people choose their network members and

is usually based on the principle of similarity. Both processes may lead to the same result, namely homophily of network members, and are usually interwoven (Veenstra et al., 2013). In this paper, we do not try to empirically cut this Gordian knot as the data reported in Study 1 is cross-sectional and Study 2 follows directly from its results, although we acknowledge the complexity and endogeneity that the relationship of cultural identification and social networks contains. We rather focus on one possible network dynamic (i.e., influence in form of complex contagions) for the purpose of theory-building.

Drawing on the literature of complex contagions, we argue that, similarly to adopting costly and risky behaviors, cultural identifications are not altered easily though they are dynamic in nature. They may change slowly over time instead of changing dramatically because of one simple contact with a certain culture. Thus, we hypothesize that when immigrants and their descendants develop and negotiate their cultural identifications, they may adopt changes in their identifications as a result of receiving repeatedly culture-relevant information from multiple network members. While, on the one hand, these different network members need to be interconnected to make social affirmation and reinforcement more likely, on the other hand, these network members need to be from different relational domains to enhance the credibility of the information. For this reason, we expect that the interconnection of same ethnicity alters from different relationship domains provides immigrants and their descendants with repeated information from different sources that may alter their cultural identifications in the longrun. Therefore, our main hypothesis is that the interconnection of same ethnicity alters across relational domains is a stronger predictor for cultural identification than the mere amount of alters belonging to a particular ethnic group.

# The Current Research

In the present work, we explored potential relationships between key acculturation variables (i.e., time in the US, ethnic and host cultural identifications, BII), and the content and the structure of personal social networks of immigrants. To do so, we conducted two studies. Using a cross-sectional, correlational design, Study 1 examines survey and personal social network data from 123 Latinos living in the US. The egocentric network data included eight alters: four friends and four colleagues (e.g., classmates, co-workers), thus tapping into two key relational domains to test the following three hypotheses:

H1: The interconnection between friends and colleagues of the same ethnicity is a better predictor for cultural identification than the size of the corresponding ethnic group.

H2: U.S. identification is positively associated with the interconnection of European-American friends and European-American colleagues.

H3: Latino identification is positively associated with the interconnection of Latino friends and Latino colleagues.

The second study builds on the findings of Study 1. Utilizing an ABM data simulation approach, we explore the dynamic ways in which network composition and structure may matter over time in predicting intrapersonal identification change.

# STUDY 1

In Study 1, Latino immigrants and their descendants living in the US were asked to complete a questionnaire about their cultural identifications, their bicultural experiences and their personal social networks.

# Method

## Participants

We relied on a community sample consisting of 123 Latino-American biculturals (41 males, 81 females, 1 transgender), aged 16–65 years (M = 28.5, SD = 9.4; 70.6% with college education or higher), who voluntarily participated in this study. All participants were first- to fifth-generation immigrants living in the US, out of which: 28.5% were first generation, 8.1% 1.5 generation (migration to the US before the age of 16), 25.2% second generation (born in the US, parents

born outside), 14.6% 2.5 generation (born in the US, one parent born in the US, the other parent born outside), 15.5% third generation (parents born in the US), 0.8% 3.5 generation (one pair of grandparents born outside of the US), 3.3% fourth generation (grandparents born in the US), and 4.1% were fifth generation (great-grandparents born in the US). Participants born abroad came mainly from Mexico (53.3%) or El Salvador (20.0%) and had spent on average 11.3 years (SD = 8.0) in the US. Other countries of origin include Brazil (n = 1), Chile (n = 1), Costa Rica (n = 2), Cuba (n = 1), Ecuador (n = 1), Guatemala (n = 1), Nicaragua (n = 2), and Spain (n = 1). The parents of the participants who were born in the US came mainly from the US (mothers: 50%; fathers: 47.4%) or Mexico (mothers: 26.9%; fathers: 25.6%). The other mothers were born in Colombia (n = 3), Cuba (n = 1), El Salvador (n = 6), Guatemala (n = 2), Nicaragua (n = 1), Peru (n = 1), or Puerto Rico (n = 3), and the rest of fathers was born in Colombia (n = 1), Cuba (n = 1), Czechoslovakia (n = 1), El Salvador (n = 6), Guatemala (n = 2), Italy (n = 1), Korea (n = 1), or Puerto Rico (n = 3).

#### Procedure

Participants for Study 1 were recruited at a public "Cinco de Mayo" street festival taking place in downtown San Francisco in 1997. Individuals present at the festival area were either politely approached by the experimenter and her assistant (all of whom were both Latino and Spanish-English bilinguals) or voluntarily came to a booth where a table sign saying "Are you bicultural? Contribute to science and our better understanding of the Latino experience" was displayed. All subjects completed the paperand-pencil survey privately and anonymously and gave written informed consent. The survey requested basic demographic information and included the measures described below. No questions about immigration legal status were asked in the survey.

The study was carried out following ethical guidelines and in accordance to UC Berkeley's Committee for the Protection of Human Subjects (Part VI, B, 3, a, i), which approved the study. The study was completely anonymous, did not include questions of sensitive nature, did not involve deception, and did not pose any anticipated risks to the participants.

#### Instruments

Participants completed a questionnaire that was made available in both Spanish and English, designed to measure the following variables:

#### **Acculturation-related measures**

Cultural identifications. Ethnic and host cultural identifications were measured with two separate items that read "I feel North-American (U.S.)" (U.S. identification) and "I feel Latino/Hispanic" (Latino identification). The response scale ranged from 1 (strongly disagree) to 6 (strongly agree). The average levels of identification were 3.9 (SD = 1.7) for U.S. identification and 5.3 (SD = 1.0) for Latino identification. The correlation between the two identification scales was r = −0.31 suggesting that, at least for this sample, identification as an Bicultural identity integration. Bicultural identity integration was measured with four force-choice items each tapping high versus low BII (e.g., "I combine both cultures" versus "I keep both cultures separate," "I don't feel caught between the two cultures" versus "I feel caught between two cultures"). For each answer option that corresponded to high BII we gave one point, zero if otherwise. The final total score ranged from 0 to 4 (ordinal alpha = 0.68).<sup>2</sup> Given the shortness of this scale (four nonredundant items tapping different facets of identity integration), this alpha is satisfactory. Overall, participants reported a BII mean level of 2.6 (SD = 1.2).

Time in the US. This variable reflects the approximate total amount of years the respondent had lived in the US at the time of the survey. Among second generation participants this variable might very closely reflect the respondent's age minus the time spent outside the US.

#### **Network-related measures**

Participants were first asked to list their four closest friends in California, with whom they had interacted with as personal friends throughout the last year and who were not family members. Second, they named four classmates, co-workers or colleagues in California with whom they had interacted with the most during the last year and who were different from their friends. Participants wrote the initials of the nominated individuals (i.e., alters) in eight circles and were then given the instruction to draw lines among all the individuals who had a relationship (described as having frequent interactions or being friends themselves). As a last step, respondents coded the ethnicity of each alter using the following categories: Latino/Hispanic, Asian, African-American, European/Anglo-American, and other (please specify). From this data, we constructed two variables measuring the networks' composition (who is in the network) and two variables measuring its structure (how are the network members connected).

Group size of Latinos. Group size of Latinos, as a compositional measure, is the absolute count of Latino alters in the network. With a network size of eight alters, the variable may take values between 0 and 8. Overall, participants listed 4.1 (SD = 2.3) Latino alters.

Group size of European-Americans. Likewise, group size of European-Americans refers to the absolute count of alters classified as European/Anglo-American and may range from 0 to 8. Overall, participants listed 2.2 (SD = 1.9) European-American alters.

<sup>2</sup>Zumbo et al. (2007) recommend using ordinal coefficient alpha when estimating the reliability based on Likert response items (i.e., binary and ordinal response scales). Ordinal alpha and Cronbach's alpha are conceptually the same with the difference that ordinal alpha is derived from a polychoric correlation matrix and not from Pearson's covariance matrix. We followed the instructions of Gadermann et al. (2012) for calculating ordinal alpha in the statistical software package R (R Development Core Team, 2011).

Interconnection of Latino friends and colleagues. This structural variable is an indicator for how well Latino friends and Latino colleagues are on average connected to each other weighted in accordance to their group size (variable referred to as inter-class tie weight in Brandes et al., 2008). It is a normalized measure that is based on the idea of the average number of neighbors between two groups. Overall, the weight of how well Latino friends F and Latino colleagues C are connected to each other was 0.4 (SD = 0.6; MAX = 3.2). The weight is calculated as:

$$\alpha(F, C) = \frac{e(F, C)}{\sqrt{|F| \cdot |C|}}$$

Interconnection of non-Latino friends and colleagues. In a like manner, this variable expresses how well non-Latino friends and non-Latino colleagues are connected to each other. Overall, participants' two groups had an interconnection weight of 0.4 (SD = 0.7; MAX = 3.5).

Correlations for all measures are provided in **Table 2**.

# Results

Our main hypothesis was that the interconnection of same ethnicity alters across relational domains would be positively linked to cultural identification. To test this, we ran two separate hierarchical multiple regressions, one set using U.S. identification as a dependent variable and another set using Latino identification as a dependent variable. We calculated three models for each regression. The first model included the acculturation-related variables time in the US and BII as predictors, and their interaction to test for a possible moderation effect of BII. We replaced the five missing cases in the variable time in the US with the overall sample mean to not lose valuable network data. BII had nine missing cases, but as it was operationalized as a composite score, we did not replace them. The reason for this is that for composite scores a big variety of replacement strategies exist, and the choice of one runs the risk of being biased in favor of the researcher's interest. However, this decision led to the reduction of our sample size to 114. In the second model, we added the predictor variables group size of European-Americans for predicting U.S. identification and group size of Latinos when predicting Latino identification. For reasons of multicollinearity, we included only one group size variable at a time. In the final model, both interconnection variables were added.

The results for U.S. identification are shown in **Table 3**. Because of our small sample size, which makes the detection of significant effects difficult, and the fact that social network variables generally tend to show great variation (Brandes et al., 2008), we treat p values below 0.10 as significant (for similar procedure see Mok et al., 2007). Throughout all the models the acculturation-related variables were significant, indicating a strong positive association between U.S. identification with both time in the US and BII. Interestingly, this relationship was stronger for individuals scoring low on BII and lower for individuals scoring high on BII (see left side of **Figure 2**). In line with our expectations, we did not find any effect for group size, but an effect for one of the interconnection variables (H1). However, Hypothesis 2 (a positive link between the interconnection of non-Latino alters across relational domains and U.S. identification) was not supported. Instead, to our surprise, we found a negative link between U.S. identification and the interconnection of Latino friends and colleagues (β = −0.18; p = 0.043). Change in R <sup>2</sup> between the models was not significant, but followed the trend of Hypothesis 1 that network structure is a better predictor than pure network content (from Model 1 to Model 2: 1R <sup>2</sup> = 0.01; p = 0.178; from Model 2 to Model 3: 1R <sup>2</sup> = 0.03; p = 0.124; from Model 1 to Model 3: 1R <sup>2</sup> = 0.04; p = 0.112). According to Akaike's Information Criterion (AIC; DeLeeuw, 1992), in which lower values indicate a better fit, Model 3 (AIC = 417.455) fitted the data best (Model 1: AIC = 417.814; Model 2: AIC = 417.909).

**Table 4** reports the regression results for Latino identification. Similar to the findings above, time in the U.S. and the interaction term were significant predictors, but at lower levels. The negative association between Latino identification and time in the U.S. was stronger for low BIIs (see right side of **Figure 2**). Again, one of the interconnection variables had a greater effect on identification than group size of Latinos when all network variables were included in the model (H1). We found a weak negative link between Latino identification and the interconnection of non-Latino friends and non-Latino colleagues (β = −0.18; p = 0.068) opposed to the hypothesized positive effect of the interconnection variable of Latino alters across relational domains (H3). In general, our models explained more variation in identification with the host culture than with the ethnic culture. Change in R <sup>2</sup> was only marginally significant from Model 1 to Model 2 (1R <sup>2</sup> = 0.02; p = 0.087) and from Model 1 to Model 3 (1R <sup>2</sup> = 0.05; p = 0.094), but not from Model 2 to Model 3 (1R <sup>2</sup> = 0.03; p = 0.176). Model 2 seemed to fit the data best (AIC = 320.277).

# Discussion

Though the results of Study 1 were not very strong, they show that situating and investigating bicultural individuals in their social contexts may be a fruitful approach for understanding the dynamic process of cultural identification. The results suggest that structural aspects of the social context predict patterns of cultural identification better than pure compositional aspects. Specifically, the interconnection between Latino friends and colleagues was linked to lower levels of U.S. identification, while the interconnection of non-Latino friends and colleagues was associated (although more weakly) with lower levels of Latino identification. This culturally inverse pattern of results seems to indicate that, at least for this sample, the social context facilitated cultural identification with the group of interaction by suppressing or lowering the identification with the other culture. Overall, the results suggest a sense of tension or opposition between identifying as an American and as a Latino, leading to the conclusion that those two cultural identifications may be subtractive or oppositional. In particular, external contextual pressures, such as the 1994 California Proposition 187 which prohibited illegal immigrants from using certain public services (e.g., non-emergency health care, public education), might have signaled to the Latino-American biculturals of our study that they

#### TABLE 2 | Correlation matrix for main variables.


1 Latino identification; 2 U.S. identification; 3 time in the US; 4 BII; 5 group size of Latinos; 6 group size of European-Americans; 7 interconnection of Latino friends and colleagues; 8 interconnection of non-Latino friends and colleagues. †p < 0.10. <sup>∗</sup>p < 0.05. ∗∗p < 0.01. ∗∗∗p < 0.001.


N = 114. Reported model coefficients are standardized betas. F/C stands for the interconnection of friends and colleagues. <sup>∗</sup>p < 0.05. ∗∗p < 0.01. ∗∗∗p < 0.001.

cannot be both. Moreover, some Latino groups in the US are highly stigmatized, which may also add to this competing pattern found for both individuals' self-reported cultural identifications and the structure of the social networks that support these identities.

One reason for why our models explained host cultural identification better than ethnic identification might be that Latino identification is less malleable and strongly influenced by variables such as family and child socialization. Bicultural individuals might develop a strong sense of ethnic identity already in the family context and then later add a sense of belonging to the larger host culture. This feeling of belonging to the host society is probably more influenced by what happens and by what immigrants and their descendants do outside the family context, thus encompassing a wider scope of experiences.

Time in the US was associated to both cultural identifications, in that time spent in the US was linked to higher levels of U.S. identification and lower levels of Latino identification. This indicates that as the amount of exposure to and engagement with U.S. culture increases with time, Latino and U.S. cultural identifications become subtractive (for subtractive pattern see de la Sablonnière et al., 2016). Certainly, this pattern seems to be at odds with a bidimensional, two-directional, multidomain definition of acculturation (see Ryder et al., 2000; Flannery et al., 2001; Schwartz et al., 2010). However, an interesting feature of these results is that the subtractive pattern is especially strong among Latinos scoring low on BII, supporting the notion that BII taps into perceptions of cultural incompatibility and conflict.

As the signs in Study 1 for the hypothesized associations between cultural identifications and the interconnection variables (H2 and H3) were different than expected, in the next study we used simulations and an agent-based modeling approach to explore in a dynamic way some of the (static) patterns examined in Study 1. The design of the second study allows us to predict different patterns of intrapersonal change in cultural identification over time based on the composition and structure of this individual's personal network. This more dynamic approach allows us not only to model the effects of social cues and social contexts (i.e., social network composition) on cultural identifications (Hong et al., 2000; Verkuyten and Pouliasi, 2006), but also to model the effects of structural aspects of the social context.

# STUDY 2

Study 2 served two main purposes: (1) to illustrate some of the static results reported in Study 1 in a dynamic way, by modeling intrapersonal change, thus tapping into the dynamic nature of cultural identification; and (2) to explore other possible multiple identities negotiation and management mechanisms. We designed an ABM and simulated data to demonstrate how an immigrant's identification with host and ethnic cultures may change depending on the composition and structure of this individual's personal network. This model may be useful in understanding complex identification outcomes evoked by simple mechanisms based on the principle of influence within networks. It further demonstrates why it might be difficult to



N = 114. Reported model coefficients are standardized betas. F/C stands for the interconnection of friends and colleagues. †p < 0.10. <sup>∗</sup>p < 0.05.

detect consistent and strong network effects with regression analysis utilizing cross-sectional data. Moreover, this model may provide a promising starting point for informing the study of how multiple cultural identities and inter-group relations dynamically interact, and how these processes might lead to the emergence of particular identity structures such as hyphenated identities or identification with a third culture, and perhaps even the development of extreme patterns of cultural identification resulting from the disidentification with either of the other two or even both cultures. Our model allows us to explore the negotiation processes of coexisting identities, conflicting identities, and a mixture of the two. Because most psychologists are unfamiliar with ABM techniques, in the next section, we will briefly describe our model following the standard ODD protocol (Overview, Design concepts and Details) that ensures an easy understanding of the model (Grimm et al., 2006, 2010).

# Method

#### Entities, State Variables, and Scales

When developing our ABM in the NetLogo software (Wilensky, 1999), we tried to stay as close as possible to the operationalization of the variables used in the questionnaire of Study 1. As a consequence, we will describe the model specifically for the context of Latino-American biculturals, though it could be applied to any two cultures. The model consists of two types of entities: one ego and eight alters. Ego is characterized by identification with Latino (ethnic) culture and by identification with U.S. (host) culture, each possibly ranging from 1 to 6. Alters are characterized by their state variables ethnicity (either Latino or European-American) and degree centrality (here the amount of ties with same ethnicity alters, ranging from 0 to 7).<sup>3</sup> The ties

<sup>3</sup> In an ABM, a state variable usually describes an attribute or a property of an agent (e.g., age, sex, size, ethnicity).

among the eight alters are undirected (meaning the alters have symmetric, reciprocal relationships that is, e.g., alters view each other as friends) and distributed randomly, ranging from 0 (alters completely disconnected) to 28 (alters completely connected). Throughout the simulation ego has 50 social interactions, always with one alter at a time. Each social interaction happens at a different point in time and may lead to a change in cultural identification. In that sense, time proceeds in discrete steps, and the length of each time step is not specified further. Composition and structure of the network are held constant through time.

#### Process Overview and Scheduling

Our model includes only one process: change of ego's cultural identifications. Ego's change in identification with Latino and U.S. cultures is traced throughout 50 time steps. At each time step, ego interacts randomly with one of the eight alters and changes level of identification with either one of the cultures, both, or none depending on the implemented rule. In total, we modeled three different mechanisms or rules: positive effect, negative effect and mixed effects (see **Table 5** for a summary).

First, the positive effect describes a mechanism similar to the one we had originally hypothesized in Study 1, namely, that a social interaction with an alter of a particular culture will increase identification with that same culture. Similarly, if ego interacts with an alter of another culture, identification with this new culture increases. As a result, social interaction will always lead to an increase in identification with the culture of the alter ego is interacting with. In this sense, both cultural identifications coexist and are independent from each other. More concretely, whenever ego interacts with a Latino alter a, ego's Latino identification increases by 0.1 + d<sup>a</sup> · 0.1, where d<sup>a</sup> (degree centrality) is the number of ties that alter a has with other Latino alters. Likewise, whenever ego interacts with a European-American alter b, ego's U.S. identification increases by 0.1 + d<sup>b</sup> · 0.1, where d<sup>b</sup> (degree centrality) is the number of ties that alter b has with other European-American alters. Basically, at each social interaction one of the cultural identifications is increased by at least 0.1. We chose 0.1 as it is a basic mathematical unit of change between 0 and 1. The increase in identification is greater than 0.1 when the alter of an interaction has at least one tie to another alter of the same ethnicity. As we wanted to model cultural identification change over time, and, thereby, avoid reaching complete identification too fast (= 6, identification is measured from 1 to 6), we multiplied the degree centrality with 0.1 as a basic unit of change. We used degree centrality to model the idea that the relationships among alters or their social interactions matter for the identification of ego. The more same ethnicity ties an alter has, the greater is the influence on ego's identification. We thus do not only examine the effects of the social context on cultural identification, but also its structural aspects.

Second, the negative effect is based on the actual finding from Study 1 showing that cultural identification with a particular culture decreases when an individual interacts with somebody of another culture. Ergo, social interaction always decreases identification with the culture ego is not interacting with. In that way, engagements with each culture coexist but are not independent from each other. As a consequence, interaction with one culture always leads to a reduction (i.e., suppression or lowering) of identification with the other culture. Precisely, this means that whenever ego interacts with a Latino alter a, ego's U.S. identification decreases by 0.1 + d<sup>a</sup> · 0.1. Similarly, whenever ego interacts with a European-American alter b, ego's Latino identification decreases by 0.1 + d<sup>b</sup> · 0.1. Again, a social interaction has more impact on the change of cultural identification when the alter of the interaction is better connected to other same ethnicity alters.

Third, the mixed effects version of the model is a combination of the former two mechanisms. At each social interaction, positive and negative effects take place simultaneously. In practical terms, this could be when both cultures are seen as coexisting with regards to one life domain (e.g., work values), but as conflicting with regards to a second one (e.g., gender roles). Another addition to the former two versions of the model is the variable α, which regulates the influence of the two effects. This variable ranges from 0 (negative effect is present, but positive effect is absent) to 1 (positive effect is present, but negative effect is absent). Only when α is equal to 0.5 both effects have the same influence on identification. For all other values, either the negative or the positive effect is stronger. Thus, whenever ego interacts with a Latino alter a, ego's Latino identification increases by α(0.1 + d<sup>a</sup> · 0.1) and ego's U.S. identification decreases by (1 − α)(0.1 + d<sup>a</sup> · 0.1). Likewise, whenever ego interacts with a European-American alter b, ego's U.S. identification increases by α(0.1 + d<sup>b</sup> · 0.1) and ego's Latino identification decreases by (1 − α)(0.1 + d<sup>b</sup> · 0.1). In all three versions of the model, ego may change both identities up to a maximum value of 6 and down to a minimum value of 1.

#### Initialization and Simulations

Before simulations began, ethnicity was assigned randomly to the eight alters. So was the distribution of their ties. At the start of each simulation, ego was set up to have moderate identification of 3.5 with Latino and US-American cultures (midpoint of the scales). Each simulation ended after 50 time steps and provided two outcomes: one value for ego's Latino identification and one value for ego's U.S. identification (later referred to as outcome identification). We simulated data by systematically varying the ratio between Latino and European-American alters (i.e., 0:8, 1:7, . . ., 8:0), and the amount of alter ties in steps of four (i.e., 0, 4, . . ., 28), resulting in 72 different combinations each for the positive and for the negative effect. Then, we tried the same combinations with the mixed effects model for α values ranging from 0 to 1 in steps of 0.1. Hence, we had 792 combinations for the third model. As running a model with certain initial values once may show only one possible development and result in only one outcome out of many (Bijak et al., 2013), we ran the simulation of each combination of starting values 50 times. Taking all three models together, we ended up with 46,800 values each for Latino identification and for U.S. identification (9ethnicity ratio · 8alter ties · (1positive + 1negative + 11α) · 50repititions).

## Selected Results

We computed means and their standard deviations for both outcome identifications and for each combination of starting

#### TABLE 5 | Formal comparison of effects.

fpsyg-08-00469 March 29, 2017 Time: 17:22 # 11


ID<sup>A</sup> refers to ego's initial identification with culture A and ID<sup>B</sup> to the initial identification with culture B. D is the degree centrality of the alter ego is interacting with and refers to the number of ties this alter has with other same ethnicity alters. α determines how strong positive and negative effects are.

values considering all 50 repetitions. **Tables 6**, **8** summarize the means of each of the two outcome identifications for positive, negative and mixed effects, respectively, and **Tables 7**, **9** their standard deviations. In the orange graphs, each of the nine orange lines represents a distinct composition of the network (ranging from dark orange with no Latinos and eight European-Americans to light orange with eight Latinos and no European-Americans). The x-axis gives information on the structure of the network (i.e., amount of alter ties), and the y-axis holds the mean of the outcome identification or its standard deviation. In the purple graphs, each of the eight purple lines represents a different network structure (i.e., amount of alter ties; the darker the line, the less alter ties), while the x-axis captures the composition of the network (i.e., amount of Latino alters). Again, the y-axis shows the mean or the standard deviation of the outcome identification given a specific network constellation (i.e., network composition and structure).

To give some examples, in the mean plot of the negative effect model of Latino identification (upper left graph in **Table 6**), the lightest orange line is parallel to the x-axis at 3.5 of the y-axis. This reads as no matter how many alter ties exist, a network composed of eight Latinos always leads to an average outcome identification of 3.5 after 50 runs of the model with 50 time steps. In the purple graph below, all lines have a positive trend. So the more Latinos there are, no matter how many alter ties exist, the higher is the average Latino outcome identification after 50 simulation runs. In addition, alter ties seem to matter the most when there are five or six Latinos. In the standard deviation plot of the negative effect model for Latino identification (upper left graph in **Table 7**), the orange line for six Latino alters approaches a v-shape form and can be read as, no matter how many alter ties exist, a network composed of six Latino alters compared to other network compositions has the widest spread of possible outcomes considering 50 simulation runs. Likewise, the purple graph below shows that no matter how many alter ties exist, the standard deviation of the mean is the highest for six Latino alters. In general, the higher the standard deviation of the mean is, the greater is also the spectrum of possible outcomes after 50 simulation runs or, differently said, the lower is the consistency of simulation outcomes. Next, we present a selection of our results by mechanism or effect.

First, the positive effect model (identical to the mixed effect model of α = 1) always led to complete Latino identification (= 6) when there were at least six Latino alters. Likewise, complete U.S. identification (= 6) was reached when there were at least six European-American alters (identical to a maximum of two Latino alters) (**Table 6**). Second, complete identification with both cultures was also reached, for instance, when there were three to five Latino alters in a completely connected network (= 28 alter ties). Third, having no alters from one ethnicity resulted in the corresponding identification to be stable at 3.5. Fourth, when the ratio between the two ethnic groups was 4:4, the maximum of both identifications was reached or almost reached (smallest identification value 5.816), no matter how many alter ties existed. Fifth, the more alters were connected to each other, the less alters of one ethnicity were needed to result in maximum identification with that ethnicity. Sixth, when having only one alter from one ethnicity, ties hardly mattered for the result of the corresponding outcome identification (value approximately between 4.1 and 4.2; standard deviation relatively stable across ties, **Table 7**). This is because there are no other same ethnicity alters that this alter could have ties with. So the influence of this alter is stable even when the total amount of alter ties in the network increases. In contrast, ties between alters mattered the most when there were two alters of one ethnicity. Then, identification with that particular culture increased with the increase in amount of alter ties. That is because with two alters of the same ethnicity there can be only up to one tie between them, which results in a degree centrality of one for both alters. The more overall alter ties there are, the higher the probability that there is a tie between these two alters. Seventh, the standard deviation of the mean was the greatest for two alters of that ethnicity, no matter how many ties existed, with reaching the maximum when there was a medium amount of alter ties, and

Measurement points are shown as lines for visual simplicity. Some of the lines may be difficult to see as they overlap.

reaching a lower value when there were either no, few, or many ties. However, the standard deviation varied the most for three alters of that ethnicity.

The negative effect model yielded a pattern of results similar to the ones above but mirrored. First, complete disidentification (= 1) with one culture was reached when at least six alters belonged to the other culture no matter how many ties existed (**Table 6**). Second, complete disidentification with both cultures was also reached, for instance, when there were between three and five Latino alters in a completely connected network. Third, the identification outcome with one culture was stable at 3.5 across number of alter ties when all alters belonged to

#### TABLE 7 | Standard deviations of means for negative and positive effects.

Measurement points are shown as lines for visual simplicity. Some of the lines may be difficult to see as they overlap.

that culture. Fourth, when the ratio of the two ethnic groups was 4:4, both identifications reached or almost reached their minimum, no matter how many alter ties were present. Fifth, the more connected to each other alters were, the less alters of one ethnicity were needed to result in minimum identification with the other ethnicity. Sixth, ties mattered the least for the change in identification with one culture when the network was composed of seven alters of that culture (e.g., Latino identification somewhat stable around 2.9 when seven Latinos were present, no matter how many ties). The amount of alter ties mattered the most for the change in cultural identification when there were six alters from the same culture (analogously to having two alters of the same culture in the positive effect). Seventh, the standard deviation of the mean across ties was the greatest for six alters of the traced culture, but varied the most across ties when five of these alters were present (**Table 7**).

Selected results for the mixed effects model are shown in **Tables 8**, **9**. When α is 1, results are the same as in the positive effect model, and when α is 0, results are the same as in the negative effect model. When α is smaller than 0.5, the positive effect is smaller than the negative effect. When α is greater than 0.5, the positive effect is bigger than the negative effect. Only when α is equal to 0.5, both effects are equally important in the

#### TABLE 8 | Means for exemplary mixed effects.

fpsyg-08-00469 March 29, 2017 Time: 17:22 # 14

Measurement points are shown as lines for visual simplicity. Some of the lines may be difficult to see as they overlap.

model. The results of two alphas that complement each other to 1 are mirrored. For example, results for α equal to 0.4 mirror the results of α equal to 0.6.

In the special case where both effects were the same, ties did not or almost not matter for the outcome identification when there were no Latinos, four Latinos or eight Latinos (Latino identification was then 1, oscillating around 3.5 or 6, respectively; inverse for U.S. identification; **Table 8**). In contrast, ties had the most effect on the increase in cultural identification when there were five alters of that culture, and the most effect on the decrease in cultural identification when three alters of that culture were present. Further, the standard deviation of the mean was highest when alters from both cultures were equally present in the network, no matter how many ties existed (**Table 9**). Generally, the standard deviation of the mean decreased the more ties alters had.

For α equal to 0.2 (negative effect stronger than positive effect), having had five alters or less from one ethnicity, decreased the corresponding outcome identification in relation to the initial identification, while six alters or more led to an increase. Next, ties mattered the most in the case of six alters (**Table 8**), and the standard deviation of the mean was highest for five and six alters (**Table 9**). The results for α equal to 0.8 (positive effect stronger than negative effect) are the mirrored results of α equal to 0.2. For an α of 0.8, having had two alters or less from one ethnicity, decreased the corresponding outcome identification in relation to the initial identification of 3.5, while three alters or more led to an increase. Also, ties mattered the most in the case of two alters (**Table 8**), and the standard deviation of the mean was highest for two and three alters (**Table 9**).

## Discussion

To further explore the processes examined in Study 1, in this second study, we modeled the dynamics between social networks and cultural identity, and tested models involving coexisting cultural identities, conflicting cultural identities, and a mixture of the two based on the principle of influence in networks. Keeping

#### TABLE 9 | Standard deviations of the means for exemplary mixed effects.

Measurement points are shown as lines for visual simplicity. Some of the lines may be difficult to see as they overlap.

the number of social interactions constant, we varied the ratio between Latino and European-American alters and the amount of their ties. The influence of the alters on cultural identification was stronger the more same ethnicity alters were connected to each other. We thus did not only model the cultural context of interaction (i.e., interaction with Latino or with European-American alter), but also its structural aspects (i.e., amount of same ethnicity ties that alter of interaction has), by giving importance to the social interactions among alters.

We showed that social network structure and content matter, but not in a homogenous, straightforward way, which might explain why these effects are difficult to detect in regression analyses involving cross-sectional data. Some network constellations may lead to very different results (expressed in a high standard deviation of the mean) depending on which alter ego interacts with; yet, some of these constellations may follow a similar trend. Other, but much fewer, constellations might even be stable in their outcomes. In certain cases, just one additional actor from one culture can make a big difference depending on the number of ties in the identity negotiation process.

This simulation enables us to identify network constellations that lead to complete disidentification with one or both cultures over time. Individuals who do not identify with their ethnic culture nor with their host culture may be of particular interest, as they might develop a sense of belonging to a third culture. While this new culture might be a more inclusive one (e.g., a global culture, a blended culture representing a unique combination of heritage and majority culture), it could also be a more extreme one (e.g., identification with a political or religious radical group). Hence, ABM could, among other things, contribute to the understanding of how acculturating individuals attain radical, extremist identifications (Verkuyten and Yildiz, 2007; Hogg and Adelman, 2013). Additionally, these simulations allow us to detect network constellations in which ego ends up with complete identification with both cultures. Which identity negotiation mechanism takes place may depend on various determinants such as the social, cultural, and political context, psychological characteristics of the individual (e.g., character traits such as dispositional openness and affiliative needs, level of BII) and the perceived or objective similarity of the two cultures.

# GENERAL DISCUSSION

fpsyg-08-00469 March 29, 2017 Time: 17:22 # 16

In the present research, we examined how key acculturation variables (i.e., ethnic and host cultural identifications, BII) relate to the composition and structure of Latino-American biculturals' personal social networks. Drawing on the idea of complex social contagion from network theory and applying it to the negotiation process of multiple identities, we argue that immigrants and their descendants may adopt changes in their cultural identifications as a result of receiving repeatedly culture-relevant information from multiple network members representing different social roles. Relying on a community sample of Latino-Americans, in Study 1, we showed that the interconnection of same ethnicity alters across different relationship domains (i.e., friends and colleagues) predicts cultural identification, while the group size of these ethnicities does not. For these participants, the interconnection of Latino friends and colleagues is negatively associated with U.S.- American identification, and the interconnection of European-American friends and colleagues is negatively linked to Latino identification. This unexpected pattern suggests that, at least for our Latino-American sample, both cultural identifications are embedded in subtractive or even conflicting social network structures. Further, for this sample, time in the US is positively related to U.S.-American culture, but negatively to Latino culture, although this effect (indicative of a subtractive or a zero-sum pattern) is stronger for Latino-Americans who perceive tension between the two cultures (i.e., biculturals low on BII). This interactive pattern lends support to the idea that biculturals who experience low BII (e.g., "I feel caught between two cultures") might manage this feeling by disidentifiying from one of the cultures over time.

While Study 1 was cross-sectional and could only show static interpersonal differences in cultural identification, Study 2 examined intrapersonal changes dynamically. In the latter study, we modeled the dynamics between social networks and cultural identification with both ethnic and host cultures over time, and tested models involving coexisting cultural identities, conflicting cultural identities and a mixture of the two based on the principle of influence. In doing so, we included the cultural context of interaction (i.e., interaction with Latino or with European-American alter) and its structural aspects (i.e., amount of same ethnicity ties that alter of interaction has). We showed that network structure and content matter, but not necessarily in a consistent or homogenous way. Still, we were able to identify network constellations that lead to complete identification or complete disidentification with one or both cultures over time. While certain network constellations may be beneficial for developing harmonious multicultural identities, others may lead to risky patterns of cultural disidentification and radicalization.

We would like to draw attention to some limitations of the current research. First, the data of Study 1 is cross-sectional, which does not allow for any causal inferences. Although we argued that the immigrant's network influences ethnic and host cultural identifications, the reverse (selection) is also possible and likely. Individuals may choose certain people to be part of their network and determine how to connect them depending on their cultural identifications (Veenstra et al., 2013). In the future, longitudinal studies could explore in what way immigrants' cultural identifications determine who becomes a network member and how these network members get connected (selection), and in what way the composition and the structure of the network influence immigrants' identifications with ethnic and host cultures over time (influence). Also, longitudinal data would provide a sequence of at least two observations, which then could be used for designing a stochastic ABM to disentangle the intertwined relationship between selection and influence.

A second possible limitation is the way we measured the network in Study 1. We elicited the network by making respondents draw their networks. As soon as the network structure gets a little bit complicated, this task becomes tricky and people might be more likely to forget relationships between their network members. This is an issue that can be solved easily in future studies by using software especially developed for collecting egocentric network data, such as the program EgoNet, that automatically and separately asks the participants about each possible alter pair, thus facilitating an accurate reporting of all possible connections between alters.<sup>4</sup>

A third limitation concerns the environment of the data collection of Study 1. As the data were collected at a Cinco de Mayo street festival, our Latino community sample is likely to have been biased in favor of immigrants with a strong Latino identification. However, notice that we do not use this data to make empirical claims about Latino multiculturalism and Latino bicultural identity; we rather use the findings from Study 1 to develop a theoretical model in Study 2 that predicts intrapersonal change in cultural identification based on different identity negotiation mechanisms. Future studies could try to balance the cultural setting to also include individuals with lower ethnic identification.

Fourth, sample size, low significance of effects, and the quality and reliability of the scales used in Study 1 may be an issue. The modest size of our sample makes detecting reliable significant effects more difficult. Especially the network variables in Study 1, for which we claim effects on cultural identification for, are of low significance. Nonetheless, we argue that the effects of network composition and structure exist, but are rather hard to show in regression analysis as the process of influence is dynamic and not straightforward, as we illustrate in Study 2.<sup>5</sup>

Fifth, some of our implicit culture-related assumptions might be debatable. For instance, due to logistical and time constrains, we were only able to gather information on the ethnicity of all alters, but not on their cultural identifications. It could be interesting to also include alters' actual cultural identifications as these may be different from their ethnicity. However, egos reported their perceived ethnicity of alters, which might be different from the real ethnicity and closer to observable aspects of the actual cultural identification.

<sup>4</sup>EgoNet is a free software developed by Chris McCarty and Martin Smith which can be downloaded at https://sourceforge.net/projects/egonet/.

<sup>5</sup> In addition, when constructing response scales for items in the questionnaire, response options directly tapping into different strengths of identification levels should be preferred to agree/disagree scales (as in the case of cultural identification) (Saris et al., 2010).

A sixth limitation refers to the model assumptions in Study 2. In real life, not all social interactions are random. While some interactions may be, others may depend on the past. A new model could include the effect of the past on future interactions, for example, by (a) making an interaction with an alter of the same ethnicity as in the previous interaction more likely, or by (b) allowing the interaction to have more influence on identity change if the past interaction was with an alter of the same ethnicity as the alter of the new interaction. Further, actors are not constant over time. They may appear and disappear from a network as ties, too, may evolve and dissolve. Nevertheless, the structure of a network seems to be rather stable even when there is a high turnover in alters, because the way people structure their networks is also affected by their personality (Kalish and Robins, 2006). In addition, even when there is an exchange of actors, certain compositional measures (e.g., percentage of women) remain relatively stable (Lubbers et al., 2010). Finally, future ABMs based on our model should report the average degree centrality of alters by ethnicity and the number of homophilous ties.

Despite all these limitations, our research has also some key strengths. First, Study 1 relied on a community sample rather than a convenience sample (e.g., university students), as is often the case in cultural and social-psychological research. Second, the network approach grasps a real-life situation of intercultural contact in contrast to commonly used self-reports, which are highly dependent on the respondent's self-awareness and are influenced by a variety of biases (e.g., social desirability, wishful thinking, lying about interactions); thereby, the social network approach is a less obstrusive and more implicit data collection mode that yields less danger of being actively manipulated by the respondent. In that sense, we combined individuals' thoughts on acculturation (self-reported cultural identifications) and their acculturative behavior (network). Third, the transference of the theoretical idea of complex social contagion to the negotiation process of multiple cultural identifications is novel. Fourth, in Study 2 we show that individuals' cultural identifications are not only influenced by their contacts, but also by the interactions that these contacts have between each other. Using an ABM approach, we were able to show that a distinct pattern of social relations (i.e., network composition and structure) does not lead to one deterministic identity outcome. Instead, in most of the cases, many different outcomes are probable, although they might follow a similar trend. Only in rare cases the exact outcome can be foreseen. The fact, that network composition and structure may affect multiple cultural identifications in many different ways, and not necessarily in a homogenous manner, might explain why it is difficult to detect network effects in regression analyses involving cross-sectional data. Fifth, experimental research has shown that, in laboratory settings, depending on the available social cues, different cultural frames become salient, and that cultural identification can shift accordingly. Our research does not only contextualize the effects of these social cues in a real-life environment, but also includes structural aspects of it.

Apart from the suggestions already mentioned, future studies could explore other immigrant populations and receiving contexts (e.g., Asian-Americans in the US, Turkish immigrants in Germany), as well as other types of biculturals (e.g., refugees, indigenous, or colonized individuals). Moreover, the boundaries of the network could be defined more openly and include additional relationship domains, such as religious, political, and community enclaves. A mixed-methods research design including network visualizations may allow the respondent to change from being observed to being the observer, and, thus, permit the addition of interpretative information on the network's content, structure and changes over time (Molina et al., 2014).

To conclude, this research contributes to the multiple identities literature, and theory on biculturalism and cultural identity negotiation more specifically, as well as the literature on egocentric social networks, by exploring the links between key acculturation variables (i.e., ethnic and host cultural identifications, BII) and the composition and structure of Latinos' personal social networks in an U.S.-American context. First, our results indicate that the social networks of Latino-American biculturals are related to these individuals' levels of cultural identifications, and that this link is not necessarily based on the composition of the network (e.g., number of Latinos or Americans in the network), as some previous research has shown, but rather on its structure (the interconnection of same ethnicity individuals across different relational domains). Second, this research illuminates the link between degree of exposure to the dominant U.S. culture and ethnic and host cultural identifications by showing that the temporal pattern of a stronger U.S. identification and a weaker Latino identification with the pass of time is particularly prominent among Latinos who perceive their cultural identities as incompatible (i.e., those lower in BII). This finding furthers our understanding of BII and solidifies its validity as a construct to understand how individuals perceive and negotiate multiple cultural involvements over time. Overall, the findings from our two social network studies speak to issues relevant to the integration of immigrants and other cultural minorities, and might be informative in developing intercultural policies and programs that foster both, cohesive social communities and harmonious multicultural identities. In an increasingly multicultural world, this involves the successful inclusion of individuals of different cultural backgrounds into individuals' social networks, and the active prevention of risky patterns of identity disidentification or radicalization (Simon et al., 2013; Lyons-Padilla et al., 2015).

# AUTHOR CONTRIBUTIONS

VB-M contributed to the design of the first study and collected the data. LR contributed to the design of the second study. She performed the data analysis and interpretation of both studies under the supervision of VB-M. LR drafted the manuscript, and VB-M provided critical revisions.

# FUNDING

The work undertaken by LR was supported by the German National Scholarship Foundation.

# REFERENCES


# ACKNOWLEDGMENT

This manuscript benefited greatly from the input provided by Gert Cornelissen, Seth Schwartz, and Thomas Woiczyk.


Kim, Y. Y. (2001). Becoming Intercultural: An Integrative Theory of Communication and Cross-Cultural Adaptation. Thousand Oaks, CA: Sage Publications.


personal network visualizations," in Mixed Methods Social Networks Research. Design and Applications, eds S. Domínguez and B. Hollstein (New York, NY: Cambridge University Press), 305–335. doi: 10.1017/cbo978113922 7193.015


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Repke and Benet-Martínez. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Self-consistency in Bicultural Persons: Dialectical Self-beliefs Mediate the Relation between Identity Integration and Self-consistency

Rui Zhang<sup>1</sup> \*, Kimberly A. Noels<sup>2</sup> , Richard N. Lalonde<sup>3</sup> and S. J. Salas<sup>4</sup>

<sup>1</sup> Department of Psychology, Dickinson College, Carlisle, PA, USA, <sup>2</sup> Department of Psychology, University of Alberta, Edmonton, AB, Canada, <sup>3</sup> Department of Psychology, York University, Toronto, ON, Canada, <sup>4</sup> Department of Theology, Wilfrid Laurier University, Waterloo, ON, Canada

Prior research differentiates dialectical (e.g., East Asian) from non-dialectical cultures (e.g., North American and Latino) and attributes cultural differences in self-concept consistency to naïve dialecticism. In this research, we explored the effects of managing two cultural identities on consistency within the bicultural self-concept via the role of dialectical beliefs. Because the challenge of integrating more than one culture within the self is common to biculturals of various heritage backgrounds, the effects of bicultural identity integration should not depend on whether the heritage culture is dialectical or not. In four studies across diverse groups of bicultural Canadians, we showed that having an integrated bicultural identity was associated with being more consistent across roles (Studies 1–3) and making less ambiguous self-evaluations (Study 4). Furthermore, dialectical self-beliefs mediated the effect of bicultural identity integration on self-consistency (Studies 2–4). Finally, Latino biculturals reported being more consistent across roles than did East Asian biculturals (Study 2), revealing the ethnic heritage difference between the two groups. We conclude that both the content of heritage culture and the process of integrating cultural identities influence the extent of self-consistency among biculturals. Thus, consistency within the bicultural selfconcept can be understood, in part, to be a unique psychological product of bicultural experience.

Keywords: biculturalism, self-concept consistency, bicultural identity integration, naïve dialecticism, East Asian, Latino

# INTRODUCTION

Hundreds of millions of people worldwide now could be considered de facto bicultural (United Nations Department of Economic and Social Affairs, 2013). Broadly speaking, biculturals are people with non-trivial exposure to and participation in at least one culture, including but not restricted to immigrants, ethnic minorities, multiracial individuals, and sojourners (Nguyen and Benet-Martínez, 2007). There are many shades of biculturalism as culture itself comes in a variety of forms (ethnicity, gender, social class, religion, etc.), but most research thus far has focused on

#### Edited by:

Clara Kulich, University of Geneva, Switzerland

#### Reviewed by:

Veronica Benet-Martinez, Pompeu Fabra University, Spain Christina Matschke, Leibniz-Institut für Wissensmedien, Germany

> \*Correspondence: Rui Zhang zhangr@dickinson.edu

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

Received: 11 November 2016 Accepted: 20 February 2017 Published: 07 March 2017

#### Citation:

Zhang R, Noels KA, Lalonde RN and Salas SJ (2017) Self-consistency in Bicultural Persons: Dialectical Self-beliefs Mediate the Relation between Identity Integration and Self-consistency. Front. Psychol. 8:321. doi: 10.3389/fpsyg.2017.00321

immigrants and ethnic minorities (e.g., Asian Americans) or people who are influenced by foreign cultures (e.g., Hong Kong Chinese). Because different cultures tend to contain relatively distinct sets of norms and ideals, a unique challenge befalling biculturals is how to manage the identities associated with their cultures within themselves. Given the profound influence of culture on the self (Markus and Kitayama, 1991), it stands to reason that relating to two cultures should also shape one's sense of self. Thus, focusing on people who are bicultural in the present article, we seek to understand the influence of managing two cultural identities on self-concept consistency, which refers to stable and non-contradictory thoughts and feelings about the self. Rooted in biculturalism research, our guiding framework regards the processes of negotiating cultural identities as being integral to being bicultural (Benet-Martínez and Haritatos, 2005; Tadmor and Tetlock, 2006; No et al., 2011; West et al., under review). In what follows, we link one specific process – individual differences in experiencing cultural identities as more or less integrated within the self – with self-concept consistency. We will show that a consistent self-concept can be understood, in part, as the unique psychological product of an integrated bicultural identity. Importantly, we argue that the experience of having an integrated bicultural identity should affect self-consistency to a similar extent across bicultural people of various ethnic heritages that nonetheless differ in terms of emphasis on the need to maintain self-consistency.

# Bicultural Identity Integration and the Bicultural Self-concept

Living in two cultural worlds requires biculturals to manage cultural complexity both socially and psychologically. Socially, being bicultural entails developing relationships with two relatively distinct communities, thus calling for effective communication with both groups and a wider linguistic or behavioral repertoire (LaFromboise et al., 1993; Noels, 2013). Psychologically, being bicultural creates accountability pressures to more than one cultural audience (Tadmor and Tetlock, 2006). The existence of different or even colliding cultural norms and modes of communication may turn inward, resulting in the need to reconcile cultures within biculturals themselves. The construct of bicultural identity integration was developed to capture individual differences among biculturals in their subjective experience of relating to both cultures (Benet-Martínez and Haritatos, 2005). Specifically, biculturals vary in how their bicultural identity is organized within the self-concept: some experience their cultural identities as harmonious or blended, whereas others experience their cultures as oppositional or compartmentalized (Benet-Martínez and Haritatos, 2005; Huynh et al., 2011). This line of research has provided extensive evidence clarifying the dimensionality, social antecedents, and psychological consequences of having a more or less integrated bicultural identity (for a recent review, see Cheng et al., 2014).

Much of the research on the bicultural self-concept has been approached from a cultural frame switching perspective. Cultural frame switching is the experience of responding to cues of cultural identity (e.g., language, cultural icons) and applying appropriate cultural knowledge and behaviors (Hong et al., 2000; also see, Chiou and Mercado, 2016). To the extent that different cultures foster contrasting senses of self (e.g., introversion vs. extraversion, Chen and Bond, 2010; selfenhancement vs. self-effacement, Lee et al., 2010), biculturals can, in principle, switch between these different self-aspects. In a similar vein, biculturals also shift their momentary feelings of cultural identification and their situated cultural identities are typically attuned to whichever culture is salient in the immediate social situation (Clément and Noels, 1992; Zhang and Noels, 2013; Noels and Clément, 2015). Research on these fronts suggest that because biculturals may be particularly prone to showing context-dependent self-views, they might be less consistent in their overall self-concept. As will be detailed below, there is a longstanding tradition in examining the ways in which the self maintains consistency in personality psychology and more recent research has shifted its focus to investigating cultural influence on self-concept consistency. Despite the common interest, however, there has been surprisingly little theoretical or empirical intersection between the two research traditions. There is virtually no research examining self-consistency among biculturals; nor is there research linking bicultural identity integration specifically to self-consistency. The present article aims to fill in this important knowledge gap.

# Self-concept Consistency and Naïve Dialecticism

Self-concept consistency or consistency within the self has figured prominently in influential perspectives on self and identity (e.g., Festinger, 1957; Block, 1961). Ever since the seminal treatises on the cultural grounding of the self (Markus and Kitayama, 1991; Shweder, 1991), however, extensive research has shown that culture moderates the importance of self-consistency. One wellstudied form of self-consistency is consistency across situations or roles, the sense that one possesses a stable set of self-defining attributes that transcends context. Relative to North Americans and West Europeans, East Asians describe themselves more differently in the presence of different people (Kanagawa et al., 2001) and report more differentiation between their general self and their role-specific selves (e.g., self as a close friend or a son/daughter; Suh, 2002; Church et al., 2008). Another form of self-consistency is internal consistency of the self, which is the perception that one's global sense of self is internally coherent and non-contradictory. East Asians are also more likely to harbor contradictory self-views or ambivalent self-evaluations simultaneously (e.g., being both extraverted and introverted; Choi and Choi, 2002; Spencer-Rodgers et al., 2004, 2009).

Not only does the cultural psychological perspective reveal cultural differences in self-consistency, it has also contributed to understanding their specific causes. A substantial amount of research indicates that cultural variations in self-consistency reflect variations in the general tendency to expect changes and tolerate contradiction in the world, which has been termed naïve dialecticism (for a review, see Spencer-Rodgers et al., 2010). Naïve dialecticism is a lay belief system that originated in East Asian (e.g., Chinese, Japanese, and Korean) philosophical traditions

and consists of three core principles: (1) what is constant in life is change, (2) everything contains within itself its opposite, and (3) everything in the universe is intricately interconnected (Peng and Nisbett, 1999). In terms of acquiring knowledge, the East Asian dialectical approach questions the existence of invariant and absolute truths. This contrasts sharply with the synthesis orientation in the Western intellectual traditions that prioritizes the pursuit of unchanging truths that transcend opposing propositions (Peng and Nisbett, 1999).

Several lines of evidence support the role of naïve dialecticism as a proximal mechanism that explains cultural differences in consistency-related cognitions and emotions, including selfconcept consistency that is of interest here. First, multi-national comparisons have shown that cross-national differences in crossrole consistency could not be accounted for by individualismcollectivism, but by naïve dialecticism (Church et al., 2008). This is because dialectical thinking appears to be unique to East Asian cultures and people from non-dialectical, collectivistic cultures (e.g., Mexico) are no less consistent than those from non-dialectical, individualistic cultures (e.g., Australia; for related evidence, see Schimmack et al., 2002; Scollon et al., 2005). Second, cross-national differences in self-consistency were replicated among ethnic groups within the United States. Asian Americans showed greater inconsistency across roles (English and Chen, 2007) than did European Americans. Asian Americans also showed more ambivalent (hence less consistent) global selfevaluations relative to European, Latino, and African Americans (Spencer-Rodgers et al., 2004). Last, direct evidence for the causal role of dialectical thinking in self-consistency can be seen from studies showing that cultural differences were mediated by dialectical self-beliefs (English and Chen, 2007; Boucher, 2011) and from experimental studies where dialectical thinking was manipulated (Spencer-Rodgers et al., 2004). Therefore, naïve dialecticism provides the most parsimonious explanation for cultural variations in self-consistency.

# The Present Research

The overarching goal of this article is to integrate key insights from research on both biculturalism and naïve dialecticism in the study of two forms of self-consistency among biculturals: consistency across roles and global consistency. Prior research on self-consistency with bicultural samples focused on the influence of heritage culture (Spencer-Rodgers et al., 2004; English and Chen, 2007), while paying scant attention to the possible effect of their unique bicultural experience. Biculturals may exhibit psychological tendencies that cannot be understood solely by examining their heritage cultures; these tendencies can also be understood by examining the processes that biculturals use to negotiate their cultural identities (Tadmor et al., 2012; Saad et al., 2013; West et al., under review). As such, the degree of self-consistency among biculturals may be driven by both how dialectical their heritage cultures are and how they negotiate their cultural identities.

Focusing on bicultural identity integration, our first hypothesis states that having an integrated bicultural identity would be associated with being more consistent across roles and having a more consistent global self-concept (H1). We formulated H1 for the following reasons. First, some prior work provides indirect supportive evidence. In studies of Latino American biculturals, identity integration was associated with a stronger perceived overlap between their personality and both typical Latino and American personality profiles (Miramontez et al., 2008). This suggests that by perceiving themselves as broadly compatible with both cultures, more integrated biculturals achieve more coherence within their self-concept. In another study of how Chinese Americans represented their cultures, less integrated Chinese Americans were found to provide more complex cultural representations (Benet-Martínez et al., 2006). Although the results are less tied to self-concept, we can deduce that a richer differentiation of cultures might translate into a more compartmentalized and internally tensioned self-concept, hence lower self-consistency. Thus, previous studies hint at identity integration leading to more self-consistency. More importantly, we believe there is a key conceptual reason why this may be the case. To integrate cultural identities presupposes the recognition of at least the potential for cultural conflict and consequently the need to resolve contradictions that arise from belonging to two cultures. From the non-dialectical perspective, integration is the ideal solution to possessing contradictory self-aspects as it draws attention to broader similarities and the bigger picture (Amiot et al., 2007). Contrarily, the dialectical perspective encourages a mere acceptance rather than resolution of contradiction within the self or in the social world (Peng and Nisbett, 1999; for related empirical work, see Chen et al., 2013). Managing to integrate cultural identities should thus be associated with higher self-consistency.

Our second hypothesis states that bicultural identity integration would affect cross-role and global self-consistency indirectly through dialectical self-beliefs (H2). **Figure 1** displays the conceptual mediation model. Based on the review of the literature provided earlier, dialecticism has been linked with both self-consistency and bicultural identity integration. On the one hand, given the evidence that dialecticism explains cultural variations in self-consistency, dialecticism can be conceived as a sufficient condition for reliably producing self-inconsistency. In the only published studies examining the effect of dialecticism on self-consistency among biculturals, dialecticism was found to predict variations in personality across contexts involving the use of different languages (Chen et al., 2014). On the other hand, because integration is an effective way of reducing contradiction (Peng and Nisbett, 1999), success in integrating cultural identities should reduce the propensity of using dialectical thinking. On the flip side, having difficulty with reconciling cultural identities could increase the general tendency to expect and even tolerate contradiction. In support of this, negative associations between these two variables were found in biculturals in Hong Kong and China (Chen et al., 2013). Therefore, bicultural identity integration was hypothesized to negatively predict dialectical thinking, which in turn would negatively predict self-consistency. It is worth mentioning that because a bicultural person's level of dialectical thinking is clearly influenced by their heritage culture, our mediation model only specifies bicultural identity integration as one of the factors affecting dialectical thinking.

It is also important to note that given the correlational nature of the studies we drew upon (Chen et al., 2013), it may well be that identity integration is the result, as opposed to the cause, of dialectical thinking. We will return to the issue of directionality of the mediation model in the section "General Discussion."

Our third and final hypothesis concerns sharpening the distinction between our predictions regarding the influence of bicultural identity integration on self-consistency and the previously documented ethnic differences among bicultural groups in self-consistency. As mentioned above, different bicultural groups differ in self-consistency because their heritage cultures differ in the extent of socializing naïve dialecticism. Thus, Asian Americans tend to have a more ambivalent self-concept relative to Latino Americans (Spencer-Rodgers et al., 2004). Given our focus on common bicultural experience, we think bicultural identity integration should affect self-consistency to a similar degree across a variety of bicultural groups. For instance, both more integrated Asian biculturals and more integrated Latino biculturals should show higher self-consistency. However, this effect of bicultural identity integration is analytically distinct from the effect of heritage culture. While more integrated Asian and Latino biculturals should show more self-consistency, this need not eliminate the average self-consistency difference between the two groups; this difference is driven largely by ethnic difference in the dialectical vs. non-dialectical tradition. Hence, our third hypothesis is as follows: biculturals of dialectical ethnic heritage would show lower self-consistency relative to those of non-dialectical ethnic heritage (H3). This hypothesis represents a conceptual replication of previous findings.

In the following, we present four studies designed to test the three hypotheses (Study 1: H1; Study 2: H1, H2, and H3; Study 3: H1 and H2; Study 4: H1 and H2). In all studies, we defined biculturals in terms of cultural exposure. By this definition, biculturals are those with involvement in two cultures (e.g., ethnic minorities), regardless of their subjective identification with them (Nguyen and Benet-Martínez, 2007). Such a demographically based definition allows us to maximize the number of bicultural participants that could be included in each study. Another issue is the dimensionality of bicultural identity integration. We followed the widely adopted distinction between the cognitive and the affective components of integration (Huynh et al., 2011). While the cognitive dimension captures whether the bicultural identity is represented jointly or separately in the self, the affective dimension reflects if the bicultural identity feels congruent or tensioned. The two dimensions are both conceptually and empirically distinct (Huynh et al., 2011). In all studies, we assessed both dimensions and explored whether there would be any difference in terms of their associations with self-consistency.

# STUDY 1

We conducted our first study to seek preliminary support for the hypothesized association between bicultural identity integration and consistency across roles (H1). We recruited a Chinese Canadian sample and assessed the two dimensions pertaining to their bicultural identity integration (Benet-Martínez and Haritatos, 2005) and how similarly they rated their personality in the context of close friendship that typically involves two groups of cultural audiences (Tadmor and Tetlock, 2006).

Ethical approval for all the studies presented here was obtained from the following ethics board: Research Ethics Office at the University of Alberta (Studies 1–3) and Office of Research Ethics at York University (Study 4). In all studies, participants gave informed consent by signing or checking a box next to the line that said "I consent."

# Method

#### Participants

Participants were 81 Chinese Canadians from introductory psychology classes (70% women, Mage = 19.23, SDage = 1.71). As citizens or permanent residents, about half (54%) of them were born in Canada; the mean length of residence among those foreign-born was 10.63 years (SD = 3.94)<sup>1</sup> .

#### Measures

In small groups, participants arrived at a computer lab and completed an online survey that assessed bicultural identity integration, self-concept consistency across roles, and demographic questions. They also completed other measures

<sup>1</sup> In all studies, we did not find any systematic generational difference in the patterns of the results. Given the relatively small sample sizes of each generation in some of the studies, however, we cannot rule out this possibility, which future research could examine (e.g., Zhang and Noels, 2013).

developed for a different study reported elsewhere (Zhang et al., 2016, Study 1). More details regarding the measures appear below.

#### **Bicultural identity integration**

fpsyg-08-00321 March 6, 2017 Time: 17:58 # 5

Participants completed the Bicultural Identity Orientation Scale (BIOS; Comanaru et al., under review) by rating their agreement with each statement on a 6-point scale (1 = strongly disagree, 6 = strongly agree). We focused on the two subscales most germane conceptually to identity integration: the extent to which one feels tension between their two cultures (conflict, four items; e.g., "I have difficulty reconciling the differences between my ethnic culture and the Canadian culture"; α = 0.70), and the extent to which one perceives oneself to be a blended product of two cultures (hybridity, four items; e.g., "I feel my identity is a hybrid of two cultures"; α = 0.79). The lower the conflict scores, the higher the hybridity scores and the more integrated one's bicultural identity is.

#### **Consistency across roles**

We adopted the Trait-Role Questionnaire that has been widely used in cross-cultural research investigating self-concept consistency (e.g., Church et al., 2008). It consisted of 40 trait adjectives tapping the Big Five dimensions (Extraversion, Agreeableness, Conscientiousness, Emotional Stability, and Openness to Experience), with each dimension indexed by eight items (see Church et al., 2008). Using a 7-point scale ranging from very uncharacteristic of me to very characteristic of me, participants then rated the extent to which the same 40 Big Five traits described them while being with a close friend who is European Canadian and while being with a close friend who is Chinese Canadian. That is, they provided contextualized personality ratings in two friendship contexts. To facilitate that, they were first asked to write down the initials of each friend and think about their relationship with the person. If they could not think of a close friend, they were instructed to write down the initials of a casual friend instead. The order of the friendship contexts and the traits within each friendship was randomized for each participant.

We operationalized self-concept consistency in terms of profile similarity, that is, the extent to which an individual's personality profile in one role is similar to his or her personality profile in another role. In our case, we computed

TABLE 1 | Descriptive Statistics from Study 1 to Study 3.

within-participant correlations between the 40 self-with-Chinese-Canadian-friend personality ratings and the 40 self-with-European-Canadian-friend personality ratings. A larger correlation between the two personality profiles indicates being more consistent across roles.

# Results and Discussion

**Table 1** lists the descriptive statistics for all the variables. As predicted, having a more integrated bicultural identity was associated with being more consistent across roles. Specifically, after the r-to-Z Fisher transformation was performed on crossrole consistency, it was correlated positively with hybridity (r = 0.31, p = 0.005) and negatively with conflict, albeit non-significantly (r = −0.18, p = 0.10). Experiencing cultural identities as discrepant with each other was more strongly associated with the perception that one acts differently with friends associated with each culture. Thus, Chinese Canadians' internal representation of cultures appears to be manifested outwardly in the form of personality perception in the close friendship domain. It was also observed that thinking rather than feeling that one's bicultural identity is integrated – the cognitively based aspect – mattered more. Because we did not predict this finding, we sought to replicate it first in other bicultural samples.

# STUDY 2

Study 2 was designed to extend it in a number of ways. First, given that people in East Asian cultural contexts tend to be more dialectical, it is unclear whether the finding was unique to only biculturals of dialectical heritage background. Thus, in addition to an East Asian bicultural sample, we recruited a Latino bicultural sample as people in Latin American cultural contexts have been shown to be collectivistic but are equally synthesis-oriented as people in European American or Canadian cultural contexts. We expected bicultural identity integration to be associated with consistency across roles with each bicultural sample. Second, we broadened the friendship context assessed in the consistency measure to encompass another common interpersonal context: interactions with one's parents. That is, we examined consistency between self-as-a-friend and self-as-ason/daughter. Third, we also assessed dialectical beliefs to test H2


DSS, dialectical self scale. <sup>a</sup>Values in this row represent correlation coefficients.

that identity integration affects consistency across roles indirectly via dialectical self-beliefs. Finally, we tested H3 that ethnic differences in self-concept consistency between the two bicultural samples should mirror differences between dialectical and non-dialectical cultures. As explained above, because different bicultural groups are socialized into different heritage cultures, their bicultural experiences might not eliminate between-group differences due to the influence of heritage cultures. Because previous research shows people of East Asian cultural heritage to be more dialectical and less consistent relative to people of Latino cultural heritage (Schimmack et al., 2002; Spencer-Rodgers et al., 2004; Church et al., 2008), we expected East Asian biculturals to be less consistent across roles, as a whole, than Latino biculturals. We also recruited a third sample as a prototype group of monocultural Canadians: European Canadians who were born in the country. Although not bicultural as commonly defined, European Canadians share the synthesis-oriented ethnic heritage with Latino Canadians (Schimmack et al., 2002; Spencer-Rodgers et al., 2004; Church et al., 2008). If it were true that ethnic differences in self-consistency are largely driven by dialecticism, both Latino and European Canadians should be more consistent relative to Asian Canadians. These ethnic differences, in turn, should be mediated by dialectical self-beliefs.

# Method

#### Participants

Participants were 65 European Canadians (68% women, one unspecified), 101 East Asian Canadians (55% women, one unspecified), and 70 Latino Canadians (61% women, one unspecified). Both European and East Asian samples as well as approximately half of the Latino sample consisted of undergraduates enrolled in introductory psychology courses who received partial course credit. The rest of the Latino sample consisted of students recruited via campus flyers or snowball sampling and they received \$10 for participation. As a result of this recruitment strategy, the Latino participants were on average older and more variable in age (M = 22.16, SD = 5.85), compared with their European (M = 19.09, SD = 1.67) and East Asian (M = 18.61, SD = 1.65) counterparts. Hence, in the following analyses concerning group comparisons, we controlled for age. All participants were Canadian citizens or permanent residents. All European, 65% of the East Asian, and 47% of the Latino participants were born in Canada, respectively. Of those foreign-born, the mean length of residence was 9.37 years (SD = 4.49) and 9.32 years (SD = 5.01) for East Asian and Latino Canadians, respectively.

#### Measures

Upon arrival in a lab, participants filled out a survey (either paper-and-pencil or on a computer) comprised of the following measures and demographic questions.

#### **Bicultural identity integration**

It was assessed with the same BIOS conflict and hybridity subscales as used in Study 1 (Comanaru et al., under review). Only the two bicultural groups completed them. The αs were 0.71 (conflict) and 0.80 (hybridity) for the East Asian Canadians; 0.72 (conflict) and 0.76 (hybridity) for the Latino Canadians.

#### **Consistency across roles**

Using an 8-point scale, all participants rated the extent to which the same 40 Big Five traits described them in two interpersonal contexts. Because we were interested in consistency across common roles people assumed, not just roles that may have specific meaning to biculturals, we broadened the close friendship context to include being with one's parent and a friend, the closeness of which was not defined. Again, the traits were presented in different orders for each role and the order of the roles was randomized for each participant. To index consistency across roles, we computed with-participant correlations between the two sets of the contextualized personality ratings.

#### **Dialectical self-beliefs**

All participants completed the 32-item Dialectical Self Scale (DSS; Spencer-Rodgers et al., 2015, Unpublished) that assesses dialectical beliefs in the domains of contradictions (e.g., "When I hear two sides of an argument, I often agree with both"), cognitive change (e.g., "I often find that my beliefs and attitudes will change under different contexts"), and behavioral change (e.g., "I often change the way I am, depending on who I am with"). Responses were recorded on a 7-point scale (1 = strongly disagree, 7 = strongly agree). The αs for the overall DSS ranged from 0.72 to 0.80. Because previous work has used either a specific subscale or the overall scale to explain cultural differences in self-concept consistency (English and Chen, 2007; Church et al., 2012), we explored the effects of both overall DSS and its subscales.

# Results and Discussion

#### Within-Group Results

In support of H1, bicultural identity integration was correlated with consistency across roles in each bicultural group. Specifically, the r-to-Z Fisher transformed consistency index was correlated positively with hybridity (East Asian: r = 0.20, p = 0.047; Latino: r = 0.11, p = 0.38) and negatively with conflict (East Asian: r = −0.17, p = 0.10; Latino: r = −0.26, p = 0.03). Interestingly, consistency was more strongly associated with the cognitive dimension of integration for East Asian biculturals, which is in line with the Study 1 finding, whereas it was more closely associated with the affective dimension for Latino biculturals.

To test H2 concerning the mediation hypothesis, we first established if DSS would be correlated with both bicultural identity integration and consistency within each group. Among the East Asian biculturals, only behavioral DSS was correlated with hybridity (r = −0.25, p = 0.013) and consistency (r = −0.40, p < 0.001). We then proceeded to test mediation with Preacher and Hayes' (2008) bootstrapping method. Five thousand bootstrapping resamples were performed. The biascorrected 95% CI for the indirect effect of hybridity on consistency via the mediator of behavioral DSS did not contain zero [0.005, 0.09], providing support for H1. In a regression model, the total effect of hybridity on cross-role consistency was reduced from β = 0.20, t = 2.01, p = 0.047 to β = 0.11,

t = 1.13, p = 0.26 once behavioral DSS was introduced to the model. Among the Latino biculturals, overall DSS was correlated with conflict (r = 0.57, p < 0.001), hybridity (r = −0.30, p = 0.013), and consistency (r = −0.44, p < 0.001). Given the contemporary perspective that a significant overall effect is no longer prerequisite for testing mediation (MacKinnon et al., 2002; Shrout and Bolger, 2002), we proceeded to test the indirect effect of hybridity on cross-role consistency via dialectical beliefs as well, even if the total effect was non-significant. With 5,000 bootstrapping resample, the bias-corrected 95% CI for the indirect effect of conflict on consistency via DSS did not contain zero [−0.17, −0.04]. In a regression model, the total effect of conflict on cross-role consistency diminished from β = −0.26, t = −2.22, p = 0.03 to β = −0.02, t = −0.13, p = 0.90 once overall DSS was introduced to the model. Similarly, the bias-corrected 95% CI for the indirect effect of hybridity on consistency via DSS did not contain zero [0.02, 0.12]. Both results are in line with H2.

To summarize, we replicated the overall positive association between bicultural identity integration and consistency among both bicultural groups, providing further support for H1. A more integrated bicultural identity (in the form of higher hybridity and lower conflict) was related to a more consistent sense of self between being with one's parent and being with one's friend. Of note, the two bicultural groups differed in which aspect of identity integration was more strongly linked with cross-role consistency. The East Asian biculturals are akin to the Chinese biculturals from Study 1 in that cross-role consistency was better predicted by how they perceived their cultural identities to be integrated, whereas how consistent Latino biculturals were across roles was more driven by their affective reactions. Furthermore, we found evidence consistent with H2 that posited dialectical beliefs as one mechanism through which identity integration influences self-consistency. Judging by differences in the strength of associations among those variables, dialectical beliefs appear to be more causally proximal to cross-role consistency relative to either hybridity or conflict. In other words, the relatively weak associations between bicultural identity integration and crossrole consistency could be an indication that the former exerts a more distal causal influence. Finally for Latino biculturals, dialectical beliefs were suggested as both a pathway from conflict to cross-role consistency and a pathway from hybridity to crossrole consistency. The latter was true despite the non-significant overall effect. The juxtaposition between a non-significant total effect and a significant indirect effect suggests that compared with the affective dimension, the cognitive dimension of identity integration might be a more distal causal agent of cross-role consistency for Latino biculturals.

#### Between-Group Results

To test H3, we first compared group differences in Fisher r-to-Z transformed consistency across roles by conducting a one-way analysis of covariance (ANCOVA) with ethnicity (European, East Asian, or Latino) as the grouping variable while controlling for age. It should be noted that results remained the same if age was not controlled for but we opted to adjust its effect because the Latino sample was older as stated before. **Table 1** lists unadjusted, raw means and standard deviations for each group. Results revealed significant group differences, F(2,230) = 10.51, p < 0.001, η 2 <sup>p</sup> = 0.08. Post hoc pairwise comparisons were conducted with Bonferroni correction. As predicted, both European and Latino Canadians were more consistent across roles than the East Asian Canadians, p = 0.002, 95% CI [−0.38, −0.07]; p < 0.001, 95% CI [−0.45, −0.12] respectively, while the former two groups did not differ from each other, p = 1.0. Next, another one-way ANCOVA comparing group differences in DSS while controlling for age showed a significant effect, F(2,230) = 11.77, p < 0.001, η 2 <sup>p</sup> = 0.09. Pairwise comparisons with Bonferroni correction were again in line with H1. East Asian Canadians endorsed dialectical beliefs more than European Canadians, p < 0.001, 95% CI [0.17, 0.58], and Latino Canadians, p = 0.001, 95% CI [0.11, 0.55]. There was no difference between the latter two, p = 1.0.

In view of the ethnic differences in dialectical self and self-concept consistency, we proceeded to test a mediation model whereby dialectical self was posited to account for the group differences in consistency. We examined this indirect effect with the MEDIATE bootstrapping procedure developed by Hayes and Preacher (2014), which allows for an omnibus test of mediation involving multiple groups. In the mediation model, the predictor was comprised of two dummy-coded ethnicity variables (East Asian = 0, European = 1, Latino = 0; East Asian = 0, European = 0, Latino = 1). The consistency scores were then regressed onto the dummy-coded ethnicity variables with DSS entered as the mediator and age as the covariate. DSS significantly predicted consistency, B = −0.27, p < 0.001, while the omnibus effect of ethnicity was reduced from F(2,230) = 10.51, p < 0.001 to F(2,229) = 4.57, p = 0.01. Specifically, the previously significant East Asian-European and East Asian-Latino differences became less significant, B = 0.13, p = 0.04; B = 0.19, p = 0.004, respectively. In support of mediation, the bias-corrected 95% confidence interval (CI) constructed with 5,000 resamples did not contain zero [−0.05, −0.007], suggesting a significant omnibus effect of mediation. More specifically, DSS significantly mediated both the East Asian-European difference, 95% CI [0.05, 0.17], and the East Asian-Latino, 95% CI [0.04, 0.16].

To summarize, H3 received full support. First, we replicated findings from previous research showing ethnic differences between Asian and European Americans in cross-role consistency (English and Chen, 2007). Second, the East Asian-Latino difference in cross-role consistency dovetails with other research that examined national differences between East Asian and Latino populations in similar domains (Schimmack et al., 2002; Church et al., 2008). Finally, the mediation results clarify the causal role of dialecticism in producing the ethnic differences above. All in all, while the within-group results reveal the working of bicultural experience, the between-group results confirm the imprint of cultural heritage on self-consistency. Importantly, the within-group dynamic does not seem to eliminate the between-group differences in self-consistency, suggesting identity integration and heritage enculturation as two relatively distinct processes that shape one's level of self-consistency.

# STUDY 3

fpsyg-08-00321 March 6, 2017 Time: 17:58 # 8

In Study 3, we examined the generalizability of Study 2's key findings in an ethnically heterogeneous bicultural sample. Specifically, we aimed to further replicate (1) the association between bicultural identity integration and consistency across roles; (2) the mediating role of dialectical self-beliefs.

# Method

#### Participants

Participants were 163 Canadians of non-European descent from introductory psychology classes (58% women, one unspecified; Mage = 18.76, SDage = 1.90). Slightly more than half (58%) identified their cultural heritage as East Asian, followed by South Asian (27%), Middle Eastern (7%), African (6%), and Latino (2%). About half (48%) were Canadian-born, with the rest reporting their mean length of residence to be 9.60 years (SD = 4.59). The original sample included two international students, who were dropped from the analyses reported herein.

#### Measures

In small-group settings, participants completed a survey that contained the same measures as in Study 1. The respective αs were: conflict = 0.78, hybridity = 0.82, and DSS (overall) = 0.79, DSS (contradiction) = 0.67, DSS (cognitive) = 0.70, DSS (behavioral) = 0.50.

# Results and Discussion

**Table 1** lists the descriptive statistics of all the variables. The r-to-Z Fisher transformed consistency index was correlated negatively with conflict (r = −0.21, p = 0.006), but not hybridity (r = 0.06, p = 0.47). Because only cognitive DSS was correlated with both conflict (r = −0.18, p = 0.02) and cross-role consistency (r = −0.19, p = 0.02), we tested mediation via cognitive DSS with 5,000 bootstrapping resamples. The bias-corrected 95% CI did not contain zero [−0.03, −0.0006]. The total effect of conflict was reduced from β = −0.21, t = −2.77, p = 0.006 to β = −0.19, t = −2.39, p = 0.02 once cognitive DSS was controlled for.

Those results depart somewhat from the specific findings from previous studies. Given that about half of the current sample were of East Asian heritage, that conflict but not hybridity was significantly associated with cross-role consistency contradicts what was found in both Studies 1 and 2. However, the core tenets of H1 and H2 were largely borne out. One of the identity integration dimensions was related to self-consistency, although it is less clear which dimension was more causally proximal. Furthermore, dialectical self-beliefs, especially components that are most conceptually germane (cognitive or behavioral DSS), mediated the effect of identity integration on self-consistency.

# STUDY 4

In the final study, we sought converging evidence for H1 and H2 by extending the previous studies in two ways. The most important change was to shift attention to another aspect of self-concept consistency: global consistency. A consistent sense of self entails not only a core set of self attributes that is stable from situation to situation but also an overall perception that one possesses coherent self-knowledge and unambiguous feelings toward the self. The latter is often understood as internal consistency of the self. By extension, a dialectical selfconcept should manifest itself in the form of exhibiting internally inconsistent judgment about the self (Choi and Choi, 2002; Hamamura et al., 2008; Spencer-Rodgers et al., 2009). Thus in this study, we tested H1 and H2 with measures tapping global self-consistency, which we operationalized as maintaining a consistent judgment of one's general personality and a noncontradictory self-evaluation. Second, because the majority of our bicultural sample in Study 3 were of Asian descent, we recruited another bicultural sample from the Toronto metropolitan area to further increase ethnic diversity. Toronto is home to the largest number of immigrants and ethnic minorities, making it the most ethnically diverse urban area in Canada. For example, ethnic minorities account for about 47% of Toronto's total population (Statistics Canada, 2011).

# Method

#### Participants

Our sample included 186 introductory psychology students of non-European descent at a Canadian university in Toronto (76% women, one unspecified; Mage = 21.15, SDage = 3.73). As such, cultural heritages of this sample were more heterogeneous than previous studies: 37% South Asian, 20% Middle Eastern, 17% Black, 10% East Asian, 7% bi- or multi-racial, 5% Latino, and 4% other. While less than half (42%) were Canadianborn, those who were foreign-born had spent an average of 10.75 years (SD = 5.44) in Canada. Because the participants completed the survey online, we interspersed it with attention checks that instructed participants to respond in a particular way (Marjanovic et al., 2015). Inattentive participants were defined as those who responded incorrectly on at least two thirds of those instructional items (Marjanovic et al., 2015). Our original sample consisted of 34 more participants who, judging by this criterion, needed to be excluded.

# Measures<sup>2</sup>

Participants completed an anonymous survey online that included the following measures and demographic questions.

<sup>2</sup> We also included the Bicultural Self-Efficacy Scale (BSES; David et al., 2009), which assesses the perceived competence for biculturals to function effectively and maintain meaningful relationships in both cultures. The original scale contains six subscales and we administered four of them: social groundedness (SG; α = 0.91) which refers to the extent to which a bicultural person's social networks consist of people from both cultures, positive attitudes (PA; α = 0.89) which refers to the extent to which a bicultural person holds both cultures in high regard, role repertoire (RR; α = 0.64) which reflects a bicultural person's range of culturally appropriate skillsets, and bicultural beliefs (BB; α = 0.80) which are the expectation that biculturals can function effectively in both cultures. The four subscales, particularly SG and PA, were also found to relate positively to global consistency in self-evaluations. Furthermore, DSS mediated most of those effects. Although bicultural self-efficacy is distinct from bicultural identity integration (David et al., 2009; in this study, rs between BII and BSES subscales ranging from 0.24 to 0.45), findings regarding BSES provide convergent evidence for our general argument that bicultural experience influences self-consistency. Detailed results can be obtained from the first author.

The order of items within each measure was randomized for each participant.

#### **Bicultural identity integration**

fpsyg-08-00321 March 6, 2017 Time: 17:58 # 9

It was assessed with the 20-item Bicultural Identity Integration Scale-Version 2, which was developed by Huynh et al. (2011; BIIS-2). BIIS-2 assesses two components of bicultural identity integration: blendedness (e.g., "I relate better to a combined culture than to my heritage or Canadian culture alone"), which is conceptually analogous to hybridity; harmony (e.g., "I feel torn between my heritage and Canadian cultures", reverse-coded), which corresponds with the opposite of conflict (see Zhang et al., 2014). Participants rated each item on a 5-point scale (1 = strongly disagree, 5 = strongly agree) and we averaged the respective items after reverse-scoring the appropriate ones to form subscale scores (blendedness: α = 0.70; harmony: α = 0.89). The higher the blendedness and harmony scores, the more integrated one's bicultural identity is.

#### **Dialectical self-beliefs**

Participants completed the same DSS on a 7-point scale. Although the overall α was acceptable (0.81), the αs for each subscale varied: contradiction (0.56), cognitive (0.72), and behavioral (0.51).

#### **Global self-consistency**

We measured global self in two domains: personality and selfevaluation. First, participants indicated how self-descriptive each of 28 personality traits was on a 7-point scale, ranging from 1 (not at all like me) to 7 (very much like me). To assess the extent to which biculturals possess contradictory traits, we followed Spencer-Rodgers et al. (2007, Study 2) in selecting half of the traits to be conceptually contradictory with the other half. As a result, there were 14 pairs (e.g., quiet/talkative, humble/proud, shy/sociable). Second, participants completed the Rosenberg Self-Esteem scale (Rosenberg, 1965) on a 7-point scale that ranged from 1 (not at all) to 7 (very much). Given our interest in evaluative contradiction, we examined the positive and negative self-evaluations separately. Thus, positive self-esteem was computed as the mean of the positive items, while negative self-esteem the mean of the negative items (see Spencer-Rodgers et al., 2004, Study 1).

To compute global self-consistency in personality judgment and self-evaluation, we adopted three indices that were developed initially in the attitudinal ambivalence literature but have been incorporated into research on the influence of dialecticism (e.g., Spencer-Rodgers et al., 2004; Boucher, 2011). Specifically, we used Similarity Intensity Model (SIM; formula: 3S − L, with S being the smaller value and L being the larger value), the Conflicting Reactions Model (CRM; formula: 2 × S), and the Gradual Threshold Model (GTM; formula: 5S <sup>0</sup>.<sup>50</sup> − L 1/S ). Although the three formulae differ in how much weight is given to the numerically smaller response in comparison to the numerically larger response, all capture the degree to which one responds in a contradictory or inconsistent way (Spencer-Rodgers et al., 2004). For instance, someone who answered "7" and "7" on a pair of contradictory items will receive the highest inconsistency score for all three indices, whereas someone who answers "1" and "7" or "7" and "1" will be assigned the lowest inconsistency score for all three. Someone rating oneself as "4" and "4" will receive an intermediate inconsistency score. For consistency in personality judgment, we applied the given formula to obtain a single value for each pair of contrasting traits and then averaged across all the 14 values to form one mean consistency index for each participant. For consistency in self-evaluation, we took the mean of the positive and negative self-evaluations, respectively, before they were entered in the given formula. This yielded a single consistency index for each participant as well. For both consistency indices, a higher score indicates being more globally inconsistent.

# Results and Discussion

To test H1, we correlated harmony and blendedness with the global consistency indices. As displayed in **Table 2**, the associations were in the predicted direction, but only four of them reached statistical significance and they all involved self-evaluative consistency. Thus, a more integrated bicultural identity (i.e., higher harmony and blendedness) was related to global consistency in self-evaluations, but not global consistency in personality judgment.

As discussed in Study 2, a mediation test can be conducted, even if the total effect was non-significant (MacKinnon et al., 2002; Shrout and Bolger, 2002). We thus proceeded to test H2. We first established that overall DSS was indeed correlated with harmony (r = −0.26, p < 0.001) and all the global consistency indices (rs ranging from 0.24 to 0.42, ps < 0.01). Separate mediation analyses with 5,000 bootstrapping resamples showed that none of the bias-corrected 95% CIs for the indirect effect of harmony on global consistency via overall DSS contained zero (see **Table 3**). Furthermore, behavioral DSS was found to correlate significantly with both blendedness (r = −0.15, p = 0.04) and the global consistency indices (rs ranging from 0.24 to 0.41, ps < 0.01). Additional mediation analyses similarly demonstrated that blendedness affected global consistency via behavioral DSS (see **Table 3**). Despite the weak or non-significant total effects of harmony and blendedness, the indirect effects via dialectical beliefs were robust. Such discrepancy in strength between the overall and indirect effects again suggests that the presumed causal effect of dialectical beliefs might be more proximal, whereas identity integration represents the more distal end of the causal process. Thus, one potential explanation for the non-significant total effects on global consistency in personality judgment might be that the causal effect of identity integration is more distal and thus smaller in magnitude than expected. We will return to this issue in the section "General Discussion" to consider other possibilities.

# GENERAL DISCUSSION

In the present research, we combine key insights from research on biculturalism and naïve dialecticism to understand self-concept consistency among biculturals. Across four studies, we examined whether and how individual differences in bicultural identity integration is associated with self-concept consistency. In line

with H1, experiencing cultural identities as being integrated into a coherent whole was associated with enacting a similar personality profile in different roles (Studies 1–3). It also related to making unambiguous evaluations of one's global self (Study 4). Despite ethnic variations within, bicultural Canadians are similarly shaped by the way in which they negotiate their cultural identities. In other words, this effect does not seem to depend on the types of culture biculturals inherit or whether both cultures are congruent with each other in dialectical traditions. It should apply to biculturals with both non-dialectical cultures (e.g., Latino Canadians), biculturals with one dialectical culture and one non-dialectical culture (e.g., East Asian Canadians or European Canadians living in East Asia), and presumably those with both dialectical cultures (e.g., Japanese of Chinese heritage). Next, our H2 regarding the indirect effect of dialectical self-beliefs received support (Studies 2–4). As cultural identities are better integrated within the self, biculturals tend to endorse dialectical beliefs less, which is in turn associated with being more consistent with regard to role enactment and one's general sense of self. This set of mediation evidence adds to the usefulness of naïve dialecticism as an explanatory tool for understanding national and ethnic differences in a wide array of consistency-related phenomena (Spencer-Rodgers et al., 2010). Last, we replicated the differences between bicultural groups of dialectical and nondialectical ethnic heritages that were found in prior research (H3). Latino Canadians (and European Canadians) showed, on average, higher consistency across roles than did East Asian

Canadians and such difference was mediated by dialectical selfbeliefs (Study 2). Thus, it appears that biculturals have at least two relatively independent sources that influence their levels of selfconsistency: their bicultural experience as well as their heritage culture.

Despite the overall evidence, however, there were some questions as to the magnitude and consistency of effects. A case in point is the magnitude of the overall effects of identity integration in Study 4: the overall effects of harmony and blendedness on global self-consistency were either non-significant or weak. As mentioned above, it may be that those direct effects represent rather distal effects, which are difficult to detect in samples that are not particularly large in size. This would suggest the causal impact of identity integration may be temporally removed such that it would take quite some time for it to spill over to other life domains. Thus, its subtle, cumulative effects may be impractical to capture in cross-sectional studies, especially among emerging adults (cf. Tadmor et al., 2012). Moreover, a related possibility is that there exist factors that moderate the effects of identity integration on self-consistency. One such moderator may be the operationalization of biculturalism. We used a broad criterion of inclusion (i.e., exposure-based definition) in all studies, but focusing on biculturals who are oriented to both cultures (i.e., identification-based definition) may strengthen the effects of identity integration as both cultures would presumably be central to the sense of self. From a socio-cognitive perspective (Tadmor and Tetlock, 2006; Amiot et al., 2007), there would be little


BII, bicultural identity integration; SIM, similarity intensity model; CRM, conflicting reactions model; GTM, gradual threshold model. <sup>∗</sup>p < 0.05, ∗∗∗p < 0.001.



BII, bicultural identity integration; DSS, dialectical self scale; SIM, similarity intensity model; CRM, conflicting reactions model; GTM, gradual threshold model.

incentive to integrating group identities within the self if one does not categorize oneself as being part of those groups or feel accountable to them (e.g., those who exclude one culture from self-categorization by adopting predominantly assimilation or separation acculturation strategy). Another moderator of the effects of identity integration on self-consistency may be the form of self-consistency. The first three studies differ from Study 4 not only in the measurement of identity integration but also the form of self-consistency assessed. Thus, it is not clear if the weaker effects of identity integration in Study 4 are attributable to either difference or both. To confirm that the effects of identity integration are stronger on cross-role than global selfconsistency, for instance, both forms of self-consistency will need to be assessed in the same study in the future. Future research may also benefit from using more spontaneous and implicit measures of global self-consistency (e.g., Spencer-Rodgers et al., 2009).

With respect to the consistency of effects, our studies varied in which dimension (cognitive vs. affective) of identity integration was more strongly linked with self-consistency in different bicultural groups. We offer three conceptually distinct explanations. To begin with, with more diverse bicultural groups (Studies 2 and 3), the affective dimension seemed to be a stronger predictor of self-consistency. This may reflect the general principle that negative affect powerfully shapes psychological experiences (Baumeister et al., 2001) or more specifically that reducing or avoiding conflict is the primary motive that guides other identity strategies (Huynh et al., 2011; Hirsh and Kang, 2016). There also appear to be some ethnic variations. In Studies 1 and 2, the cognitive dimension was a stronger predictor than the affective dimension among East Asian biculturals, whereas the opposite was true for Latino biculturals. However, it is unclear what may underlie such ethnic difference. Finally, there may be a matching effect such that the effect is stronger when the cognitive dimension is matched with a cognitive outcome or when the affective dimension is matched with an affective outcome. So the effects of affect could have been more consistent, had cross-role consistency been assessed in the form of feeling good vs. bad across contexts. In support of this reasoning, because affect can signal identity motives (Zou et al., 2008), it is identity conflict that predicted why some biculturals reacted contrastively to cultural primes (e.g., Mok and Morris, 2009). Similarly, the effects of cognitive representation are most likely to be observed with outcomes more tied to identity structure (Miramontez et al., 2008; Saad et al., 2013; Zhang et al., 2014). Future research could tease apart those possibilities to better understand the differential effects of the two components of identity integration.

In addition to the magnitude and consistency of the predicted effects, a main limitation of this research is the correlational nature of our studies. We could not show unambiguously that bicultural identity integration causes less endorsement of dialectical self-beliefs, which in turn leads to higher self-consistency. A viable alternative model is that identity integration and self-consistency are related because they are both consequences of holding dialectical self-views. There is some evidence suggestive of causality flowing from dialecticism to identity-related constructs. Sanchez et al. (2009) found among multiracial adults that relative to other mixedrace individuals, Asian/White individuals altered their racial identities more across contexts. Such group difference was possibly the consequence of exposure to dialecticism among Asian/White biracial individuals. This interpretation brings up the influence a bicultural person's heritage culture could have on the need to integrate different cultural identities. The reverse directionality may be particularly true of East Asian biculturals as their dialectical heritage might set them up for being more tolerant of unintegrated cultural identities in the first place. Thus, there may well be a reciprocal relation between identity integration and dialectical beliefs. To investigate this, researchers could employ a longitudinal panel design to test the respective strengths of identity integration predicting dialectical selfbeliefs vs. dialectical self-beliefs predicting identity integration in different bicultural groups.

# Implications

A novel contribution of this research lies in shedding new light on self-consistency that has primarily been studied in the context of cultural comparisons along national and ethnic lines. It is the first to demonstrate the connection of self-consistency with a process biculturals use to negotiate their cultural identities, thus filling in an important knowledge gap. To be specific, the present findings exemplify the value of exploring unique psychological products of being bicultural (Tadmor et al., 2012; Saad et al., 2013; West et al., under review). Bicultural experience may give rise to psychological characteristics that go beyond what may be predicted by each source culture in isolation. In our case, H1 and H2 represent novel predictions derived from the view that regards self-consistency as being shaped by the process of integrating identities associated with multiple cultures; instead, H3 operated from the prevailing assumption that self-consistency is largely influenced by one's cultural heritage alone. We showed that self-consistency is grounded in the dual-culture maintenance challenge shared among various bicultural groups in addition to heritage cultures that vary from one to another. As a matter of fact, without an explicitly bicultural perspective, it would not have been evident that there may be more than one cultural source that shapes biculturals' level of self-consistency. Therefore, to predict a bicultural person's level of self-consistency, we would first need to know his or her cultural heritage (i.e., whether dialectical or not) that serves as a baseline. His or her bicultural experience (i.e., identity integration) would then help fine-tune how far and in which direction the individual level of self-consistency deviates from the baseline.

More generally, in studying the experience of being bicultural, researchers need to be more mindful of connecting the psychological phenomenon of interest with specific aspects of the cultural contexts surrounding biculturals. In addition to heritage cultural norms and beliefs, biculturals could be unique in term of their identity negotiation processes, social status, obstacles to full inclusion, cultural adaptation, and so forth. In the case of Asian Canadians, for instance, whether it is their higher levels of social anxiety (Hsu et al., 2012), experience of lower relational mobility (Zhang and Li, 2014), or compensatory conformity to the European Canadian norm (Tafarodi et al., 2002), there are

likely more than one socio-cultural factor that account for those psychological characteristics. To be sure, depending on the form of culture, being bicultural could entail juggling any two or multiple important social identities, such as being a mother and being a career woman with contrastive prescriptions for what it means to succeed (Hodges and Park, 2013; Meeussen et al., 2016). Future research could examine if a similar process to what was found here operates at the intersection of social identities that are receiving increasing attention as social change progresses.

How do the findings relate to the work on dialectical thinking and biculturalism? We think the present research furthers our understanding of individual differences and movement in dialectical thinking in an ever-changing world. Like much cultural psychological research, research on naïve dialecticism construes individual differences in dialectical thinking as largely product of one's ethnic or national culture. However, our findings suggest that one need not to be explicitly steeped into East Asian philosophies to become dialectical: a dialectical view could arise out of recognizing contradiction and uncertainty in one's life. For some biculturals (e.g., Latino Canadians or Americans), the complexity of traversing two cultural worlds that are not perfectly aligned with each other could gradually prime them to think and feel in a more dialectical manner, even if neither culture promotes dialecticism. As for biculturalism research, the present research underscores the need to further investigate the psychological and socio-cognitive consequences of bicultural identity integration (e.g., Cheng et al., 2008; Saad et al., 2013) and to differentiate the cognitive from the affective dimension in their downstream effects (e.g., Miramontez et al., 2008; Mok and Morris, 2009).

Finally, results from this research have implications for understanding biculturals' psychological adjustment. Overall, previous literature shows a more integrated bicultural identity to contribute to greater well-being, but the mechanisms are not well-understood (Downie et al., 2006; David et al., 2009; Huynh et al., 2011). Given that one or both dimensions of identity integration was found to relate to dialectical self-beliefs (either the overall DSS or at least one subscale) in the studies reported here, one specific reason why a lack of identity integration reduces well-being may be that it encourages the tendency to accept both positive and negative sides of oneself and the surrounding social world. Holding dialectical beliefs is indeed somewhat detrimental to well-being, even in East Asian cultures

## REFERENCES


(Spencer-Rodgers et al., 2010). Further supporting the role of dialecticism, there is some evidence that dialectical self-beliefs mediated the effect of identity integration on well-being, at least among biculturals in Hong Kong and China (Chen et al., 2013). Coincidentally, because we used Rosenberg's self-esteem scale to assess self-evaluative ambiguity in Study 4, global self-esteem scores could be computed to index well-being. Unsurprisingly, global self-esteem was correlated negatively with overall DSS (r = −0.52, p < 0.001) and specifically harmony (r = −0.26, p < 0.001). A mediation test confirmed that the effect of harmony on global self-esteem occurred via DSS: 95% CI [0.09, 0.34]. Therefore, dialectical self-beliefs may be one mechanism through which a lack of identity integration comes to negatively impact biculturals' psychological well-being.

# CONCLUSION

It is a bicultural mind a consistent mind? The answer appears to depend on two factors. On the one hand, cultural content clearly matters. Because a dialectical cultural tradition attaches less importance to the need to maintain consistency, a bicultural who is enculturated to that tradition tends to develop a less consistent self-concept than one who is enculturated to a nondialectical tradition. On the other hand, our research highlights the process of integrating more than one culture within the self, irrespective of cultural content. A bicultural mind is a consistent mind when one manages to weave disparate cultural identities into a coherent whole. Ultimately, what unites these two factors is that dialectical beliefs seem to be the chief mechanism by which both come to shape self-consistency in biculturals.

# AUTHOR CONTRIBUTIONS

RZ and KN designed the studies. RZ, SS, and RL collected the data. RZ performed the data analysis. RZ drafted the manuscript and KN and RL provided critical feedback. The authors would like to acknowledge the Dickinson College Research & Development Committee and the Social Sciences and Humanities Research Council of Canada for their support of this publication.




**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Zhang, Noels, Lalonde and Salas. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Counter-Stereotypes and Feminism Promote Leadership Aspirations in Highly Identified Women

Carola Leicht <sup>1</sup> \*, Małgorzata A. Gocłowska<sup>2</sup> \*, Jolien A. Van Breen<sup>3</sup> , Soledad de Lemus <sup>4</sup> and Georgina Randsley de Moura<sup>5</sup>

<sup>1</sup> Kent Business School, University of Kent, Canterbury, United Kingdom, <sup>2</sup> Department of Work and Organizational Psychology, University of Amsterdam, Amsterdam, Netherlands, <sup>3</sup> Department of Psychology, University of Groningen, Groningen, Netherlands, <sup>4</sup> Department of Psychology, University of Granada, Granada, Spain, <sup>5</sup> School of Psychology, University of Kent, Canterbury, United Kingdom

#### Edited by:

Mark Hallahan, College of the Holy Cross, United States

#### Reviewed by:

Ilka H. Gleibs, London School of Economics and Political Science, United Kingdom Eric Mayor, University of Neuchâtel, Switzerland

#### \*Correspondence:

Carola Leicht acl23@kent.ac.uk Małgorzata A. Gocłowska goclowska.m@gmail.com

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

> Received: 27 October 2016 Accepted: 15 May 2017 Published: 02 June 2017

#### Citation:

Leicht C, Gocłowska MA, Van Breen JA, de Lemus S and Randsley de Moura G (2017) Counter-Stereotypes and Feminism Promote Leadership Aspirations in Highly Identified Women. Front. Psychol. 8:883. doi: 10.3389/fpsyg.2017.00883 Although women who highly identify with other women are more susceptible to stereotype threat effects, women's identification might associate with greater leadership aspirations contingent on (1) counter-stereotype salience and (2) feminist identification. When gender counter-stereotypes are salient, women's identification should associate with greater leadership aspiration regardless of feminism, while when gender stereotypes are salient, women's identification would predict greater leadership aspirations contingent on a high level of feminist identification. In our study US-based women (N = 208) attended to gender stereotypic (vs. counter-stereotypic) content. We measured identification with women and identification with feminism, and, following the manipulation, leadership aspirations in an imagined work scenario. The interaction between identification with women, identification with feminism, and attention to stereotypes (vs. counter-stereotypes) significantly predicted leadership aspirations. In the counter-stereotypic condition women's identification associated with greater leadership aspirations regardless of feminist identification. In the stereotypic condition women's identification predicted leadership aspirations only at high levels of feminist identification. We conclude that salient counter-stereotypes and a strong identification with feminism may help high women identifiers increase their leadership aspirations.

Keywords: women, gender identity, gender stereotypes, feminism, leadership

# INTRODUCTION

Recent years have seen a proliferation of campaigns promoting gender equality. In 2005, the web designer Gretchen Cawthon launched a website called "Girls Can't What?". The website sells merchandise featuring women in counter-stereotypic professions: e.g., women construction workers, women firefighters, and women scientists. In another campaign, in 2011, the gender studies department at the Massachusetts Institute of Technology promoted women's greater identification with feminism. As part of this campaign, members of the university received bags and badges promoting feminism, and were photographed holding signs saying, "This is what a feminist looks like"<sup>1</sup> . Events like these highlight an increasing recognition that gender disparities are detrimental to societies and economies, and suggest approaches that can be taken to address inequality. Our paper investigates two of these approaches: promoting gender counter-stereotypes, and increasing women's identification with feminism, and looks at the consequences they may have to women's leadership aspirations.

Studies suggest that campaigns stressing feminism and counter-stereotypes may help women to resist gender inequality, and that this is most likely in high women identifiers. Generally speaking, those who identify strongly with a group are most likely to protest against group-based inequality (Abrams and Randsley de Moura, 2002; Van Zomeren et al., 2008). However, as research shows, this is not always the case for gender. In fact past studies have found that high levels of gender identification associate with a greater susceptibility to threatening stereotypes, increasing the need for interventions targeting high women identifiers in particular. This suggests that additional conditions need to be met, to achieve high women identifiers' greater empowerment. While increased identification (e.g., with women) makes identity-related issues more relevant to the individual (Kaiser and Hagiwara, 2011), and can generally energize and motivate action (Van Knippenberg, 2000; Ellemers et al., 2004), the exact direction of those effects may depend on several moderators. For instance research on group identification has shown that its effects are contingent on salient identity cues (James and Greenberg, 1989; Van Knippenberg, 2000), and on specific identity content (Becker and Wagner, 2009) that may lead individuals to perceive certain actions (e.g., in this case, greater leadership aspirations) as beneficial for oneself or the in-group (Van Knippenberg, 2000). In consequence, the energizing effect of group identification on women's empowerment should depend on the salience of gender counter-stereotypic (vs. stereotypic) cues (Hoyt and Murphy, 2016), and on the presence of a politicized identification, like feminism, which sees resistance to stereotypic content as beneficial (Van Breen et al., in review). Based on this premise we asked whether counter-stereotype (vs. stereotype) salience, and feminist identification can moderate the relationship between women's identification and their leadership aspirations.

# Counter-Stereotypes and Leadership Aspirations

Because leadership is typically associated with agency and masculinity (Sczesny and Kühnen, 2004; Hogue and Lord, 2007) and womanhood with nurturance and warmth (Eagly et al., 1992; Eagly and Karau, 2002), prevalent stereotypes create psychological barriers to women's participation in leadership (Eagly and Johnson, 1990; Rudman and Glick, 1999). Moreover, a heuristic tendency to prefer leaders that are perceived as embodying the ingroup's image (Hogg, 2001) makes it difficult for women to rise to leadership positions in male dominated domains. As such, stereotypes represent a considerable barrier to women's leadership aspirations. Indeed, past research has shown that exposing women participants to gender stereotypes can dampen their leadership aspirations (Davies et al., 2005). Likewise, under stereotype salience women report lower perceived performance, diminished leadership aspirations, diminished feeling of belonging, greater perceived task difficulty, and feelings of inferiority on leadership task (Davies et al., 2005; Cheryan et al., 2012, 2013). Additionally, when gender stereotypes are salient women may fear and/or experience backlash when they display leadership behavior (Eagly and Carli, 2007; Rudman and Fairchild, 2007; Amanatullah and Morris, 2010; Williams and Tiedens, 2015), leading them to avoid leadership roles to an ever greater extent. Conversely, increasing the salience of counter-stereotypes can maintain and even increase leadership aspirations in women (provided that these counter-stereotypes do not evoke threatening upward comparisons; Dasgupta and Asgari, 2004; Rudman and Phelan, 2010; Hoyt and Simon, 2011), and reduce the tendency to focus on in-group similarity (e.g., Leicht et al., 2014). One reason for these effects may be that counter-stereotypic content signals that it is safe, possible, and perhaps even desirable for women to behave in gender counter-stereotypic ways.

Taken together, when counter-stereotypes are salient, high identification with gender could energize women and increase their leadership aspirations. But women's behavior is not merely a function of group identification and of stereotype (vs. counterstereotype) salience. Other factors are known to help women break out of stereotypic roles, and overcome disadvantage. This research suggests that women's reactions to gender issues (e.g., gender stereotypes) are determined not only by gender identification, but also by identification with feminism (Van Breen et al., in review, Van Breen et al., manuscript in preparation).

# Feminist Identification and Stereotype Resistance

Gender constitutes one of people's most salient group identities, giving individuals a sense of belonging, and providing social roles and norms to live by (Tajfel, 1974; Banaji et al., 1993; Abrams and Hogg, 2004). But whilst much research on gender and performance focuses on the extent of women's gender identification (Kaiser and Hagiwara, 2011), how women construe gender identity is rarely considered. Becker and Wagner (2009) (also see Condor, 1986) proposed that when studying social identities researchers should consider not only the strength of people's identification, but also specific identity content (Becker and Wagner, 2009). This issue may be particularly important in the case of gender identification, because as research shows, feeling strong ties with other women does not preclude the perpetuation of gender inequality (Klandermans, 2014).

In an attempt to better understand how women construe their identity in relation to gender, recent research distinguished women's identification with other women from a second, orthogonal factor, namely women's identification with feminism. This research demonstrated that identification with feminism and with women constitute distinct and unique social identities, and two independent psychological constructs (Roy et al., 2007;

<sup>1</sup>For the website of the Girls Can't What campaign follow this link: https://www. girlscantwhat.com. For images from the MIT campaign, follow this link: https:// www.flickr.com/photos/59603194@N02/sets/72157626631504524/

Van Breen et al., in review). Extending those findings, we propose that in addition to women's gender identification, also identification with feminism may shape how women react to gender stereotypes. One possibility in particular, is that when stereotypic content is salient, high levels of feminist identification will direct high women identifiers' motivation toward actions that are considered beneficial for the group—such as stereotype resistance (Van Knippenberg, 2000; de Lemus et al., 2013; Van Breen et al., in review). This effect could be similar to what is observed in empowered women. For example, women entrepreneurs (Gupta et al., 2008) and women in high power positions (Hoyt and Blascovich, 2007) have been found to perform better than men when gender divisions are salient. A similar resistance effect may also be observed in high feminist identifiers. In other words, it is possible that, when stereotypes are salient and feminist identification is high, women's identification may associate with greater leadership aspirations.

In sum, our literature review suggests that women's identification should associate with greater leadership aspirations through two different "routes" (1) under conditions of counterstereotype salience (in all research participants) and (2) under conditions of stereotype salience, but only provided that feminist identification is also high.

# This Research

As highlighted, gender stereotypes (vs. counter-stereotypes) and feminist identification could moderate the link between women's identification and leadership aspirations. To test this idea we looked at the association between women's identification with other women and their leadership aspirations under conditions of stereotype (vs. counter-stereotype) salience, and in high (vs. low) feminist identifiers. We predicted that when gender counterstereotypes are salient, women's identification should associate with greater leadership aspiration regardless of feminism, while when gender stereotypes are salient, women's identification would predict greater leadership aspirations contingent on a high level of feminist identification. We chose leadership aspirations as our dependent variable because leadership is counter-stereotypic for women and persistence in this domain signals women's greater resistance to stereotypic content. For exploratory reasons, we also measured effects to participants' fear of backlash.

# PRE-TEST STUDY

Before running the main study we developed and pre-tested a novel manipulation of stereotype (vs. counter-stereotype) salience. We wanted to manipulate attention to stereotypes (vs. counter-stereotypes) without evoking threatening social comparisons (e.g., asking participants to think of a highly successful female CEO is potentially threatening, and may wipe out the effect of the counter-stereotype; Hoyt and Simon, 2011). To achieve this, we decided to present participants with neutral images of women and ask them to focus on how the women targets are portrayed in a stereotypic (vs. counterstereotypic) manner. This type of manipulation, where the content of the image is kept constant across both conditions, has the additional benefit of being "clean" and confoundfree. Typical manipulations of stereotype and counter-stereotype exposure ask participants to think of different targets, for instance Angela Merkel vs. Bill Clinton (Latu et al., 2013), or a "female mechanic" vs. a "male mechanic" (Gocłowska et al., 2014; Leicht et al., 2014), potentially confounding stereotype (vs. counterstereotype) exposure with other features of the target (e.g., gender). Our new paradigm avoids these potential limitations, allowing us to assess the impact of stereotype (vs. counterstereotype) salience and nothing else.

We pretested four images of women (available from authors upon request). Image A depicted a woman standing at a large window, image B a woman in front of a laptop holding a cell phone and a baby, image C a woman eating a salad in front of a computer, and imaged D a woman aircraft pilot with a military aircraft in the background. Next, we asked several questions about the target's stereotypicality and femininity.

The first goal of the pretest was to select (for the main study) two images that are most neutral in terms of stereotypicality and femininity. Using neutral imagery would ensure that the targets used in our manipulation are relevant to all research participants, and that they can be perceived as both stereotypic as counter-stereotypic, depending on instructions used. The second goal was to test the effectiveness of our new stereotype (vs. counter-stereotype) attention manipulation. Asking participants to describe the stereotypic (vs. counter-stereotypic) features of the (relatively neutral) female targets, should change participants' perceptions of those targets in line with the manipulation. Most likely, focusing on stereotypic (vs. counter-stereotypic) content should amplify ratings of stereotypicality and femininity. Competence, likability and attractiveness, emotional reactions and task difficulty could also potentially be affected, and were thus measured as well.

# Methods

US based MTurk workers (N = 259; 101 male, 157 female, 1 gender not specified) took part in an online survey (Mtime = 12 min, \$1 reward)<sup>2</sup> . Demographic questions indicated that 12% of participants were students, 13% were self-employed, 59% were full or part-time employed, 10% were homemakers, and 6% were jobseekers. We also inquired about gender distribution in the research participants' main organization (company, university etc.). According to participants' estimates, among "ordinary" workers with no managerial responsibilities 53% were female, among lower and middle management employees 48% were female, while among upper management employees 37% were female. Those numbers seem to confirm the general observation that women are underrepresented in upper management positions.

At the beginning of the study participants were allocated randomly to one condition in a 4 (Image: A vs. B vs. C vs. D) by 2 (Attention: counter-stereotypic vs. stereotypic) betweenparticipants design. Within each condition we asked participants

<sup>2</sup>During study completion 16 participants failed attention checks, and their data were excluded from further analysis, leaving a final sample of 242 participants (96 male, 145 female, 1 gender not specified; Mage = 38.15, SD = 12.74).

to describe the image they were presented with for at least 90 s. In the stereotypic condition we asked participants to describe the image focusing on its "traditional and female gender stereotypic aspects", whilst in the counter-stereotypic condition we asked participants to focus on "non-traditional and female gender counter-stereotypic aspects". After participants described the image, we asked about their perception of the task and of the stimuli.

All items were measured on a 7-point Likert-type scale (1, not at all; 7, very much). To measure target stereotypicality we used three items: traditional, stereotypic or contrary to what society expects (α = 0.73). Target femininity was measured across various domains (e.g., behavior, posture, fashion; α = 0.87). For exploratory purposes we asked participants to indicate how attractive, how competent and how likeable the target was (we treated those as separate variables), and to rate positive (e.g., happy, enthusiastic, α = 0.89) and negative (e.g., disappointed, anxious, α = 0.90) emotions evoked by the target. Finally, we also asked participants to indicate how easy or pleasant the task was (e.g., How much did you enjoy describing the image? How easy it was to describe the image?; 5-item measure, α = 0.85).

# Results

**Table 1** represents mean ratings and pairwise comparisons of the four types of images. Images differed significantly in terms of stereotypicality [F(3, 238) = 29.48, p < 0.001, η <sup>2</sup> = 0.27] and femininity [F(3, 238) = 25.76, p < 0.001, η <sup>2</sup> = 0.25]. Pairwise comparisons (post-hoc LSD test) showed that image C (woman eating a salad) was perceived as overly stereotypic and feminine,

TABLE 1 | Differences in the ratings of the four stimuli in Pretest Study and in Study 1.

while image D (woman pilot) was overly counter-stereotypic and relatively low on femininity. Based on those results we decided to only use Image A (woman in front of the window) and Image B (woman with baby and laptop) in the main study.

**Table 2** represents mean image ratings in the counterstereotypic and stereotypic condition. As we hoped for, images in the stereotypic (vs. counter-stereotypic) condition were perceived as more stereotypic [F(1, 240) = 21.78, p < 0.001, η <sup>2</sup> = 0.08] and more feminine [F(1, 240) = 12.43, p < 0.001, η <sup>2</sup> = 0.05]. Additional exploratory analyses demonstrated that the stereotypic condition decreased ratings of competence [F(1, 240) = 8.07, p = 0.005, η <sup>2</sup> = 0.03], and positive emotions evoked by the target [F(1, 240) = 7.93, p = 0.005, η <sup>2</sup> = 0.03]. The manipulation had no effect on ratings of attractiveness or likability or on negative emotions and the perceived task difficulty. Overall, this set of findings suggests that our manipulation changed perceptions of targets in line with gender stereotypic (vs. counter-stereotypic) content. In addition, targets in the counter-stereotypic condition evoked greater (positive and negative) emotion and were seen as more competent.

# MAIN STUDY

Having pre-tested our manipulation and selected the two most neutral images, we set out to run the main study. We asked participants to attend to stereotypic (or counter-stereotypic) features of the selected stimuli, and measured leadership aspirations and identification with women and with feminism.


Means with a different superscript are significantly different from one another at p < 0.05. Image A represented a woman looking out of a window, Image B a woman with a baby, mobile phone, and a laptop, Image C a woman eating lunch in front of a computer, and image D a woman pilot in front of a military aircraft. Images can be obtained from authors' upon request.


TABLE 2 | Differences in ratings between the counter-stereotypic and stereotypic condition in the Pretest Study and in Study 1.

Means with a different superscript are significantly different from one another at p < 0.05. Pretest ratings represent aggregated ratings of four images (A, B, C, and D) while Study 1 ratings represent aggregated ratings of two images selected from the pretest (A, B).

Our goal was to test whether the three-way interaction of women's identification, feminist identification and stereotype salience would predict women's leadership aspirations.

# Methods

#### Participants, Design, and Procedure

Using MTurk, 238<sup>3</sup> female US based workers (Mage = 36.06, SD = 12.23) were recruited to take part in an online study on "image perception and description," and were allocated randomly to the stereotypic or counter-stereotypic attention condition. The questionnaire took, on average, 31 min to complete, and participation was rewarded with \$1.5.Participants' professional status was the following: 9% students, 12% self-employed, 55% full or part-time employed, 16% homemaker, 8% jobseeker. Also in this sample, gender distribution was perceived as being unequal in the highest ranks in the organization. Among "ordinary" workers with no managerial responsibilities 51% were thought to be female, among lower and middle management employees 49% were female, while among upper management employees 38% were reported to be female.

After completing informed consent participants were presented with one of our two selected images. We used and counter-balanced (between-participants) two images to ensure greater external validity and generalizability of our findings<sup>4</sup> . As in the pre-test study we asked participants to identify image elements portraying the woman in a stereotypic/traditional or non-stereotypic/ non-traditional way. Participants had to spend at least 90 s on this task (more time was allowed as well). We then measured leadership aspirations. We instructed research participants to imagine a workplace scenario and indicate to what extent they would like to and would feel comfortable taking a leadership role in that situations. For exploratory purposes we also measured participants' fear of backlash. Both variables were measured in a counterbalanced way, to account for order effects.

Following our manipulation and the workplace scenario, we measured several individual difference moderators and covariates. This part of the questionnaire inquired about participants' gender, their identification with women and with feminism, and several individual difference measures related to feminism. To account for individual differences in people's beliefs about gender (Morton et al., 2009; Napier et al., 2010; Okimoto and Brescoll, 2010), we asked participants to fill in measures of gender essentialism, gender system justification, benevolent sexism and liberal feminist attitudes (Glick and Fiske, 1996; Jost and Kay, 2005; Napier et al., 2010; Connelly and Heesacker, 2012; Brescoll et al., 2013; de Lemus et al., 2015). We wanted to use those measures as covariates, to see whether the effects uncovered are specifically due to feminist identification, rather than merely feminist beliefs. If effects of feminist identification (and it's interaction with other variables) held regardless of other gender related constructs, this would indicate that our effects are specific to feminist identification.

Because this part of the questionnaire came after the manipulation, we took precautions to ensure that participants report their stable beliefs and level of identification. Namely, before measuring identification and gender-based beliefs we instructed participants (several times) to "express to what extent you generally agree with the statements below" and to provide answers based on "your attitudes, beliefs and experiences most of the time in your everyday life." Furthermore, we also inspected the data to see whether the manipulation had any effects on the individual differences: none effects were uncovered (ps > 0.52). This finding is consistent with that of other research measuring identification after the manipulation (Jimenez-Moya et al., 2015).

Finally, we reminded participants of the stereotype (vs. counter-stereotype) salience manipulation, and asked them to complete the same measures as in the pre-test, except that this time we treated stereotypicality and femininity as manipulation checks.

#### Dependent Variables

All measures in the study were captured using a Likert-type scale with 1 indicating "not at all" and 7 indicating "very much." For correlations of all the measures in this study see **Table 3**.

<sup>3</sup>Thirty participants failed the two attention checks included in our experiment and were therefore excluded from our sample, leaving us with a final sample of 208 participants for analysis (Mage = 36.60, SD = 12.46). Conducting the analyses with the full dataset did not change the significance of the results.

<sup>4</sup>The study also included a third factor: guided (vs. unguided) focus. In the guided condition participants were asked to focus on specific aspects of the image while focusing on stereotypic or counter-stereotypic features. In the unguided condition no specific instructions were given. This factor is of little theoretical significance, and did not emerge as a significant predictor.


TABLE 3 | Bivariate correlations between dependent variables, moderators, covariates, and manipulation checks.

\*\*p < 0.001; \*p < 0.01; GSJ, Gender System Justification; LFAIS, Liberal Feminist Attitudes and Ideology Scale.

We embedded the leadership aspirations and fear of backlash measures in an imagined work scenario. First, we asked participants to think of a place where they frequently are, or have been involved in group projects. We asked them to focus on a place that is most current and relevant to what they do, and to write the name of this place down in the box provided. Common responses given by participants were "work," "office," "school," "university," "college," or "church". On the next page we referred to this location, and asked participants to imagine the following situation:

"On a typical day at [location name provided by research participant] you were selected to be part of a new project team, responsible for completing a specific task. You know who your team members are, and you know that all of you have similar expertise and experience on this type of project."

We then asked participants to answer several questions about their leadership aspirations and fear of backlash in the context of this event. Leadership aspirations were measured by asking participant to indicate on 13 statements whether they would like to take on the leadership role within this group (e.g., "I would like to be selected as a leader for this task."). These statements were mostly self-created, with a few borrowed from previous research (α = 0.95) (Hoyt and Simon, 2011). For an overview of the items see **Appendix** at the end of this paper. Fear of backlash was also assessed within this imaginary work workplace scenario, however, rather than assessing whether participants would like to take on the leadership role, we asked about participants' level of (dis)comfort in leadership positions. Using four items borrowed from previous research (Moss-Racusin and Rudman, 2010) we asked each participant how (un)comfortable she would feel being appointed as a leader (α = 80; e.g., "I would be concerned that I might be disliked."). For fear of backlash, the instruction additionally stressed that the research participant was the group member with the highest expertise for the task.

#### Moderators

Identification with women and with feminists was measured using two parallel sets of items. Three items measured participants' identification with women, (e.g., "Being a woman is an important part of who I am"; α = 0.82), and three similar items were used to measure identification with feminism, replacing the word "woman" with "feminist," α = 0.99 (de Lemus et al., 2015).

#### Covariates

Gender System Justification (GSJ), Gender Essentialism (GE), Benevolent Sexism (BS) were measured using well-established measures with good reliability, αGSJ = 0.72, αBS = 0.91, αGE = 0.83. (Glick and Fiske, 1996; Jost and Kay, 2005; Okimoto and Brescoll, 2010). In order to distinguish between identification with feminism, and attitudes toward feminism, we also asked participants to fill in the 14 item Liberal Feminist Attitudes and Ideology Scale (LFAIS; α = 0.89). Here participants were asked to indicate the extent to which they agreed with statements such as "A woman should have the same job opportunities as a man." (Levonian Morgan, 1996).

#### Manipulation Checks

Using items identical to those in the pretest we measured stereotypicality (α = 0.58), femininity (α = 0.85), competence (one item), likeability (one item), and attractiveness (one item) of the model as well as positive (α = 0.91) and negative emotions (α = 0.91), and task difficulty (α = 0.85). For correlations between all measures see **Table 3**.

# RESULTS

# Collinearity Check

## Gender Identity, Feminist Identity, and Feminist Attitudes

Our hypotheses were tested with moderated regression analysis in PROCESS macro (Hayes, 2013). Before running these analyses, we conducted a collinearity check on our two identity predictors and a content related co-variate of feminist attitudes. This would ensure that gender identification, feminist identification, and feminist attitudes are distinct constructs. First, we conducted a Factor Analysis with all items from all scales included in the analysis. With a varimax rotation, the analysis revealed a four factor structure explaining 70.45% of variance. The first two factors were formed out of the liberal feminist attitudes scale, with all positive phrased items loading on Factor 1 and all negative phrased items loading on Factor 2. The third factor consisted out of the three gender identification items, whereas the fourth factor consisted out of the three feminist identification items. Correlation analyses (see **Table 3**) additionally revealed that the liberal feminist attitude scale was related to feminist identification, but not with women's identification or leadership aspirations.

In summary, all three scales formed distinct factors that corresponded with our variables, gender identification, feminist identification, and liberal feminist attitudes, none of the items cross loaded, and correlations were non-significant between our two predictor variables, gender identification and feminist identification. This confirms that feminist identification and women's identification are distinct factors, and that that including these variables as predictors and covariates in a regression analysis is not problematic (Antonakis et al., 2010).

# Manipulation Check

Results of the manipulation check were very similar to what we found in the pretest. **Table 1** (bottom panel) shows mean ratings across the two types of images used. As in the pretest, image A and image B did not differ significantly in terms of stereotypicality or femininity. More importantly to our argument, mean image ratings in the counter-stereotypic and stereotypic condition (**Table 2**) demonstrated that the manipulation worked as intended. Namely, images in the stereotypic condition were perceived as more stereotypic [F(1, 206) = 49.06, p < 0.001, η 2 = 0.19] and more feminine [F(1, 206) = 14.96, p < 0.001, η <sup>2</sup> = 0.07]. The targets in the stereotypic condition were also seen as less competent [F(1, 206) = 8.64, p = 0.004, η <sup>2</sup> = 0.04], and elicited less positive emotions [F(1, 206) = 7.26, p = 0.008, η <sup>2</sup> = 0.03]. Thus, as intended, our manipulation changed perceptions of targets in line with gender stereotypic (vs. counter-stereotypic) content.

# Hypothesis Test

The main goal of our study was to examine under what conditions women's gender identification links with greater leadership aspirations. We tested this idea using the PROCESS macro Model 3 (Preacher and Hayes, 2008). We entered leadership aspirations as the dependent variable, and gender identification, feminist identification and attention (0 = attention to counter-stereotypic information, 1 = attention to stereotypic information) as predictor variables.

TABLE 4 | Results of the key regression analyses with leadership aspirations entered as the dependent variable.


Sensitivity Analysis represent the key finding when controlling for gender relevant variables. Significant predictors are marked in bold. Fem.Id., Feminist Identification; Wom.Id., Identification with Women; Wom.Id. x Fem.Id., Interaction term of identification with women and with feminists; Condition: counter-stereotypic coded as "0" and stereotypic coded as "1"; Wom.Id. x Condition, Interaction term of identification with women and of condition; Fem.Id. x Condition, Interaction term of identification with feminism and of condition; Wom.Id. x Fem.Id. x Condition, three way interaction term of the two identifications and the condition; GSJ, Gender System Justification; LFAIS, Liberal Feminist Attitudes and Ideology Scale.

#### Leadership Aspirations

The analysis revealed a significant main effect for gender identification, B = 0.32, SE = 0.13, t(200) = 2.50, p = 0.013 95%CI [0.07;0.57], showing that identification with gender was positively associated with leadership aspirations (also see **Table 4**, left panel). There were no main effects of feminist identification and attention on leadership aspirations (all p's > 0.46). The effect of gender identification was qualified by a significant three-way interaction between gender identification, feminist identification and attention, B = 0.28, SE = 0.11, t(200) = 2.48, p = 0.01 95%CI [0.06;0.51], which significantly increased the total amount of variance explained, R <sup>2</sup> = 0.07, p = 0.03; R <sup>2</sup> change = 0.03, p = 0.01. To break down the interaction term, we split the sample by condition and entered feminist identification and women's identification as the independent variables, and leadership aspirations as the dependent variable (see **Figure 1**).

In the counter-stereotypic condition there was a significant main effect of women's identification: B = 0.42, SE = 0.18, t(101) = 2.33, p = 0.02 95%CI [0.06;0.77], but feminist identification and the interaction terms did not emerge as significant predictors (ps > 0.22). Notably, inspecting the standardized regression coefficients revealed that the effect of women's identification on leadership aspirations was higher in the counter-stereotypic condition (β = 0.239), than in the overall sample (β = 0.186). In other words, women who were highly identified with other women (vs. those who were not) showed increased leadership aspirations after attending to counter-stereotypic information.

stereotypic condition (lower).

In the stereotypic condition the significant three way interaction was explained by a two way interaction between gender identification and feminist identification: B = 0.18, SE = 0.07, t(99) = 2.34, p = 0.02, 95%CI [0.03;0.33]. When further broken down, this significant interaction in the stereotypic condition resulted in one significant slope. Gender identification was associated with higher leadership aspirations in the stereotypic condition, however this was only the case if identification with feminism was also high, B = 0.68, SE = 0.26, t(99) = 2.66, p = 0.01, 95%CI [0.16;1.20]. Put differently, while under counter-stereotype salience women's identification associated with greater leadership aspirations regardless of feminism, under stereotype salience women's identification associated with greater leadership aspirations only when feminist identification was high.

To test for the robustness of those results, we re-run our analyses with Gender System Justification, Benevolent Sexism, Gender Essentialism, positive and negative emotions, and the Liberal Feminist Attitudes Scale as covariates (See **Table 4**, right panel). When controlling for these variables, the effect of the three-way interaction term of stereotype exposure and women and feminist identification was still significant: B = 0.33, SE = 0.11, t(194) = 2.86, p < 0.01, 95%CI [0.10;0.55].

#### Fear of Backlash

For exploratory purposes we conducted identical analyses with fear of backlash as a dependent variable. The analyses revealed a significant main effect of feminist identification on fear of backlash, B = 0.09, SE = 0.04, t(200) = 2.45, p = 0.02, 95%CI [0.02;0.17], so that higher feminist identification was associated with greater fear of backlash. There was also a marginal negative effect of gender identification, B = −0.20, SE = 0.11, t(200) = −1.90, p = 0.06, 95%CI [−0.42;0.011]. Attention to stereotypes was not significant, and there were no significant interaction effects (all ps > 0.26).

## GENERAL DISCUSSION

Counter-stereotypes and feminism are thought to inspire and motivate women to challenge gender inequality, and are often used in public campaigns and interventions. Could such means be helpful in raising the leadership aspirations of highly identified women? To answer this question we tested whether stereotype (vs. counter-stereotype) exposure and feminist identification moderate the effect of gender identification on women's leadership aspirations. A review of the literature suggested that high gender identification should associate with greater leadership aspirations (1) under conditions of counter-stereotype salience (in all research participants) and (2) under conditions of stereotype salience, conditional on greater levels of feminist identification. Those predictions were based on what we know about women's reactions to non-threatening counter-stereotypes (Hoyt and Simon, 2011; Latu et al., 2013), and based on the literature on stereotype resistance (de Lemus et al., 2015; Van Breen et al., in review). In line with these expectations we uncovered that stereotype (vs. counter-stereotype) salience and participant's identification with women and with feminism interacted to predict female participants' leadership aspirations. The three-way interaction term was further explained by a significant effect of women's identification (regardless of feminist identification) on leadership aspirations in the counterstereotypic condition, and by a significant effect of women's identification on leadership aspirations in the stereotypic condition (but only in high feminist identifiers). These findings are in line with recent research showing that gender identification and feminist identification are orthogonal identity constructs, and that women's reactions to gender stereotypes and to counterstereotypes depend the interaction of these two identities (de Lemus et al., 2015; Van Breen et al., in review). Our research extends those findings by showing that women's and feminist identification interact with stereotype salience to affect behavioral intentions such as leadership aspirations.

# Limitations and Future Directions

There are a number of limitations regarding the operationalization of theoretical constructs, as well as the generalizability of the research results. For example, leadership aspirations were measured using self-report, rather than actual leadership behavior. Agreeing with this criticism, we perceive the present set of findings as a preliminary step, rather than a definitive statement, about the role that identification and stereotypes (vs. counter-stereotypes) play in women's leadership aspirations. We advocate that in the future more research is needed to investigate whether the effects found in the present study carry over to women's performance on actual leadership tasks, such as tasks that require of women to lead on a team project, or to presenting a speech in front of their team members (Latu et al., 2013). Furthermore, more research on this topic could be conducted in organizational settings and in active women leaders. For example looking at whether in traditional organizations (where gender stereotypes abound) women who endorse progressive identification are more often promoted to leadership roles would garner more support for and extend the present set of findings. Although correlational in nature, this type of test would suggest that not only do high women high feminist identifiers have greater leadership aspirations, but that they also successfully realize these aspirations. This type of test would allow for a better understanding of the extent to which the processes investigated herein can also be found in "real-life" settings.

# Theoretical and Practical Implications

The current findings have several implications for theory and practice. First of all, they demonstrate, in line with other social identity research, that higher levels of identification with one's group can fuel individuals' motivation, but that the direction of those effects depends on variables related to identity salience and to one's perception of what is beneficial to oneself and the in-group. In line with this reasoning, our results show that the effect of gender identification apparent in our study is contingent on the salience of counter-stereotypic or stereotypic identity cues (James and Greenberg, 1989; Worchel et al., 1998; Van Knippenberg, 2000; Hoyt and Murphy, 2016), and on a politicized identification (feminism) that encourages resistance toward stereotypic norms (Van Breen et al., in review). Altogether these findings support the idea that counter-stereotypes and feminism can help increase women's aspirations, and that this is especially the case for women highly identified with their in-group.

Secondly, consistent with recent research on women's identification (Van Breen et al., in review), our study supports the idea that gender identification and feminist identification are unique constructs, and that they can shape women's reactions to stereotypic or counter-stereotypic information. Van Breen et al. (in review) proposed that whilst some women have a more traditional gender identification, and identify highly with women, but not with feminism ("traditional women"), others have a more progressive understanding of their gender identity, and identify highly with both women and with feminism ("dual identifiers"), only with feminism ("distinctive feminists"), or with none ("low identifiers"). Interpreting our results through this lens suggests that the distinction between gender and feminist identification is especially important in a salient gender stereotypic context. Namely, in this context we see that when stereotypes are salient, progressive women (who identify highly with women and with feminism) have greater leadership aspirations than distinctive identifiers (who identifying highly with feminism but not with women). This finding is very interesting as it opens up new interpretations on the role of group identification in stereotypic contexts. Stereotype threat literature has shown some pessimistic indications that highly identified women are at more risk from stereotype threat (Kaiser and Hagiwara, 2011). While we believe this to be true, we also think that gender identification can have beneficial effects for women—provided that it is supplemented with some kind of belief, or a second type of identity, that inoculates women against the debilitating effects of stereotype salience. The present set of findings suggests that when feminist identification is high, stereotypic content fuels leadership aspirations amongst those who are strongly identified with the group. We believe that this finding opens up a new and exciting possibility for stereotype threat research: that fostering a feminist identification can inoculate women against threatening stereotypic content. Moreover, the current set of findings may even suggest that feminist identification promotes leadership aspirations, as much as counter-stereotypes do. This finding emphasizes the importance of resistance to stereotypes as a motivated response to protect the interests of the group (de Lemus et al., 2015; Van Breen et al., in review).

At this point in the discussion an inquisitive reader may ask why, in the present study, women's identification was related to leadership aspirations (contingent on counter-stereotype salience and on feminism), but feminism was not. Since feminism is often associated with a greater readiness for collective action (Bliuc et al., 2007; Klandermans, 2014), and with the endorsement of women in gender counter-stereotypic roles, one might expect a link between women's feminist identification and their greater leadership aspirations. Our study, on the other hand, shows that feminist identifiers do endorse greater feminist beliefs (lower gender system justification, gender essentialism and benevolent sexism, and higher level of liberal feminist attitudes), but not greater leadership aspirations. In fact, as the correlations in

**Table 3** indicate, leadership aspirations are not correlated with any of the feminist beliefs or gender beliefs. The only correlation that emerges with gender content is that of a positive relation between leadership aspirations and gender system justification. In addition, the data suggests that high leadership aspirations associate with greater perceived liking of the task, greater amount of positive emotions evoked by our image stimuli, and with greater ratings of model likeability. This may suggest that women with greater leadership aspirations are more dominant, confident and comfortable leaders, and that they see things in a more positive light. Thus, greater leadership aspirations are not directly linked to feminist beliefs, and can even lead to support for greater gender inequality. This observation is consistent with the "Queen Bee" phenomenon, showing that female leaders can sometimes be perpetrators of discrimination (e.g., Derks et al., 2015), and our data seem to support this point. Thus, if anything, leadership aspirations probably reflect one's resilience in light of societally imposed norms or barriers, rather than a politicized motivation to challenge gender inequality.

Next to these more theoretical contributions, our research also has several practical implications. It highlights the benefits of recent endeavors of policy makers and practitioners to increase greater the salience of gender counter-stereotypic behavior and women's identification with feminism. While most research focuses on the role of feminism to collective action, hereby we extend the benefits of feminist identification to organizational outcomes, such as leadership aspirations. Overall, our findings indicate that having a more progressive gender identity, that

# REFERENCES


combines identification with women and with feminism, could have positive effects women's career progression, by motivating them to increase engagement in gender counter-stereotypic domains such as leadership.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the American Psychological Association, with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the ethics committee of the School of Psychology at the University of Kent.

# AUTHOR CONTRIBUTIONS

CL and MG made an equal contribution to this project: they designed and conducted the studies, analyzed the data, and wrote the paper. SdL, JVB, and GRdM contributed to the project conception and provided critical comments and revisions.

# FUNDING

This work was supported by Grant no. PSI2016-79971-P from the Spanish Ministry of Science and Technology (AEI/FEDER, UE) awarded to MG, JVB, and SdL, and by a Marie-Curie Fellowship awarded to MG (FP7-PEOPLE-2013-IOF, 622331, CREA.TA).


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Leicht, Gocłowska, Van Breen, de Lemus and Randsley de Moura. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# APPENDIX

# LEADERSHIP ASPIRATIONS ITEMS


# Multiple Social Identities Enhance Health Post-Retirement Because They Are a Basis for Giving Social Support

Niklas K. Steffens\*, Jolanda Jetten, Catherine Haslam, Tegan Cruwys and S. Alexander Haslam

School of Psychology, The University of Queensland, Brisbane, QLD, Australia

We examine the extent to which multiple social identities are associated with enhanced health and well-being in retirement because they provide a basis for giving and receiving social support. Results from a cross-sectional study show that retirees (N = 171) who had multiple social identities following (but not prior to) retirement report being (a) more satisfied with retirement, (b) in better health, and (c) more satisfied with life in general. Furthermore, mediation analyses revealed an indirect path from multiple social identities to greater satisfaction with retirement and better health through greater provision, but not receipt, of social support to others. These findings are the first to point to the value of multiple group membership post-retirement as a basis for increased opportunities to give meaningful support to others. We discuss the theoretical and practical implications for the management of multiple identities in the process of significant life transitions such as retirement.

#### Edited by:

Soledad De Lemus, University of Granada, Spain

#### Reviewed by:

Katherine Stroebe, University of Groningen, Netherlands Antonio Bustillos, National University of Distance Education, Spain

> \*Correspondence: Niklas K. Steffens

n.steffens@uq.edu.au

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

Received: 23 May 2016 Accepted: 20 September 2016 Published: 17 October 2016

#### Citation:

Steffens NK, Jetten J, Haslam C, Cruwys T and Haslam SA (2016) Multiple Social Identities Enhance Health Post-Retirement Because They Are a Basis for Giving Social Support. Front. Psychol. 7:1519. doi: 10.3389/fpsyg.2016.01519 Keywords: multiple identities, retirement, group membership, social identity, health, well-being, life satisfaction, social support

# INTRODUCTION

Adjustment to retirement is a major challenge. Not least, this is because for many people retirement involves giving up a work identity with which they are highly engaged, and which provides them with a sense of respect and self-worth. In this context, it is not surprising that there is a lot of variation in people's adjustment to retirement and that a substantial number of retirees report compromised health and well-being in the transition to retirement (Wang, 2007; Wang and Shi, 2014). Nevertheless, such findings beg the question of how best to manage the transition to retirement so as to allow people to maintain, and possibly enhance, their health and wellbeing. Speaking to this issue, the social identity approach suggests that individuals' social group memberships — and the sense of 'we' and 'us' (or social identities) that they derive from these are a basis for optimizing health and well-being (Haslam et al., 2009; Jetten et al., 2012; Tewari et al., 2012; Cruwys et al., 2014; Sani et al., 2015; Steffens et al., 2016c). Moreover, while a focus on social group memberships provides a useful theoretical framework for understanding why retirement puts people at risk (because it entails losing valued group memberships), it also points to potential ways to promote successful adjustment — namely, compensating for the loss of work group membership by strengthening memberships in other social groups post-retirement.

In the present research we draw on findings in the retirement, social identity, and social support literatures and propose that retirees' multiple identities — that is, the number of social groups that they regard as self-defining — are associated with better adjustment and greater well-being. More

specifically, we propose that having more (rather than few) social identities post-retirement can have benefits for health because these afford opportunities to make meaningful contributions to the lives of others through the provision of social support.

# Multiple Identities and Retirement Adjustment

Retirement constitutes a major life change and involves the loss of an activity that people have typically engaged in for a large part of their lives. Given the scale of the transition, it has the capacity to alter (for good or ill) how people live and feel (Pinquart and Schindler, 2007; Wang, 2007). To understand this process, scholars have focused on examining the factors that influence people's ability to adjust to retirement (Coe and Zamarro, 2011; Wang and Shi, 2014). Specifically, evidence suggests that adjustment tends to be more successful where people engage in financial decision-making (Topa et al., 2009; Adams and Rau, 2011; Feldman and Beehr, 2011) and in planning the timing of their exit from work (Litwin, 2007; Brockmann et al., 2009; Wu et al., 2016), as well as participating in post-retirement leisure activities (Dorfman and Kolarik, 2005; Nimrod, 2007) and physical exercise (Lord et al., 2003; Nimrod et al., 2008). Yet despite considerable attention to these factors in retirement planning, a substantial number of people fail to adjust successfully (Wang, 2007). So there is clearly more to the story.

In addition to these factors, we propose that our capacity to adjust to retirement is also likely to be impacted by our social connectedness to others, and particularly to groups of others. Indeed, this proposition follows from the Social Identity Model of Identity Change (SIMIC; Jetten et al., 2009) which provides a way of conceptualizing the role that group memberships, and the social identities that underpin them, play in life changes such as retirement. This suggests that belonging to, and maintaining membership in, multiple social groups helps protect people's health and well-being in the face of life transitions. In line with this general proposition, empirical evidence shows that belonging to multiple groups has health-related benefits for those experiencing a range of life-changing events such as starting life as a university student (Iyer et al., 2009), recovering from a stroke (Haslam et al., 2008), and adjusting to life following brain injury (Jones et al., 2011).

More directly relevant to our present concerns, there is evidence that membership of multiple social groups also reduces the risk of premature death in retirees. Speaking to this issue, Steffens et al. (2016a) investigated the contribution of social group ties to mortality in the 6 years following retirement in a nationally representative sample of over 400 individuals in England. One of the study's most striking findings was that retirees who maintained multiple group ties during this transition were far less likely to die prematurely. More specifically, if people had two group memberships prior to retirement, they had only a 2% risk of early death if they maintained both, but this increased markedly if they lost one or both groups (to 5% and 12%, respectively). Importantly, these results controlled for factors that might otherwise be expected to restrict a person's ability to maintain group memberships (e.g., physical health and age). Furthermore, the study also found that 6 years post-retirement life satisfaction was 10% lower for every group membership that people lost but did not replace.

Yet while this study demonstrates the importance of multiple social groups for people in retirement, it was not able to explore the contribution of self-identified groups to the observed relationships (i.e., those which are important and central to a person's self-concept). This is relevant to the core tenet of self-categorization theory — that groups determine people's psychology to the extent that they self-categorize and see themselves as part of a given group. Accordingly, attempting to index this construct through a list of pre-determined groups identified by researchers (rather than participants) will often fail to fully capture the diverse range of possible social groups to which people feel a psychological sense of belonging (for a more detailed discussion, see Cruwys et al., 2016). In particular, this approach may (a) miss groups that people see themselves as members of, but also (b) enhance the salience of a particular group by virtue of mentioning it, or (c) change the content and number of groups with which people identify in response to being primed with a list. It follows that any estimations based on pre-selected groups will tend to provide an inaccurate measure of multiple group membership and, ultimately, an inaccurate test of theoretical assertions related to this construct. In contrast, eliciting details of groups that participants generate themselves allows for a more accurate operationalization of the theoretical construct (than pre-determined groups) and, in turn, should provide a better index with which to test relationships of theoretical interest.

Even more important is the fact that this research was not able to examine the mechanisms that might underlie the beneficial effects of multiple identities. In particular, while there is strong evidence that social support is likely to be a critical mediator of the relationship between social identification and well-being (for a review, see Haslam et al., 2012), the role that this plays in helping retirees with multiple identities adjust to, and maintain their health in, retirement remains to be tested. The present research addresses both these limitations. In so doing it also simultaneously examines the degree to which both the provision and the receipt of support are implicated in the health-protecting benefits of multiple identities.

# Multiple Identities, Provision of Social Support, and Health

Theoretical and empirical research have suggested that self-categorization (and social identification) with a social group provides access to important sources of social support (Wiesenfeld et al., 2001; Levine et al., 2002, 2005; Reicher et al., 2006; Haslam et al., 2012). Added to this, research has demonstrated that this feeling of being socially supported is a key factor that explains (i.e., mediates) the effect that identification with a group has on people's health. In this regard, there is abundant evidence of an indirect effect of social identification on health through received social support — both in groups in organizational contexts (van Dick et al., 2004; Haslam et al., 2005; Haslam and Reicher, 2006; McKimmie et al., 2009; Avanzi et al., 2015; Ketturat et al., 2016) as well as in leisure and other nonwork group contexts (Branscombe et al., 1999; Crabtree et al.,

2010). Yet, it has been suggested that if one group is beneficial for social support, then more groups are likely to further enhance the experience of social support. Indeed, the same pathway from identity to health via received social support has also been found in studies of multiple identities (Haslam et al., 2016b; Walter et al., 2016). From this plethora of studies, we can thus conclude that identification with (single and multiple) social groups is associated with better health and well-being in part because it provides people with a basis for being supported, and, more particularly, perceiving that they are supported, by others.

Yet while receiving support may be important in accounting for the link between social identity and health, another body of research indicates that it is often not the act of receiving but the act of giving support that is important in promoting health. Along these lines, evidence indicates that individuals' health and wellbeing increase when they give support to others — be it in the form of (a) money (Dunn et al., 2008, 2010; Aknin et al., 2009, 2013), (b) time (Wheeler et al., 1998; Thoits and Hewitt, 2001; Campbell et al., 2009; Sarid et al., 2010; Choi and Kim, 2011), or (c) emotional help (Brown et al., 2003; Abolfathi Momtaz et al., 2014). Indeed, Brown et al.'s (2003) seminal study which examined older married couples (all males were older than 65 years) over a period of 5 years — found that giving support to one's spouse was more important in reducing mortality than receiving support from them (see also Abolfathi Momtaz et al., 2014). This then also suggests that changes in people's lives that result in the loss of opportunities to give support (e.g., to engage in volunteering) can have negative health and well-being implications (e.g., by lowering people's life satisfaction; Meier and Stutzer, 2008). Furthermore, it is noteworthy that the healthpromoting effects of giving are particularly pronounced among older adults (Van Willigen, 2000; Musick and Wilson, 2003). For example, Greenfield and Marks (2004) found that the well-being of older adults who suffered from a loss of a role identity benefited to a great extent (and more than their younger counterparts) from volunteering their time to help others.

Drawing on the findings from these literatures, we argue that multiple identities in retirement are likely to be associated with enhanced health not only because they provide a basis for receiving support but also because they provide opportunities to give support. Indeed, of these two forms of social support, one might expect that the provision of social support will be especially protective in light of evidence that older adults derive greater health and well-being benefits from giving support than receiving it.

# The Present Research

In the present research, we investigate the link between multiple identities, retirement satisfaction, subjective quality of life, and subjective health in the transition to retirement. We hypothesize that multiple identities will be associated with retirees' (a) better adjustment to the retirement transition (H1a), (b) better subjective health (H1b), and (c) greater subjective quality of life (H1c). We also examine these associations as they relate to people's multiple identities before and after retirement in order to better understand the contribution of identities to health and well-being outcomes in the course of the transition. In line with findings by Steffens et al. (2016a) — who found that social group memberships following, but not prior, to retirement impacted mortality — we expect that multiple identities following retirement will be more important than those prior to retirement (H2). Furthermore, we examine the contribution that the nature of social support (i.e., giving and receiving) makes to the relationship between multiple identities and these three outcomes. In light of previous research, we hypothesize that not only receiving (H3) but also giving (H4) social support will mediate the effects of multiple identities on (a) retirement satisfaction, (b) subjective health, and (c) subjective quality of life.

# MATERIALS AND METHODS

# Participants and Procedure

The study received ethics approval by the Behavioral and Social Sciences Ethical Review Committee at the first author's university (Ref: 2012001231). We recruited a sample of 171 (71 female; 98 male; 2 undisclosed) retired individuals residing in Australia. Participants were recruited via different organizations that distributed an online survey to individuals who had recently retired (i.e., individuals who had stopped work and were not undertaking bridging or part-time work). This included the Ageing Mind Initiative's [AMI] (2016) 50+ Registry, which has a database of people over 50 years who are willing to take part in research projects on ageing, and the Online Research Unit [ORU], 2016, the largest professional online research organization in Australia. Participants had an average age of 67.38 years (SD = 4.45) and they had been retired for an average of 3.63 years (SD = 3.86). Participants were invited to participate in a study titled "Transition to Retirement Survey." After providing their informed consent, participants responded to the survey measures as indicated below. For all scales, unless stated otherwise, participants indicated their level of agreement with each item on scales ranging from 1 (not at all) to 7 (completely). Following completion of the survey, participants were debriefed and thanked for their involvement in the study.

# Measures Multiple Identities

Participants used a multiple group membership listing task (Haslam et al., 2008) to identify the social groups that they belonged to and that they saw as important to their sense of self both pre- and post-retirement. Participants were given examples of categories and types of social groups (e.g., leisure or social groups, which could include book and gardening clubs; community groups, which could include church groups; workgroups which could include work teams), before being asked to provide up to a maximum of six social groups to which they belonged separately for the period before and after retirement. On average, participants reported being members of three groups prior to retirement (M = 2.99; SD = 1.89) and following retirement (M = 3.08; SD = 1.88). For each group that they identified, they also indicated on a scale ranging from 1 (not important) to 7 (very important) how important each group was to them.

#### Received Social Support

fpsyg-07-01519 October 14, 2016 Time: 18:55 # 4

Participants responded to four items assessing their received social support (α = 0.92; adapted from Jetten et al., 2012; "People that are important to me help me with my tasks"; "People that are important to me provide me with emotional support"; "People that are important to me listen to me if I need to talk"; "People that are important to me make me feel loved and cared for").

#### Provided Social Support

We adapted the same four items that were used to measure received social support to assess provided social support (α = 0.86; adapted from Jetten et al., 2012): "I help people that are important to me with their tasks," "I provide emotional support to people that are important to me," "I listen to those people that are important to me if they need to talk," and "I make those people that are important to me feel loved and cared for."

## Satisfaction with Retirement

Participants responded to three items assessing their satisfaction with retirement, adapted from Spector's (1997) Job Satisfaction Scale (α = 0.84). These items were: "In general, I am satisfied with being retired," "In general, I don't like being retired" [reversecoded], and "In general, I like the fact that I am retired."

#### Subjective Health Status

Participants responded to three items assessing their subjective health status (α = 0.97; adapted from Eriksson et al., 2001). These were: "At the moment, my health is very good," "At the moment, my health is in an excellent condition," and "At the moment, I feel very healthy."

### Subjective Quality of Life

Quality of life is a multi-dimensional construct that comprises both subjective and objective elements (Fernández-Ballesteros, 2011). In line with points made in the Introduction, in the present research we focused on subjective quality of life and used Diener et al. (1985) Satisfaction with Life Scale to assess this construct (α = 0.89). Items were as follows: "In most ways my life is close to ideal," "The conditions of my life are excellent," "I am satisfied with life," "So far I have gotten the important things I want in life," and "If I could live my life over, I would change almost nothing."

# RESULTS

# Main Analyses

Means, standard deviations (SDs), and zero-order correlations between measures are presented in **Table 1**. We ran a series of hierarchical linear regressions in which we entered multiple identities pre-retirement at Step 1 and then multiple identities post-retirement at Step 2 as predictors of satisfaction with retirement, subjective health status, and subjective quality of life.

#### Satisfaction with Retirement

As shown in **Table 2**, supporting H1a, linear regression analyses indicated a significant effect of multiple identities pre-retirement at Step 1 (β = 0.17, p = 0.040) that accounted for 3% of the variance in satisfaction with retirement. When entering multiple identities post-retirement as an additional predictor at Step 2, multiple identities pre-retirement was no longer a significant predictor (β = 0.03, p = 0.776) and only multiple identities post-retirement significantly predicted satisfaction in retirement (β = 0.21, p = 0.038). This suggests that multiple identities prior to retirement is a significant predictor on its own but it ceases to be a significant predictor once one controls for the shared variance between multiple identities following retirement and retirement satisfaction (supporting H2). Furthermore, it is noteworthy that the variance accounted for by the model that included both predictors (at Step 2) increased significantly to a total of 6%.

#### Subjective Health Status

Consistent with H1b, multiple identities pre-retirement was a significant predictor of subjective health status at Step 1 (β = 0.24, p = 0.003), accounting for 6% of the variance. However, it was not a significant predictor at Step 2 when both multiple identities prior to and following retirement are entered (β = 0.10, p = 0.338). Supporting H2, at Step 2 only multiple identities postretirement was a significant predictor (β = 0.22, p = 0.028) and this model accounted for a greater proportion (a total of 9%) of the variance in subjective health status.

#### Subjective Quality of Life

Supporting H1c, analysis revealed a marginally significant effect of pre-retirement multiple identities at Step 1 (β = 0.16, p = 0.053) that explained 2% of the total variance. At Step 2, and consistent with H2, pre-retirement multiple identities were no longer a significant predictor (β = −0.01, p = 0.989) but post-retirement multiple identities were (β = 0.25, p = 0.017). Together the predictors explained an even greater share (i.e., 6%) of the total variance. In the context of the transition, these results indicate that if a retiree had three group memberships prior to retirement but then lost two group memberships (so that they now had one) their mean subjective quality of life was 5.40 (on a 7-point scale). However, a retiree's mean subjective quality of life was 6.05 if they maintained three group memberships after retirement and 6.70 if they gained two group memberships (so that they now had five). These results are presented in **Figure 1**.

#### Mediation via Received and Provided Social Support

We examined the extent to which the relationship between multiple identities and satisfaction with retirement was mediated by received and provided social support. For this purpose, we ran bias-corrected bootstrapping multiple mediation analyses with 5000 resamples using PROCESS (Model 4; Hayes, 2013). Multiple mediation analysis has the advantage that it can test the indirect effect through given and provided social support, while simultaneously controlling for the impact of the other variable.

Results indicated that the indirect path from multiple identities post-retirement to satisfaction with retirement through provided social support was significant, γ = 0.04, SE = 0.03,


TABLE 1 | Means, standard deviations, and intercorrelations between variables.

†p < 0.10, <sup>∗</sup>p < 0.05, ∗∗p < 0.01; N = 159–171.

fpsyg-07-01519 October 14, 2016 Time: 18:55 # 5

95%CIs [0.01,0.11], while the indirect path through received social support was not significant, γ = −0.01, SE = 0.02, 95%CIs [−0.07,0.02], R 2 Model = 0.10, 1R 2 AdditionofMediators = 0.03, providing support for H4a (but no support for H3a). These mediation results are displayed in **Figure 1**.

Furthermore, supporting H4b, results for subjective health status showed a significant indirect path from multiple identities post-retirement through provided social support, γ = 0.05, SE = 0.03, 95%CIs [0.01,0.13]. However, they revealed a non-significant indirect path through received social support, γ = 0.01, SE = 0.02, 95%CIs [−0.03,0.07], R 2 Model = 0.19, 1R 2 AdditionofMediators = 0.09, providing no support for H3b.

Finally, and as shown in **Figure 2**, results supported H4c by showing that the indirect path from multiple identities post-retirement to quality of life through provided social support was significant, γ = 0.05, SE = 0.03, 95%CIs [0.01,0.13]. At the same time, there was no support for H3c in so far as the indirect path through received social support was non-significant, γ = 0.01, SE = 0.02, 95%CIs [−0.03,0.07], R 2 Model = 0.23, 1R 2 AdditionofMediators = 0.14. Consistent with H4, these mediation analyses thus indicate that the relationship between multiple identities and health and well-being outcomes was not mediated by retirees' receipt of support but was mediated by their provision of it to others.

In addition, we examined whether received social support was a significant mediator when examined on its own (without controlling simultaneously for provided social support). This analysis indicated that the indirect path from multiple identities post-retirement to satisfaction with retirement through received social support was significant, γ = 0.02, SE = 0.02, 95%CIs [0.00,0.07], R 2 Model = 0.08, 1R 2 AdditionofMediator = 0.02. Furthermore, results revealed that the indirect effect of multiple identities post-retirement through received social support was

TABLE 2 | Hierarchical regression analyses: Multiple identities (pre- and post-retirement) predicting satisfaction with retirement, subjective health, and quality of life.


†p < 0.10, <sup>∗</sup>p < 0.05, ∗∗p < 0.01; Multiple Identities (Pre) and (Post) refer to multiple identities (number of social group memberships) prior to and following retirement, respectively.

also significant in the analysis for subjective health status, γ = 0.05, SE = 0.02, 95%CIs [0.01,0.10], R 2 Model = 0.13, 1R 2 AdditionofMediator = 0.05, and subjective quality of life, γ = 0.06, SE = 0.03, 95%CIs [0.02,0.13], R 2 Model = 0.20, 1R 2 AdditionofMediator = 0.11. In sum, results indicate that, consistent with previous research, received social support is a significant mediator of the link between multiple identities and health when examined on its own. In addition, results also indicate that in the present sample the link between multiple identities and health through received social support is accounted for by the shared variance with provided social support.

## Sensitivity Analyses

We conducted four further tests to establish the robustness of the patterns observed above. First, it is possible that the effects of possessing more social identities might be linear but also that they saturate such that once a person has many rather than few groups any additional group might provide fewer health benefits. To address this question, we examined whether there was evidence of a curvilinear relationship between multiple identities and dependent variables in order to examine whether the benefits of each additional social group membership decreases as the number of social group memberships increases. This involved computing the quadratic term of multiple identities following retirement (to reduce collinearity between predictors, multiple identities were Z-standardized prior to calculating the quadratic term) and then running a series of linear regression analyses in which the (Z-standardized) linear term was added at Step 1 and the quadratic term was added as a predictor at Step 2. Analysis revealed that the quadratic term was unrelated to, and did not account for additional variance beyond that accounted for by the linear term in the prediction of, satisfaction with retirement, b = −0.03, 95%CIs [−0.26,0.20], β = −0.02, 1R <sup>2</sup> < 0.001, p = 0.780, subjective health status, b = −0.21, 95%CIs [−0.45,0.02], β = −0.13, 1R <sup>2</sup> = 0.018, p = 0.075, and subjective quality of life, b = −0.10, 95%CIs [−0.30,0.11], β = −0.72, 1R <sup>2</sup> = 0.005, p = 0.364. In sum, there was no evidence of a curvilinear relationship between multiple identities and dependent variables, suggesting that there were no diminishing marginal returns associated with multiple identities and that the relationships never become negative.

Second, to gain insight into the directionality of the relationships in our focal mediation model, we ran additional sensitivity analyses examining the reversed mediation from better health via social support to multiple identities. Results indicated that none of the reversed indirect paths from health measures (as independent variables) via received and provided social support to multiple identities post-retirement (as the dependent variable) were significant. More specifically, the (reversed) indirect paths to multiple identities post-retirement through both received and provided social support were not significant when examining (a) satisfaction with retirement as the independent variable (received support: γ = 0.03, SE = 0.03, 95%CIs [−0.01,0.12]; provided support: γ = 0.01, SE = 0.03, 95%CIs [−0.06,0.08], R 2 Model = 0.10, 1R 2 AdditionofMediator<sup>s</sup> = 0.03), (b) subjective health status as the independent variable (received support: γ = 0.05, SE = 0.05, 95%CIs [−0.02,0.16]; provided support: γ = −0.01, SE = 0.05, 95%CIs [−0.11,0.10], R 2 Model = 0.11, 1R 2 AdditionofMediators = 0.02), or (c) subjective quality of life as the independent variable (received support: γ = 0.07, SE = 0.07, 95%CIs [−0.05,0.21]; provided support: γ = 0.01, SE = 0.07, 95%CIs [−0.12,0.15], R 2 Model = 0.10, 1R 2 AdditionofMediators = 0.02).

Finally, we ran additional analyses that included the number of highly important groups (rated above the mid-point of the scale — i.e., 5 or higher on the 1 to 7 scale). Results indicated that the number of highly important identities post-retirement (while controlling for those pre-retirement) were positively associated with all health outcomes (predicting a similar or a slightly greater proportion of the variance across the variables): (a) satisfaction with retirement, β = 0.22, p = 0.022, R 2 Model = 0.05, (b) subjective health status, β = 0.34, p < 0.001, R 2 Model = 0.13, and (c) subjective quality of life, β = 0.27, p = 0.004, R 2 Model = 0.08. This confirms the point that it is people's sense of psychological connectedness with groups that determines health outcomes.

# DISCUSSION

In the present research, we examined the extent to which retirees' multiple identities — both before and after the transition to retirement — are associated with their adjustment to retirement and their health. Results show that the more social group memberships retirees had following retirement, (a) the more satisfied they were with retirement, and (b) the better their subjective health and (c) the better their subjective quality of life (supporting H1). Furthermore, evidence indicated that these health benefits primarily derived from multiple identities following (but not prior to) retirement (supporting H2). The present findings demonstrate that being less well connected before retirement need not be detrimental as people can still

experience good health in retirement to the extent that they develop and build multiple social identities.

Sensitivity analyses also indicated that there was no evidence of a curvilinear relationship between multiple identities and retirement satisfaction and health, suggesting that there were no diminishing marginal returns of multiple identities and, more particularly, that the relationships never become negative. Finally, speaking to the mechanism through which multiple identities protect health and well-being in retirement, results revealed an indirect effect through giving social support but not through receiving social support (when controlling simultaneously for giving social support). Specifically, retirees who had multiple identities indicated that they provided more social support to others and this in turn was associated with greater satisfaction with retirement, better health, and greater subjective quality of life (providing support for H4).

The present findings make at least three important contributions to the literatures on retirement, multiple identities, and health. First, we are only just starting to understand how social group memberships are implicated in people's ability to adjust to a life change that most individuals experience at some point in their lifetime — namely, retirement (Steffens et al., 2016a). Yet this prior research provided limited insight into the extent to which multiple identities impact retirees' adjustment and health because the research (a) relied on groups whose specification was fixed and pre-determined, and (b) did not consider retirement satisfaction (as a direct indicator of retirees' psychological adjustment to retirement). Our results extend this research and provide more conclusive support for the idea that individuals' multiple identities — subjective group memberships that they self-identify as being part of who they are — are associated with successful retirement. Retirees who felt connected to a greater number of social groups not only had

greater subjective quality of life but also were more satisfied with retirement and experienced better health.

Second, our results shed some light on the processes through which multiple identities are associated with better health in retirement. Previous research that explored these issues has shown that social identities are health-enhancing because they provide people with a sense that they are supported by other people (Haslam et al., 2005, 2016b; Crabtree et al., 2010; Avanzi et al., 2015; Walter et al., 2016). However, the present study extends our understanding of the various ways in which multiple identities are associated with health. Specifically, by examining the provision and receipt of social support simultaneously, our findings indicate that retirees' multiple identities indirectly impacted their satisfaction with retirement and health primarily because these are a basis for providing social support. This is not to say that receiving social support is unimportant for health and well-being. Indeed, if one does not control for provided social support, received social support is a significant mediator. Nevertheless, in line with previous research (Brown et al., 2003; Abolfathi Momtaz et al., 2014), it appears that, when examined simultaneously, the provision of social support explains more variance in health than its receipt, and that provision of social support is a particularly powerful mechanism in the multiple identities– health link. More generally, then, this suggests that in lifechanging contexts such as retirement, where individuals are stripped of one important way to contribute to society, multiple identities provide a vehicle for continuing to make a significant contribution to the lives of others and thereby to sustain a healthy life.

Third, the present results suggest that the process of retirement might be enhanced through facilitation of retirees' social group-based connections to others. Thus, in addition to making financial, residential, medical care, and physical activity plans, our findings suggest that it is also important for retirees to plan how they might stay (or become) socially connected to groups that can provide them with a sense of communality and belonging. Moreover, the fact that beneficial effects were strongest for retirees who had a greater number of social identities post-retirement suggests that these plans need not be constrained by a person's social circumstances prior to retirement. For, as the present data show, building one's network of social group memberships after retirement can bring health and well-being gain. While this network may be more difficult to develop and embed for those less well connected before they retire, there is evidence that they can be strengthened by social interventions that explicitly target social disconnection (e.g., GROUPS 4 HEALTH; Haslam et al., 2016a). There is clearly mileage in future research seeking to establish whether such an intervention — whose goal is explicitly to build and maintain multiple social identities — can be successfully adapted to promote the health and well-being of retirees.

# Limitations and Future Research

The present research is not without limitations. Primary amongst these is the fact that, because the study has a cross-sectional design it cannot establish causality in the associations investigated here. Future research should therefore employ longitudinal and experimental (intervention) studies in order to shed light on directionality (as do Steffens et al., 2016a). In this regard it is worth noting, though, that we do not rule out the possibility that health may also increase people's willingness to provide social support — even though in the present study there was no evidence of this reversed mediation. Indeed, research on the link between giving and happiness shows that this link is likely to be mutually reinforcing (bidirectional) such that giving increases happiness, which in turn, increases giving (Thoits and Hewitt, 2001; Aknin et al., 2012). Nevertheless, it would clearly be useful for long-term panel studies to shed light on this issue by exploring the strength of these different pathways. In this regard too, we cannot exclude the possibility that part of the variance may be explained by individual differences (for example, people who are cognitively more flexible or more open to experience may be more likely to join new groups in retirement). Having said this, research from other fields shows that effects of multiple group memberships are not reducible to stable individual differences. In this regard, an experimental study by Jones and Jetten (2011) showed that reflecting on many, rather than few, group memberships led to subsequent increases in resilience, while an intervention study by Haslam et al. (2016a) that focused on the maintenance and development of social group memberships produced improvement in participants' health. Furthermore, Steffens et al. (2016b) controlled for the Big Five in their studies that examined the link between multiple social identities and creativity and found that results could not be reduced to differences in personality. Nevertheless, it is possible that retirees' willingness and capacity to develop new multiple identities may in part be impacted by other variables, a possibility that future research should examine.

Moreover, it is certainly the case that one's social life involves not only group-based social ties but also those with significant individuals, which have been shown to be important in adjusting to retirement. In this context, spousal relationships have been highlighted as particularly important (Kupperbusch et al., 2003; Davey and Szinovacz, 2004; Bushfield et al., 2008), but as these were not the focus of the present research, we cannot comment on the contribution that (particular) individual relationships make relative to group-based relationships. Interestingly, though, relationships that people have with significant others tend not to be confined to one-on-one interaction, but often occur in the context of wider group-based interaction (with the family in the case of a spouse, or a work team in the case of a professional colleague). While we are not discounting the important role that one-on-one relationships have for health and well-being, there is a growing body of evidence from a variety of (healthy, vulnerable, and clinical) populations which indicates that social connectedness to groups is a unique and reliable predictor of several health outcomes (for reviews, see Cruwys et al., 2014; Jetten et al., 2014; Steffens et al., 2016c). Along these lines, Sani et al. (2012) have shown that health effects derive from feeling oneself to be

part of a social group and not merely from one's amount of social contact. Furthermore, Haslam et al. (2014) have shown that it is group ties and not one-on-one ties that predict subsequent cognitive health among older adults, and Glei et al. (2005) have found that group-based activities are more important than individual engagement with significant others (with a spouse, close relative, friend) in protecting cognitive health over time. Having said this, further research is needed to disentangle and quantify the extent to which retirees' health benefits derive from different aspects of their social group memberships as well as other aspects of their social life. For instance, future research should examine the extent to which adjustment to retirement is impacted by the perceived compatibility between the multiple groups that retirees are part of (Iyer et al., 2009) as well as the sense of continuity in their group memberships and associated sense of self (Sani, 2008; Sani et al., 2008).

Furthermore, there would also be value in future work that expands the scope of the potential outcomes that flow from having multiple identities. For example, it would be interesting to investigate not only retirees' sustained mental and physical health but also other aspects of an engaged healthy life such as cognitive function (e.g., memory and creativity; Haslam et al., 2014; Steffens et al., 2016b) and resilience (e.g., the capacity to bounce back from significant set-backs; Jones et al., 2011; Cruwys et al., 2013). Similarly, it would be worthwhile disentangling the active ingredients in the provision of social support that are associated with better health and well-being. In this regard, it is noteworthy that the association between social support and the various health outcomes was overall of moderate magnitude, while the association between multiple identities and both forms of social support was only in the weak to moderate range (rs = 0.16 and 0.23). To provide a better understanding of these relationships, future research might shed light on the extent to which (a) multiple identities are related to different aspects of provision of support, such as one's motivation, actual supportive behaviors, or the impact that giving has on others (Bolino and Grant, 2016) and (b) these different elements have benefits for people's health and wellbeing.

Finally, in the present research we focused on subjectively experienced (and reported) support and health (rather than objective indicators of support and health) and future research might extend the present examination by including other more objective indicators of these measures (see also Fernández-Ballesteros, 2011). It is important to note though, that people are generally capable of reporting meaningfully the way they experience their health and reviews indicate that reported quality of life and subjective health status are powerful predictors of a range of other third-party rated and objective health measures (including mortality; Pavot et al., 1991; Idler and Benyamini,

# REFERENCES

1997). Accordingly, while previous research might lead one to expect weaker effects for objective physical than subjective psychological health measures (e.g., see Steffens et al., 2016c), there are no grounds for expecting the pattern of findings to be substantively different from that observed in the present research.

# CONCLUSION

The present research investigated how multiple identities in the transition to retirement are associated with successful adjustment, as well as health and well-being. Our results show that multiple identities following retirement are associated not only with greater satisfaction with retirement but also with better subjective health and enhanced subjective quality of life. Furthermore, the study is the first to demonstrate that the health-protective benefits of having multiple identities arise in part from the fact that they give people more opportunity to contribute to the lives of others through the provision of social support. This observation is important for theoretical reasons but also has the potential to inform practical interventions — in pointing to the importance of cultivating multiple social identities among retirees with a view to fostering their resilience.

Together, these various findings provide a new dimension to our understanding of the importance of networks of social group memberships in retirement. In particular, they suggest that, in the process of giving up membership in work-related group(s), access to multiple social identities allows people to continue to be active members of other valued communities and thereby to continue to have meaningful and healthy lives. Indeed, in ways that would surely please John F. Kennedy, we can see that, in retirement, multiple group memberships provide a basis for people to reap the benefits not of what others can do for them but of what they can do for others.

# AUTHOR CONTRIBUTIONS

NS, JJ, CH, TC, and AH developed the study concept and designed the research. NS performed the statistical analyses. NS drafted the manuscript. All authors edited the manuscript. All authors read and approved the final manuscript.

# FUNDING

This work was supported by three grants from the Australian Research Council awarded to JJ (FT110100238), CH (DP160102514), and AH (FL110100199).


Abolfathi Momtaz, Y., Ibrahim, R., and Hamid, T. A. (2014). The impact of giving support to others on older adults' perceived health status. Psychogeriatrics 14, 31–37. doi: 10.1111/psyg.12036



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Steffens, Jetten, Haslam, Cruwys and Haslam. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The More (Social Group Memberships), the Merrier: Is This the Case for Asians?

Melissa X.-L. Chang\*, Jolanda Jetten, Tegan Cruwys, Catherine Haslam and Nurul Praharso

*School of Psychology, The University of Queensland, Brisbane, QLD, Australia*

While previous studies have consistently shown that belonging to multiple groups enhances well-being, the current research proposes that for Asians, multiple group memberships (MGM) may confer fewer well-being benefits. We suggest that this is due, in part, to Asian norms about relationships and support seeking, making Asians more reluctant to enlist social support due to concerns about burdening others. Overall, MGM was associated with enhanced well-being in Westerners (Study 2), but not Asians (Studies 1–3). Study 2 showed that social support mediated the relationship between MGM and well-being for Westerners only. In Study 3, among Asians, MGM benefited the well-being of those who were least reluctant to enlist support. Finally, reviewing the MGM evidence-base to date, relative to Westerners, MGM was less beneficial for the well-being of Asians. The evidence underscores the importance of culture in influencing how likely individuals utilize their group memberships as psychological resources.

#### Edited by:

*Clara Kulich, University of Geneva, Switzerland*

#### Reviewed by:

*Rui Zhang, Dickinson College, USA Tomoko Ikegami, Osaka City University, Japan*

\*Correspondence: *Melissa X.-L. Chang xue.chang@uqconnect.edu.au*

#### Specialty section:

*This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology*

> Received: *19 March 2016* Accepted: *17 June 2016* Published: *12 July 2016*

#### Citation:

*Chang MX-L, Jetten J, Cruwys T, Haslam C and Praharso N (2016) The More (Social Group Memberships), the Merrier: Is This the Case for Asians? Front. Psychol. 7:1001. doi: 10.3389/fpsyg.2016.01001* Keywords: multiple group membership, social support, well-being, culture, social identity

# INTRODUCTION

We know that social relationships matter for psychological well-being. The research shows that people who belong to more social groups have better psychological well-being than those with fewer social group memberships (Thoits, 1983; Helliwell, 2003; Brook et al., 2008; Haslam et al., 2008; Cruwys et al., 2013; Jetten et al., 2015; Sani et al., 2015). This is due, in part, to social groups providing an important means to access psychological resources such as social support (see Jetten et al., 2014). Social support, in turn, has been found to be associated with mental health benefits that include better adjustment to stressful events (House et al., 1988; Thoits, 1995). It follows logically then, that the more groups an individual belongs to, the more potential social support resources they have at their disposal when encountering challenges or stressors in their life.

Although much of the empirical work on multiple group membership and well-being has generated important insights, these studies have been conducted primarily in Western societies and it remains to be examined whether the findings generalize to non-Western cultures (Henrich et al., 2010). Put simply, will "the more the merrier" effect, in the case of multiple groups, hold for Asians? There are good reasons to believe that such findings may not easily generalize to Asian populations. For example, studies have suggested that cultural norms about relationships in the Asian cultures could make Asians more sensitive to the negative relational consequences of support seeking when compared to European Americans (e.g., burdening others, disrupting group harmony, Taylor et al., 2004; Kim et al., 2006, 2008). To the extent that this applies, multiple group memberships (MGM) may not be associated with greater support seeking in times of difficulty and stress for Asians. These cultural differences in the role that multiple group memberships play raise questions about the nature of its relationship with psychological well-being in Asian cultures. Addressing this issue, the present research examines the role of MGM in Asian and Western contexts.

# Multiple Group Memberships and Well-Being—"The More the Merrier"

The Social Identity Approach (SIA; Haslam et al., 2009; Jetten et al., 2012, 2014) provides an explanation for the relationship between MGM and well-being. Fundamental to this approach is the idea that social group memberships are critical in initiating a shared sense of identification. When social groups are perceived as meaningful and relevant to characterizing the self, they become psychologically internalized and help to understand the self and one's place in the world (Turner et al., 1987). Hence, social groups have the power to furnish people with a sense of themselves as part of a larger collective ("us") rather than as merely unique individuals ("I"; Turner, 1982).

Accordingly, when one incorporates groups of others (e.g., one's family, friendship, community, and recreational groups) into one's sense of self, one will feel psychologically connected with those others, such that their interests become one's own (Haslam et al., 2012). More importantly still, such internalized group membership provides a meaningful basis to receive and benefit from various forms of social support (Cohen and Wills, 1985; Underwood, 2000; Postmes and Branscombe, 2002; Haslam, 2004; Haslam et al., 2009). This means that it is only when people perceive themselves to share a mutual sense of common group membership with others that they are more likely to give, and be open to receiving, support and other resources from them; facilitating constructive helping between individuals. Applying this logic, social groups should only enhance positive social support and well-being when (and to the extent that) individuals identify strongly with them. A large body of empirical work has substantiated this by showing that shared identity is indeed what makes social support possible and effective (Haslam, 2004; Haslam et al., 2005, 2012; Levine et al., 2005). For illustration, in a study by Haslam et al. (2005), it was found that shared group memberships had a positive impact on well-being among hospital patients recovering from heart surgery because such group memberships served as a basis for the receipt of effective support from others.

Following from this, the SIA posits that if group membership serves as a psychological resource (e.g., providing a basis for social support), then it is likely that MGM should enhance this resource (Haslam et al., 2008; Cruwys et al., 2014; Jetten et al., 2015). Individuals who belong to multiple groups are therefore likely to have more potential sources from which to draw social support in times of difficulty and stress, in turn protecting and enhancing their well-being (Iyer et al., 2009; Jetten et al., 2009, 2014). Indeed, several studies have consistently shown that having more group memberships (e.g., belonging to family, friendship, community, recreational groups) is associated with greater psychological well-being.

For example, in one study conducted with individuals who had recently experienced a stroke, Haslam et al. (2008) showed that life satisfaction and well-being were higher for those who belonged to more social groups before their stroke. Furthermore, belonging to multiple groups has also been found to be protective against developing depression (Cruwys et al., 2013), and to be associated with reduced depression and distress among university students and healthy adults during times of important life transitions (Thoits, 1983; Linville, 1985, 1987; Iyer et al., 2009). Along these lines, a longitudinal study with British students entering university showed that having a greater number of group memberships before the move to university predicted better adjustment and well-being once students had been at university for a few months (Iyer et al., 2009). Moreover, there was a clear relationship between the numbers of groups that these students were members of and the amount of social support they reported receiving, such that MGM provided individuals with more sources of support that could be drawn on to provide stability in this period of transition. While the aforementioned studies involved correlational or longitudinal examinations, the beneficial effect of MGM has also been demonstrated experimentally by manipulating the psychological availability of group memberships (i.e., getting participants to think of one, three, or five group memberships; Jones and Jetten, 2011, Study 2).

However, the relationship between MGM and well-being has mostly been demonstrated in Western cultural contexts. This is important because research suggests that individuals in Asian cultural contexts may not utilize their MGM in the same way as individuals in Western cultural contexts. We therefore question the generalizability of the multiple group membership effect and ask in this research whether group memberships play a different role in Asian cultures.

# Multiple Group Memberships and Well-Being: A Cross-Cultural Perspective

Research has shown that there is cultural variation in how people construe the self and their relationships with others (e.g., relative emphasis on personal goals or group goals; Markus and Kitayama, 1991), and in turn, their expectations of those relationships. In Western contexts, individualism is emphasized, and individuals are encouraged to promote their uniqueness and act according to their own volition (Heine et al., 1999). By contrast, the emphasis is on collectivism in Asian contexts. Individuals in these contexts are encouraged to maintain closeness and harmony within social groups and view group goals as primary and personal beliefs, needs, and goals as secondary (Markus and Kitayama, 1991; Heine et al., 1999). These differences in cultural norms and expectations about relationships are likely to impact on the extent to which individuals seek social support (Taylor et al., 2004; Kim et al., 2006; Sherman et al., 2009). Specifically, in Asian cultures, individuals are more careful with regards to extracting support from others because they are more concerned about the potential negative relational implications of support seeking (Kim et al., 2008). That is, social support seeking in this cultural context is more likely to be seen as an imposition, burdening the provider of the support because drawing on another person for support can tax that other person's time and attention resources (Seidman et al., 2006). This can also be costly in the sense that it may undermine and disrupt group harmony. In contrast, individuals in Western cultures view social support as personal resources, making them more likely to actively solicit them in times of difficulty (Uchida et al., 2008; Chen et al., 2012).

Evidence supporting this argument comes from Taylor et al. (2004) who showed that Asians were less willing than European Americans to draw on social support from their social networks to cope with stress. Furthermore, they found Asians to be more sensitive to the relational consequences of support seeking than Americans, and that these relational concerns accounted for the cultural differences in the use of social support for coping with stress. This suggests that among Asians, cultural norms (e.g., concerns about burdening others) could discourage the active engagement of one's social support network to cope with stress.

Building on these previous findings, to the extent that MGM are seen as another form of relationship with others (Jetten et al., 2015), it is likely that cultural variations in shared assumptions about relationships and support seeking may influence the way individuals utilize their MGM. Specifically, this may affect the degree to which one feels it is appropriate to draw upon support resources derived from shared group memberships, which would in turn influence one's well-being. Consequently, MGM may confer fewer benefits to the psychological wellbeing of Asians because cultural norms on relationships and support seeking in Asian cultural contexts might lead to reluctance to tap into social support resources from their group memberships (see **Figure 1**). Importantly, this line of work extends on previous research examining cultural differences that has focused largely on relationships and support seeking from significant individuals (e.g., a family member, friend; Kim et al., 2008). Here, we consider the particular contribution that relationships and social support from groups of others makes to health and well-being—where group memberships can comprise a range of diverse relationships such as those with family and friendship groups as well as religious, community, and recreational groups. More importantly, we focus on those group memberships that are meaningful and which individuals feel psychologically connected to (i.e., social identification; Jetten et al., 2014).

# The Present Research

In four studies, we investigated the extent to which cultural normative expectations about support seeking would influence the degree to which individuals derive psychological resources and well-being benefits from belonging to multiple groups. We examined the effect of MGM on a wide range of well-being indicators—notably, life satisfaction, happiness, depression, anxiety, and stress. Study 1 used an exploratory correlational design to examine the role that MGM play in the well-being of Asians. In Studies 2 and 3, we investigated a potential cultural underpinning that might account for why individuals from different cultures (Asian vs. Western) may utilize their group membership resources in different ways. Here, we predicted cross-cultural variation in drawing on support from shared group memberships, such that reluctance to enlist support from relationships may lead to fewer support resources being derived from belonging to multiple groups. Lastly, to examine the robustness of the relationship between MGM and well-being in different cultural contexts, we conducted a review of the empirical literature assessing this cross-sectional relationship (contrasting Asians with Westerners) in Study 4.

# STUDY 1

In a first exploratory correlational study, we aimed to explore the relationship between MGM and psychological well-being among international Asian students who had just transitioned from their own culture to a Western culture to commence their studies.

Ethical clearance for all studies was provided by the Psychology Ethics committee at the University of Queensland. In all studies, participants ticked a box before starting the survey, indicating their informed consent.

to be weaker among Asians than among Westerners.

#### Chang et al. Culture, Multiple Group Memberships, and Well-Being

# Method

#### Participants

Participants were 180 international students at a large Australian university who had only been in Australia for 1 month (54 males, 124 females, 2 did not indicate their gender). Participants had a mean age of 22.0 years (SD = 4.00) and were born in Asia (i.e., China, Hong Kong, Indonesia, Malaysia, and Singapore).

#### Measures

Participants completed a survey that included demographic questions, and scales measuring multiple group membership and well-being (i.e., life satisfaction, depression, anxiety, stress).

#### **Multiple group membership**

Multiple group membership was assessed with four items from the Exeter Identity Transition Scale (EXITS; see Haslam et al., 2008). These included "I belong to a lot of groups," "I join in the activities of lots of different groups," "I have friends who are members of lots of groups," and "I have strong ties with lots of different groups." Each item was rated on a 7-point scale (1 = Strongly Disagree, 7 = Strongly Agree; α = 0.91).

## **Satisfaction with life scale**

Satisfaction with life scale (SWLS; Diener et al., 1985)**.** Five items (e.g., "I am satisfied with my life") assessed one's judgment of satisfaction with one's life (α = 0.87). Each item was rated on a 7-point scale (1 = Strongly disagree, 7 = Strongly agree).

### **Depression anxiety stress scales**

Depression anxiety stress scales (DASS-21; Lovibond and Lovibond, 1995). Depression, anxiety, and stress were measured using the DASS-21. The DASS-21 consists of three 7-item subscales that measure depression (e.g., I felt down-hearted and blue; α = 0.88), anxiety (e.g., I felt I was close to panic; α = 0.80), and stress (e.g., I found it hard to wind down; α = 0.85). It has excellent reliability and validity in both clinical and nonclinical samples (Henry and Crawford, 2005; Crawford et al., 2009). Participants were asked to indicate how much the items had applied to them during the past week on a scale ranging from 0 = Did not apply to me at all, to 3 = Applied to me very much, or most of the time. For each subscale, responses were summed and multiplied by two in accordance with recommended practice (Lovibond and Lovibond, 1995).

# Results and Discussion

Descriptive statistics and bivariate correlations of key variables are presented in **Table 1**. Multiple group membership was not significantly related to life satisfaction (r = 0.121, p = 0.105), depression (r = −0.104, p = 0.167), anxiety (r = −0.105, p = 0.162), or stress (r = −0.070, p = 0.350), suggesting that belonging to multiple groups may not be associated with enhanced well-being for Asian participants.

This finding contrasts with mounting evidence of the wellbeing benefits of MGM (Jetten et al., 2012, 2014), and provides initial support for our argument that Asians may derive fewer well-being benefits from multiple groups relative to Westerners. However, aside from the fact that one should be careful in drawing inferences from an absence of a significant relationship


*Study 1; N* = *180.*

\**p* < *0.05,* \*\**p* < *0.01.*

(i.e., confirming the null hypothesis), another limitation is that having only explored this relationship in Asians, questions about (a) the extent to which the multiple group membership effect differs across cultures, and (b) the mechanism underlying this cultural difference (if any), remain. We address these limitations in the following studies.

# STUDY 2

Study 2 directly compared Western and Asian samples to investigate cultural differences in the relative benefit to psychological well-being of belonging to multiple groups. Our prediction was that culture would moderate the positive effect of MGM on well-being. Specifically, belonging to multiple groups would benefit the well-being of Asians to a lesser extent than Westerners. A second goal of Study 2 was to examine a potential mechanism underlying this cultural difference—the notion that Asians are less likely to derive support resources from their MGM (Taylor et al., 2004; Kim et al., 2006; Sherman et al., 2009). We predicted that only for Western participants, social support would mediate the effect of MGM on psychological well-being. We expected that this mediational path would not be significant for Asian participants. In our examination of these predictions, we adopted happiness and depression as indicators of well-being.

# Method

#### Participants

Participants were 137 undergraduate students at a large Australian university; 60 international students who were born and raised in Asia and described their ethnicity as Chinese (17 males and 43 females), and 77 Australians Caucasians (28 males and 49 females). Participants either received course credit or 10 Australian dollars for their participation. Asian participants (n = 60) were on average 23.2 years old (SD = 4.01) and the average length of time spent in Australia was 23.73 months. Western participants (n = 77) had a mean age of 19.6 years (SD = 3.69).

#### Measures

Participants completed a survey that included demographic questions, and scales measuring multiple group membership, social support, and well-being.

#### **Multiple group membership**

An abbreviated two item scale (i.e., "I am a member of lots of different social groups" and "I have friends who are in lots of different social groups"), assessed the extent to which individuals belong to multiple groups (Jetten et al., 2010; r = 0.68). These items have been used in previous research and have been found to demonstrate good internal reliability (see Jetten et al., 2010). Each item was rated on a 7-point scale (1 = Strongly Disagree, 7 = Strongly Agree).

#### **Social support**

Participants' level of social support was measured using four items adapted from Van Dick and Haslam (2012; e.g., "Do you get the help you need from other people?"; α = 0.94), to which participants responded using a 7-point scale (1 = Not at All, 7 = Completely).

#### **Well-Being**

Two measures indexed this construct. A single item of overall well-being, indexing happiness, was used. Previous research has argued that single-item measures of happiness are not only valid but also, produce similar findings to multi-item scales of this construct (see Abdel-Khalek, 2006; Jetten et al., 2012). Participants were asked to respond to the item "Presently would you describe yourself as:" using a 5-point scale from 1 "Very unhappy," to 5, "Very happy."

The Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977) comprising 20 items (e.g., "I was bothered by things that usually don't bother me") assessed levels of depression (α = 0.91). Each item was rated on a 4-point scale (0 = Rarely or none of the time, to 3 = Most or all of the time).

## Results

Descriptive statistics for the sample by culture are provided in **Table 2**. While multiple group membership was not associated with happiness (r = −0.058, p = 0.657) in Asian participants, it was associated with greater levels of happiness (r = 0.343, p = 0.002) in Western participants. The difference between these correlations was statistically significant, Z = −2.36, p = 0.018.

Results also revealed that whereas multiple group membership was not associated with depression (r = 0.063, p = 0.632) in Asian participants, belonging to multiple groups was associated with lower levels of depression (r = −0.317, p = 0.005) in Western participants. The difference between these correlations was also statistically significant, Z = 2.22, p = 0.026.

TABLE 2 | Descriptive statistics as a function of culture.


*Study 2; N* = *137.*

#### Culture Moderates the Positive Effect of Multiple Group Membership on Well-Being

A moderation analysis [Hayes, 2013, model 1; significance levels were calculated using unstandardized values in Hayes PROCESS model 1, as recommended by Hayes and Preacher (2014)] was conducted to examine the influence of culture as a moderator of the relationship between multiple group membership and happiness. Multiple group membership was included as a continuous predictor, with happiness as the outcome variable. Culture (Asian vs. Western) was entered as the moderator. Results showed that culture, B = 0.29, 95% CI [0.131, 0.456], t = 3.57, p < 0.001, but not multiple group membership, B = 0.11, 95% CI [−0.019, 0.233], t = 1.67, p = 0.097, significantly predicted happiness. The interaction between culture and multiple group membership was significant, B = 0.13, 95% CI [0.006, 0.254], t = 2.07, p = 0.041. Simple slopes analysis revealed that multiple group membership was associated with higher levels of happiness for Western participants, B = 0.22, 95% CI [0.041, 0.400], t = 2.42, p = 0.017, but not for Asian participants, B = −0.04, 95% CI [−0.211, 0.132], t = −0.45, p = 0.650 (see **Figure 2**).

A second moderation analysis was conducted to examine the influence of culture as a moderator of the relationship between multiple group membership and depression. Analysis revealed no significant effect of culture, B = −0.04, 95% CI [−0.128, 0.042], t = −1.01, p = 0.314, or multiple group membership, B = −0.06, 95% CI [−0.122, 0.012], t = −1.62, p = 0.107, on depression. However, the interaction between culture and multiple group membership was significant, B = −0.07, 95% CI [−0.135, −0.0008], t = −2.00, p = 0.047. Simple slopes analysis revealed that multiple group membership was associated with lower levels of depression for Western participants, B = −0.11, 95% CI [−0.208, −0.021], t = −2.42, p = 0.017, but not for Asian participants, B = 0.02, 95% CI [−0.075, 0.117], t = 0.43, p = 0.665 (see **Figure 3**).

#### Social Support Mediates the Positive Effect of Multiple Group Membership on Well-Being for Western, but Not Asian, Participants

To test whether the positive effect of multiple group membership on happiness could be explained by a difference between cultural groups in the mediating role of social support, a moderated mediation analysis with 10,000 bootstrap samples was conducted [Hayes, 2013, model 8; significance levels were calculated using unstandardized values in Hayes PROCESS model 8, as recommended by Hayes and Preacher (2014)]. Multiple group membership was included as a continuous predictor, with happiness as the outcome variable. Social support was entered as a continuous mediator, and culture (Asian vs. Western) was entered as the moderator. Analysis revealed no significant effect of multiple group membership, B = 0.14, 95% CI [−0.031, 0.307], t = 1.61, p = 0.109, or culture, B = 0.18, 95% CI [−0.013, 0.374], t = 1.85, p = 0.067, on social support. However, the interaction between multiple group membership and culture on social support was significant, B = 0.17, 95% CI [0.002, 0.335], t = 2.00, p = 0.048. Conditional indirect effects (IE) revealed a significant indirect effect of multiple group membership on

happiness via social support for Western participants, IE = 0.10, standard error [SE] = 0.05, 95% CI [0.026, 0.205], but not for Asian participants, IE = −0.02, standard error [SE] = 0.04, 95% CI [−0.107, 0.059]. Results showed that the indirect effect through social support was significantly different between Asian and Western participants, B = 0.12, 95% CI [0.017, 0.267]. Culture did not moderate the effect of multiple group

membership on happiness when the mediator was in the model, B = 0.07, 95% CI [−0.032, 0.170], t = 1.35, p = 0.181 (See **Figure 4**). This is consistent with our prediction that it is the capacity of MGM to provide support that is moderated by culture, and suggests that this mechanism fully explains the cultural differences in the relationship between MGM and wellbeing.

Frontiers in Psychology | www.frontiersin.org July 2016 | Volume 7 | Article 1001 |

To examine whether social support would mediate the effect of multiple group membership on depression for Western, but not Asian, participants, a second moderated mediation analysis with 10,000 bootstrap samples was conducted (Hayes, 2013, model 8). The interaction between multiple group membership and culture on social support was significant, B = 0.17, 95% CI [0.002, 0.335], t = 2.00, p = 0.048. Conditional indirect effects (IE) revealed a significant indirect effect of multiple group membership on depression via social support for Western participants, IE = −0.05, standard error [SE] = 0.02, 95% CI [−0.104, 0.013], but not for Asian participants, IE = 0.01, standard error [SE] = 0.02, 95% CI [−0.029, 0.055]. Results showed that the indirect effect through social support was significantly different between Asian and Western participants, B = −0.06, 95% CI [−0.136, −0.008]. Culture did not moderate the effect of multiple group membership on depression when the mediator was in the model, B = −0.04, 95% CI [−0.094, 0.020], t = −1.28, p = 0.202 (See **Figure 5**).

We also tested the most plausible alternative mediation model, which was that individuals with better well-being would be more likely to engage in drawing social support from different social groups, and thus, more likely to feel a part of multiple groups (Hayes, 2013, model 4). Happiness was included as a continuous predictor, with multiple group membership as the outcome variable. Social support was entered as a continuous mediator. However, this model was not consistent with the data. In the Western subsample, both the indirect effects (IE) of happiness, IE = 0.19, standard error [SE] = 0.16, 95% CI [−0.070, 0.562], and depression, IE = −0.34, standard error [SE] = 0.22, 95% CI [−0.855, 0.035], on multiple group membership were not significant. Similarly, in the Asian subsample, both the indirect effects of happiness and depression on multiple group membership were not significant.

# Discussion

In Study 2, we found that belonging to multiple groups was associated with greater levels of psychological well-being for Western participants, but not for Asian participants, and this was true across two indexes of well-being: happiness and depression. More specifically, we found that only for Western participants, the impact of belonging to multiple groups on well-being was fully mediated by social support.

It thus appears that there are boundary conditions for the effect of MGM on well-being. Even though MGM serve as a psychological resource from which individuals can receive and benefit from the support provided by fellow group members to enhance well-being (Haslam, 2004; Jetten et al., 2012, 2014), for Asian participants, we observed that they were not deriving as much support from their group memberships, reducing the positive effect of multiple group membership on well-being. We propose that this lack of social support from MGM arises from Asian cultural norms about relationships and support seeking in which there is a reluctance to seek support because one should not burden their social networks (Taylor et al., 2004; Kim et al., 2006, 2008). We test this prediction in Study 3 by including measures indexing concerns about burdening others.

A potential limitation of Studies 1 and 2 is that the studies' samples comprised Asian international students who find themselves in a unique situation. They might belong to many different groups in their home countries (e.g., family, friendship, community, recreational groups), but experience difficulty in utilizing them for support resources and deriving associated wellbeing benefits because they are not easily accessible (i.e., located overseas in their home countries). Also, the group memberships to which Asian international students belong in Australia may be quite different from those of Australians. That is, the new group memberships may be weaker, less stable or less meaningful given they have only been part of them for a shorter duration; possibly limiting the extent to which support resources can be derived from these memberships. This may explain why Asians may derive fewer well-being benefits from multiple groups relative to Westerners in our studies. We address this limitation in Study 3.

# STUDY 3

In Study 3, we measured perceived burden as a proxy to assess one's reluctance to draw upon relationships for support resources. In line with the above reasoning, we tested whether the effect of MGM on well-being was dependent on the extent

to which one is reluctant to seek social support from their relationships. Specifically, we expected that MGM would only be associated with enhanced psychological well-being for those Asian participants who were least reluctant to draw on support from others. Importantly, Study 3 also addressed a limitation of Studies 1 and 2 by investigating this prediction in a sample of Asian students at a university in Singapore residing within their cultural context.

# Method

#### Participants

Participants were 105 students at a Singapore university (43 males and 62 females). Participants received 6 Singapore dollars for their participation. On average, participants were 21.4 years old (SD = 1.54) and all described their ethnicity as Chinese.

#### Measures

Participants completed a survey that included demographic questions, the multiple group membership scale (see Study 2; Jetten et al., 2010; r = 0.59.), and the Satisfaction with Life Scale (see Study 1; Diener et al., 1985; α = 0.85).

## **Reluctance to enlist social support**

Three items were developed to measure the extent to which one is reluctant to enlist social support from relationships due to concerns about burdening others (i.e., "It is important for me not to burden others with my problems"; "One should avoid troubling others," and "I make an effort not to impose on others"; α = 0.80). Each item was rated on a 7-point scale (1 = Strongly disagree, 7 = Strongly agree).

# Results and Discussion

**Table 3** provides descriptive statistics and results of bivariate correlations between key variables. Multiple group membership (M = 5.14, SD = 1.07) was not significantly related to life satisfaction (M = 4.57, SD = 1.12), r = 0.130, p = 0.186.

A moderation analysis (Hayes, 2013, model 1; significance levels were calculated using unstandardized values) was used to examine whether the degree of reluctance to enlist social support would moderate the relationship between multiple group membership and life satisfaction. Multiple group membership

#### TABLE 3 | Descriptive statistics and correlations.


*Study 3; N* = *105.*

\**p* < *0.05,* \*\**p* < *0.01.*

was included as a continuous predictor, and life satisfaction as the outcome variable. Reluctance to enlist social support was entered as a continuous moderator. There was no significant effect of multiple group membership, B = 0.13, 95% CI [−0.072, 0.332], t = 1.28, p = 0.205, or reluctance to enlist support, B = 0.02, 95% CI [−0.208, 0.252], t = 0.19, p = 0.850, on life satisfaction. Consistent with our prediction, the interaction term between belonging to multiple groups and reluctance to enlist support was significant, B = −0.18, 95% CI [−0.352, −0.010], t = −2.10, p = 0.038. Simple slopes analysis revealed that belonging to multiple groups was associated with greater levels of life satisfaction for those who were least reluctant to enlist social support, B = 0.32, 95% CI [0.034, 0.597], t = 2.22, p = 0.029, but was unrelated to life satisfaction for participants who were more reluctant to enlist social support, B = −0.06, 95% CI [−0.308, 0.197], t = −0.44, p = 0.663 (see **Figure 6**).

Here, we replicate the findings from previous studies that MGM are not associated with psychological well-being for Asian participants. However, additionally, we found that one's reluctance to enlist social support moderated the relationship between multiple group membership and psychological wellbeing, highlighting that belonging to multiple groups was in fact, associated with greater well-being among those Asian participants who were less reluctant to enlist social support. As the findings from our moderation analysis suggest, belonging to multiple groups may benefit the well-being of Asians to a lesser extent, especially when they are more sensitive to concerns about burdening others. This potentially increases their reluctance to draw on support resources from their relationships, reducing the positive effect of MGM on well-being.

Taken together, the findings across the three studies show that for Asians, multiple group membership is not strongly associated with psychological well-being, with correlations ranging between 0.058 and 0.130 across three studies and five indicators of wellbeing. This suggests that for Asians, the relationship between multiple group membership and well-being is weaker than it is for Westerners. To further interrogate this point, we conducted an empirical review of the work assessing the relationship between MGM and well-being.

# STUDY 4

Study 4 aimed to (1) investigate the robustness of the effect of MGM on well-being in the literature (both the Asian and Western cultural contexts), and (2) determine whether this effect is smaller in Asian, compared to Western, cultural contexts. Addressing these aims, we obtained effect sizes by identifying and integrating relevant studies in the literature examining the relationship between MGM and psychological well-being. PsycINFO and PubMed were searched for relevant studies using the following keywords: multiple group membership, social group memberships, and greater number of group identification. Eligible studies were published between 1990 and August 2015, in English and in a peer-reviewed journal, and included quantitative measures or manipulations of number of group memberships across different life domains (e.g., leisure or social group, community group, family) along with a dependent quantitative measure of psychological well-being (i.e., life satisfaction, happiness, self-esteem, psychological distress, mood, depression, anxiety, and stress).

The search produced a total of 735 references in PsycINFO and 1030 references in PubMed, of which 1564 were unique references. Of these, we identified 15 relevant articles reporting 23 different studies, with a total of 14,063 participants. None of these studies compared Westerners with Asians, and only four were conducted with Asian participants. More detail on each of the studies is provided in **Table 4**. In addition to these studies, we included our data sets from Study 1 to Study 3 in the integrative analysis. This resulted in 27 independent studies for the analysis.

# Results and Discussion

The meta-analytical procedure reported in Borenstein et al. (2009) was used to determine the overall relationship between MGM and well-being. From the studies identified, 27 effect sizes were calculated with a total sample size of 14,063 participants and correlations ranging from −0.06 to 0.48. A random effects model (which assumes that the true effect size varies across studies and follows a normal distribution around the mean) revealed an overall mean effect size of r = 0.22, 95% CI [0.175, 0.267], Z = 9.13, p < 0.001.

To compare the effect size of studies consisting primarily of Westerners to studies involving Asians, subgroup analyses were conducted. For the studies with Asian participants (n = 7), 7 effect sizes were calculated with a total of 1153 participants and correlations varying between −0.06 and 0.26. The random effects model revealed a mean effect size of r = 0.13, 95% CI [0.072, 0.187], Z = 4.38, p < 0.001, indicating that the effect size was small (Cohen, 1988). For the studies with Western participants (n = 20), 20 effect sizes were calculated with a total sample size of 12,910 participants and correlations ranging between 0.08 and 0.48. A random effects model showed a mean effect size of r = 0.25, 95% CI [0.194, 0.302], Z = 8.67, p < 0.001, for these studies. More specifically, in 95% of all possible meta-analyses, the true mean effect would probably fall in the range of 0.194– 0.302 for Westerners, indicating a robust effect. In contrast, for Asians, the true mean effect would usually fall in the range of 0.072–0.187, suggesting there is a higher probability that the true effect is weaker for Asians, and likely to be null in some analyses (Cohen, 1988). As it is predicted that multiple group membership processes would be different across cultures, in calculating the Q-value, the between-studies variance, T2, was computed within subgroups and used as a separate estimate for


TABLE 4 | The relationship

 between multiple group

memberships

 and

psychological

 well-being,

 as reported in 27 studies.


*(Continued)*


each subgroup. The Q-value for the difference was 8.75 with 1 df and p = 0.003.

The present analysis is the first review applying metaanalytic techniques to interrogate the relationship between MGM and indices of well-being. We found some evidence that multiple group membership was more strongly related to well-being in the studies involving Westerners than in studies involving Asians. This strengthens our argument that individuals in cultural contexts that privilege drawing support from their relationships may reap greater well-being benefits from their MGM.

Nevertheless, some caution is warranted in interpreting the results of this review. The purpose of this review was not to provide an all-encompassing meta-analysis, but rather an initial interrogation of the existing literature applying metaanalytic principles. A potential limitation is the small number of studies included in the Asian subgroup analysis. Of note, Studies 1–3 contributed 3 out of the 7 studies in this analysis, and whilst these all had confidence intervals including 0, the other 4 studies did not. The reason for this is not clear, but publication bias against null effects may play a role (Franco et al., 2014). If so, the mean effect size for Asians in reality may be even weaker, relative to Westerners. Despite this, our findings suggest that compared to Westerners, Asians may generally derive fewer well-being benefits from belonging to multiple groups.

# GENERAL DISCUSSION

In this current research, we apply a novel theoretical framework to understand some important cross-cultural differences in the relationship between MGM and psychological well-being. Overall, we found support for our argument that, belonging to multiple groups may confer little, or fewer, well-being benefits for Asians relative to Westerners. Importantly, our findings provide further evidence for a growing body of research claiming that MGM can form the basis of a "social cure"; the sense of shared social identities derived from belonging to multiple groups are an important basis for social support, which is a critical social factor in protecting well-being (Haslam et al., 2009; Jetten et al., 2012). Of note though, we provide some of the first evidence for the boundary conditions of this effect, showing that this relationship is more likely to be observed for individuals from Western than Asian cultural contexts.

For Asians, we found that multiple group membership was not significantly, or only weakly, associated with well-being. This was replicated across 3 survey studies in samples comprising Asian students from Australia and Singapore using multiple indices of well-being (Studies 1–3), and was also reflected in the findings of our review applying meta-analytic techniques (Study 4). In particular, in our cross-cultural study (Study 2) and review (Study 4), we showed that belonging to multiple groups was comparatively less beneficial to the psychological well-being of Asians, relative to Westerners. This leads us to conclude that cultural differences in the extent to which individuals derive well-being benefits from belonging to multiple groups can exist. Specifically, whilst belonging to multiple groups is likely to confer well-being benefits for individuals from Western cultures, the "more the merrier" effect may hold less truth for individuals from Asian cultures. As these findings suggest, it is important to consider an individual's cultural context to obtain a full understanding of their psychology (Kim and Markus, 1999).

Critically though, in examining the cultural underpinnings of multiple group membership processes, the present analysis sheds new light on the contribution that cultural background makes to understanding how individuals derive psychological resources from their group memberships in different ways. Utilization of social group resources (e.g., support) is in fact interconnected with cultural norms about relationships. Specifically, it was found in Study 2, that social support mediated the effect of multiple group membership on well-being for Westerners but not Asians. Also, Study 3 showed that among Asians, multiple group membership was only associated with well-being for those who were least reluctant to enlist social support due to concerns about burdening others. Our current work therefore provides the first empirical demonstration that, compared to Westerners, Asians may reap fewer well-being benefits from belonging to multiple groups. This is possibly because of an emphasis on different cultural norms on relationships and support seeking across cultures (i.e., the extent to which it is perceived as appropriate to draw support resources from others), thereby leading support seeking to be perceived differently in different cultures (Kim et al., 2008). Among Asians, the cultural notion that it is important not to be a source of burden on others may cause them to worry about imposing on their relationships by requesting support (Kim et al., 2008) which, in turn, strengthens their reluctance to draw on such resources from their MGM resulting in reduced well-being benefits being derived from these memberships. Conversely, the present analysis suggests that individuals from Western cultures appear to have fewer concerns about deriving support from their social group memberships, which in turn is associated with enhanced well-being. In fact, these findings are in line with the literature showing cross-cultural differences in group processes, where it is argued that self-concepts and group behaviors are construed in more relational ways in Asian cultural contexts than in Western cultural contexts (Yuki and Takemura, 2013). In groups, Asians may be more focused on maintaining harmonious and reciprocal relationships, and therefore, they may be more reluctant to seek support from their group members in order to maintain these delicate interpersonal ties. By contrast, maintaining interpersonal ties and social harmony in groups may be less of a concern in Western cultural contexts. In this way, Westerners may feel more able to obtain support from their MGM.

Despite these cultural variations in notions about support seeking, it is noteworthy that Study 2 found social support to be associated with greater well-being for both Asians and Westerners. This finding is in line with an extensive body of research documenting the mental health benefits of social support (House et al., 1988; Thoits, 1995). What is interesting from the findings though, is that while Asian cultural norms about relationships and support seeking can make Asians more reluctant to enlist social support (Taylor et al., 2004; Kim et al., 2006, 2008), this does not appear to prevent them from perceiving such support as beneficial. Specifically, for Asians, the findings suggest that the costs involved in utilizing social support (e.g., highlighting one's incompetence; over-taxing the resources of others; Bolger et al., 2000; Seidman et al., 2006) were more likely to occur in the process of deriving support from others, rather than after such support was provided. This is reflected in the nonsignificant relationship between multiple group membership and social support, but significant positive relationship between social support and well-being. In contrast, we did not find evidence of this "cost" in our sample of Westerners. Given that previous research has highlighted that Asians can benefit from seeking and receiving support from relationships which are perceived as mutual and interdependent (i.e., the individual had previously provided support to the relationship partner; Wang and Lau, 2015), future research could examine these processes in greater detail. For example, Asians may possibly derive support resources from their mutual relationships when belonging to multiple groups, thereby reaping well-being benefits from MGM.

We note that the current studies were not without limitations. First, the cross-sectional nature of the data does not permit us to make causal inferences, and thus, directionality of effects cannot be inferred. However, it is worth noting that the alternative model that we tested in Study 2 did not provide a superior fit for the data. Also, because our study samples comprised only university students (Studies 1–3), the findings may not generalize to a nonuniversity population or to clinical populations who may be more socially isolated. All these limitations should be addressed in future work by examining the generalizability of findings to other populations and by examining evidence for the predicted relationship longitudinally.

Another limitation is that our hypotheses in Studies 1 and 2 (i.e., weaker relationship between MGM and well-being for Asians, relative to Westerners) were examined using samples of Asian international students. Directly comparing Asian international students to Australian Caucasian students may have created a potential confound in Study 2—that these Asian participants' unique situation may have influenced the ease and extent to which they derived support and gained well-being benefits from their group memberships. The present data do not allow us to rule out this alternative explanation. However, our findings from Study 3 speak to this potential confound. In this study, we replicated the findings with a sample of Asian students in Singapore who were within the context of their home culture, and showed that multiple group membership was not associated with well-being. This finding, obtained in an un-confounded study context, substantiates our argument that Asians may generally derive fewer well-being benefits from belonging to multiple groups than Westerners. Nevertheless, it is recommended that future research explore and replicate these effects in other samples, for instance, (a) Asians who are within their cultural context or (b) Asian international students could be compared to Western international students (e.g., Europeans) in the Australian context.

Lastly, for practical reasons, social support was measured using four items adapted from Van Dick and Haslam (2012) instead of the much longer COPE Inventory (Carver et al., 1989), a commonly used measure to assess cultural differences in social support seeking (Kim et al., 2006). This social support measure has in fact been shown to demonstrate high reliability in Western samples (Van Dick and Haslam, 2012), though it has not previously been used in Asian samples. Apart from this, the other measures used in our studies have previously been used in both Asian and Western samples with adequate psychometric properties. In general, nonetheless, it is recommended that future work should examine the hypotheses using other scales to extend as well as replicate our findings.

Overall, our findings are consistent with the literature demonstrating cross-cultural differences in group processes (e.g., variations across cultures in how individuals relate to groups; Yuki and Takemura, 2013). However, there are several possible interpretations of our findings. For instance, research has suggested that MGM lower one's well-being because groups can demand too much effort and deplete one's limited time and energy (Bolger et al., 1989). In line with the strong motivation to maintain harmonious and reciprocal relationships, and relatedly, the strong emphasis on the fulfillment of role-based obligations in relationships in Asian cultural contexts (Rothbaum et al., 2000), it is therefore possible that Asians derive fewer wellbeing benefits from belonging to multiple groups because of their need to fulfill obligations (e.g., to reciprocate support). This explanation should be examined in future work, ideally with populations other than undergraduate students. In addition, future research could examine these cross-cultural differences in group processes in greater detail to understand how people conceptualize their MGM. This may help to shed light on cultural differences in what constitutes MGM and what it means to belong to multiple groups.

Nevertheless, the current work has generated important insights about the function of multiple group membership processes in different cultural contexts. More specifically, it has provided the first empirical evidence on the circumstances in which these group memberships can sometimes fail to have any positive effects on well-being. This has important practical implications for psychosocial interventions involving attempts to enhance well-being by (1) advocating the building and development of social group memberships or (2) facilitating individuals to draw support resources from their group memberships (Jetten et al., 2014; Newlin et al., 2015). Our findings suggest that it may be important to tailor these interventions to the needs of Asians. For instance, when working with Asian clients, therapists may need to work with clients' reluctance to draw upon support resources from their relationships and help them tap these resources in a way that is consistent with Asian culture normative standards.

# Conclusion

Although the relationship between MGM and psychological well-being appears to be relatively well-established, our findings indicate that there is cross-cultural variation in this association. The well-being benefits from MGM are not only determined by how much support can be derived from belonging to multiple groups, but also how easy it is for an individual to draw upon these resources. Given that Asian cultural norms prescribe individuals not to draw on such support lightly, unfortunately, Asians may not fully benefit from the psychological resources their group memberships encompass.

# AUTHOR CONTRIBUTIONS

MC, JJ, TC, and CH contributed to the study design. MC and NP collected the data. MC performed the data analysis and interpretation under the supervision of JJ, TC, and CH. MC

# REFERENCES

(References marked with an asterisk indicate articles included in Study 4).


drafted the manuscript, and JJ, TC, and CH provided critical revisions.

# ACKNOWLEDGMENTS

The work undertaken by the second and fourth authors on this paper was supported by an Australian Research Council Future Fellowship (FT110100238) and by the Canadian Institute for Advanced Research Social Interactions, Identity and Well-Being Program, respectively.


and academic adjustment of urban youth. Appl. Dev. Sci. 13, 199–216. doi: 10.1080/10888690903288755


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Chang, Jetten, Cruwys, Haslam and Praharso. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Multiple Group Membership and Well-Being: Is There Always Strength in Numbers?

#### Anders L. Sønderlund1,2 \*, Thomas A. Morton<sup>1</sup> and Michelle K. Ryan1,3

<sup>1</sup> Social, Environmental and Organisational Research Group, Department of Psychology, University of Exeter, Exeter, United Kingdom, <sup>2</sup> Department of Public Health, University of Southern Denmark, Odense, Denmark, <sup>3</sup> Department of Human Resource Management and Organizational Behavior, University of Groningen, Groningen, Netherlands

A growing body of research points to the value of multiple group memberships for individual well-being. However, much of this work considers group memberships very broadly and in terms of number alone. We conducted two correlational studies exploring how the relationship between multiple group membership and well-being is shaped by (a) the complexity of those groups within the overall self-concept (i.e., social identity complexity: SIC), and (b) the perceived value and visibility of individual group memberships to others (i.e., stigma). Study 1 (N = 112) found a positive relationship between multiple group membership and well-being, but only for individuals high in SIC. This effect was mediated by perceived identity expression and access to social support. Study 2 (N = 104) also found that multiple group memberships indirectly contributed to well-being via perceived identity expression and social support, as well as identity compatibility and perceived social inclusion. But, in this study the relationship between multiple group memberships and well-being outcomes was moderated by the perceived value and visibility of group memberships to others. Specifically, possessing multiple, devalued and visible group memberships compromised well-being relative to multiple valued group memberships, or devalued group memberships that were invisible. Together, these studies suggest that the benefits of multiple group membership depend on factors beyond their number. Specifically, the features of group memberships, individually and in combination, and the way in which these guide self-expression and social action, determine whether these are a benefit or burden for individual well-being.

#### Keywords: social identity complexity, stigma visibility, identity compatibility, multiple identities, well-being

# INTRODUCTION

Social and cultural diversity in most Western countries is expanding as a result of rising social mobility, immigration, and international trade and investment (UN, 2009). In response to this increasing diversity, the social contexts in which people operate are becoming more complex, and as a consequence individual self-concepts are also changing. Whereas individual self-definitions were historically embedded within stable local contexts (Karner, 2011), in modern globalized society – where individuals cross geographical, cultural, and social borders more than ever – people express themselves in terms of a broader and more shifting array of group memberships and social categories (Crisp and Hewstone, 2006). These include categories based on demographic groupings

#### Edited by:

Soledad de Lemus, University of Granada, Spain

#### Reviewed by:

Toon Kuppens, University of Groningen, Netherlands Alexandra Vázquez, Universidad Nacional de Educación a Distancia (UNED), Spain

> \*Correspondence: Anders L. Sønderlund sonderlund@gmail.com

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

> Received: 28 October 2016 Accepted: 06 June 2017 Published: 21 June 2017

#### Citation:

Sønderlund AL, Morton TA and Ryan MK (2017) Multiple Group Membership and Well-Being: Is There Always Strength in Numbers? Front. Psychol. 8:1038. doi: 10.3389/fpsyg.2017.01038

like gender, ethnicity, nationality, or profession as well as on specific opinions, preferences, and shared activities (Tajfel et al., 1979; Crisp and Hewstone, 2006). Importantly, these categories now intersect in novel and interesting ways. For example, the movement of people across traditional boundaries, such as women and ethnic minorities into professional domains from which they were previously excluded, has challenged traditional stereotypes and created new hybridized identities (Crisp et al., 2001). All of these changes have the potential to impact on how individuals conceptualize the self, which in turn is likely to affect individual well-being (Cross et al., 2003; Crisp and Hewstone, 2006). The focus of this paper is on the specific nature of this relationship.

A growing body of research suggests that multiplicity of the self is generally a good thing. For example, membership in multiple groups has been associated with not only improved emotional well-being (Binning et al., 2009; Jetten et al., 2010), but also mental and physical resilience (Jones and Jetten, 2011), quality of life and coping (Haslam et al., 2008), and stress and social adaptation (Iyer et al., 2009). These positive effects are commonly attributed to the idea that identifying with multiple social groups grounds people more firmly in their social world, and provides them with multiple connections to similar others (Haslam et al., 2008; Jetten et al., 2012). The meaning and social support that follow from these connections, in turn, provide resources from which individuals can draw personal strength, resilience, and guidance in terms of values, attitudes, and behavior (e.g., Jetten et al., 2015; Chang et al., 2016; Steffens et al., 2016). Because of these properties, the greater the number of group memberships one has access to, the better one is likely to function.

Despite demonstrations of the psychological value of multiple group memberships, the exact mechanisms through which these contribute to well-being remain unclear. Indeed, research in this area begs the question of whether the reported benefits accumulate as a direct, linear function of the number of groups to which a person belongs, or whether additional factors are also necessary for membership in multiple groups to lead to positive outcomes. Intuitively, the features of specific group memberships, and the varying meaning of these to the individual in his or her social context, should be at least as important for well-being as the sheer number of groups to which they belong. For example, other theoretical models propose that the uniqueness of the specific identities that make up an individual's self-concept should moderate any relationship between multiple group membership and individual outcomes, including wellbeing (Roccas and Brewer, 2002). Within this framework, only membership in multiple, distinctive groups should increase the complexity of the self and through this increase the psychological resources available to the individual (see also Linville, 1987).

The psychological consequences of group membership are also likely to be contingent on the value attached to these – both by the individual and others around them. Illustrative of this point, belonging to a low-status or stigmatized group has been found to compromise a person's well-being (Puhl and Brownell, 2006). While it might be that membership in additional groups could weaken the impact of stigma on the individual, this is likely to be true only if those alternative bases for self-definition are valued rather than equally stigmatized (e.g., Rydell and Boucher, 2009). Similarly, membership in groups that are perceived to be in conflict or otherwise incompatible with one another (e.g., identifying as a male college football player, say, and ballet dancer) may also negatively affect well-being (Brook et al., 2008; Miramontez et al., 2008) – that is, simultaneously striving to live up to the standards of groups that have opposing values is more likely to be a strain than a source of individual strength. While it may be the case that membership in additional groups could buffer against the impact of stigma on the individual, this is only likely to be true if those alternative bases for self-definition are valued rather than equally stigmatized (e.g., Rydell and Boucher, 2009).

In the ways described above, various lines of previous research suggest that any relationship between multiple group memberships and well-being is unlikely to be simple, positive, and independent of other factors. The core concern of this paper is to provide further empirical investigation of this point. Specifically, we argue – and show – that the features of identities within the individual self-concept (e.g., whether individual identities are distinctive or not; and whether multiple identities are compatible or incompatible), and within the individual's social context (e.g., whether identities are socially valued or devalued; easily expressed or complicated) are important for shaping their psychological consequences when combined. Before presenting the two studies that test these ideas, we elaborate on the literature that we have already begun to sketch out above, as well as outline in more detail our own perspective.

# Social Identity Complexity

As we have previously alluded to, the literature on (multiple) group memberships already contains theoretical frameworks that highlight the need to attend to the interrelationships among group memberships, rather than their sheer number alone. In particular, Roccas and Brewer (2002), have elaborated on the processes through which group membership might contribute to social identity complexity (SIC). From this perspective, SIC refers to the subjective perception of overlap between different self-defining groups or categories – that is, the degree to which different social groups or categories (and the identities associated with these) share members (Roccas and Brewer, 2002; Brewer, 2008; Schmid and Hewstone, 2011). This perceived overlap can depart significantly from more objective measures of category overlap (Schmid and Hewstone, 2011). As **Figure 1** illustrates, perceiving one's social groups as highly overlapping may lead to the formation of a decidedly exclusive overall ingroup (for example, all 'real' Barcelonans are FC Barcelona fans), indicative of low SIC (Brewer and Pierce, 2005; Schmid and Hewstone, 2011). Alternately, others may perceive very little overlap, and recognize that ingroup members on one dimension are not necessarily ingroup members on another dimension (some Barcelonans may be Real Madrid fans or Manchester United fans or neither or both). This kind of representation would signify high SIC (Roccas and Brewer, 2002).

The SIC model was originally directed toward addressing questions about intergroup relations, rather than individual wellbeing. That is, high SIC was argued to be a source of positive intergroup perceptions and behavior because high SIC entails awareness of the cross-cutting relationships between groups. By contrast, the social cognition of those with low SIC is likely to be dominated by more simplified assumptions about group coherence and boundaries. Indeed, evidence suggests that high SIC is negatively associated with perceptions of intergroup threat, and prejudicial attitudes (e.g., Brewer and Pierce, 2005; Schmid et al., 2009). Although high SIC has also been suggested to be adaptive and beneficial in terms of individual well-being (Roccas and Brewer, 2002), to our knowledge, only limited research has been directed toward testing this possibility.

Nonetheless, SIC provides a useful theoretical starting point from which to explore the relationships between multiple group membership and well-being. As mentioned above, belonging to multiple distinct groups (high SIC) has been argued to be a source of identity strength, resilience, and well-being (Douglas, 2012; Jones et al., 2012). This may be because the more multifaceted and unique a person's self-concept is, the richer and firmer their sense of self becomes. For example, an individual whose self-concept comprises many distinct, non-overlapping groups, likely has a broader and more varied foundation from which to access belonging and social support, and through this derive well-being (Jetten et al., 2015). Conversely, belonging to multiple overlapping groups, one does not have access to the spread of multiple distinct sources of identity, support, and connection, that might additively combine to improve overall well-being (Zimet et al., 1988; Jetten et al., 2012, 2015). Instead, this individual would have a relatively narrow and singular basis from which to conceptualize the self and on which to rely for group-based support and solidarity. In these ways, the SIC perspective elaborates on the factors that might shape how multiple group memberships relate to individual well-being. However, the exact question of how SIC translates into individual psychological outcomes (beyond intergroup orientations) remains relatively underdeveloped within this framework, and direct tests of this prediction are, at least to our knowledge, currently absent from the literature.

# Social Identity Compatibility and Stigma

Although the SIC framework provides a solid foundation for investigating the well-being consequences of multiple group membership, the complexity approach nonetheless maintains a focus on the number of individual group memberships (albeit in conjunction with their distinctiveness) as a primary determinant of individual psychological outcomes. The broader meaning of these group memberships – both to the individual and within their social context – is not explicitly theorized in this framework. Nonetheless, it seems plausible that factors associated with the meaning of identities – both in the eyes of the individual and of others – should play a role in determining links between multiple group memberships and well-being. For example, belonging to multiple, distinctive but devalued groups should be a different experience than belonging to multiple, distinctive, and valued groups – although both these configurations are high in identity complexity. To unpack this idea, we now turn to the literature on stigma and well-being.

Stigmatization has been found to have significant adverse impacts on psychological well-being in terms of depression (Lee et al., 2002; Lewis et al., 2003), reduced life satisfaction (Markowitz, 1998), anxiety and hopelessness (Lee et al., 2002), and psychological distress (Kang et al., 2006). While there are some studies demonstrating that the negative impact of stigma can sometimes be ameliorated – for example, when individuals can attribute their stigma to the prejudice of others rather than their own perceived deficits (Major and O'Brien, 2005) – the majority of research on the topic has demonstrated an adverse impact of stigma on psychological and physical health outcomes (Carr and Friedman, 2005; Ellemers and Barreto, 2006; Beals et al., 2009; Puhl and Heuer, 2009; Schmitt et al., 2014).

Notwithstanding the overall negative impact of stigmatized group memberships on well-being, some studies have suggested that multiplicity of the self could mitigate these negative effects (Mussweiler et al., 2000). To explain this pattern, it has been

theorized that when any specific aspect of the self comes under threat or is devalued, individuals who have multiple selves on which to draw can more readily shift away from the threatened identity to an alternative one (Mussweiler et al., 2000; Benet-Martínez and Haritatos, 2005; Rydell et al., 2009; Sacharin et al., 2009; Jetten et al., 2012). Thus, multiple identities may buffer the individual against the effects of stigma attached to any single identity. On the other hand, an individual who belongs to a stigmatized group, but who has few alternative group memberships to draw on to define their self, is unlikely to be able to so easily disengage from the implications of their stigmatized group membership (Linville, 1987; Mussweiler et al., 2000; Roccas and Brewer, 2002).

Contrary to the idea of a buffer effect protecting the selfconcept, however, research also suggests that belonging to a stigmatized group can preclude the individual from accessing the benefits associated with alternative group memberships. In particular, a stigmatized group membership might eclipse all other group memberships, and in this way dominate, rather than be cushioned by them. This is because stigmatized identities can become highly central to the self and to one's interactions with others (Goffman, 1963). Specifically, due to their heightened salience, stigmatized identities can complicate the process of connecting to other groups and of maintaining meaningful relationships with others (Goffman, 1969; Goldstein, 2002). For example, being Black in a majority White (and perhaps racist) society can both restrict access to other social categories (e.g., occupation, community, education, etc.), and prevent the individual from being perceived accurately by others in terms of their other identities. Someone may be a doctor, a father, and a football fan, but the fact that he is also Black may 'blind' others to these additional group memberships, effectively undermining expression of these identities and thus inhibiting active participation in the associated social categories. This form of "identity constraint" might compromise the benefits of multiple group membership and place limits on individual well-being (Postmes and Branscombe, 2002). Indeed, past research has indicated that well-being is related to the extent to which the individual feels that he or she can self-express, and, in effect, be perceived by others accurately (Swann et al., 1989).

Stigmatized identities may also become a source of internal conflict and thereby place limits on the extent to which the individual can achieve a clear and cohesive self-concept. For instance, the group memberships of woman and midwife would be considered by most as highly compatible due to the fact that this profession is perceived as traditionally feminine and its workforce is predominantly female (Dimond, 2002). By contrast, being female and a doctor, or male and a midwife, may be considered more incompatible – and indeed the unexpected intersection of these categories can itself be a source of stigma (e.g., Purdie-Vaughns and Eibach, 2008; Meeussen et al., 2016; von Hippel et al., 2016; Veldman et al., 2017). Integrating these ideas with the multiple group membership literature, Brook et al. (2008) found that identity quantity (i.e., multiple group memberships) was positively correlated with well-being, but only when the specific identities involved were both important to the individual and perceived to be compatible with one another. When important identities were instead perceived to be incompatible, multiple identities were associated with reduced well-being. Thus, the compatibility of group memberships within the individual's overall self-concept appears to structure the implications of these for individual well-being. In this way, we argue that both the perceived value of group memberships to others (i.e., stigma) and the sense of incompatibility that can arise from membership in multiple stigmatized groups, should modify previously observed links between multiple group memberships and well-being.

# The Present Research

While previous research clearly demonstrates a connection between multiple group membership and a range of wellbeing factors, we argue that this relationship is likely to be complex rather than straightforward, and contingent on a variety of additional factors. In the preceding sections, we identify two general areas of identity perception and integration that are likely to moderate the relationship between multiple group memberships and well-being in some way. These include the configuration of various identities within the self-concept (i.e., complexity), and the social value attached to those identities (i.e., stigma). These things may shape the degree to which possessing multiple group memberships supports (vs. undermines) individual action and interaction in the world, and the degree to which this allows for multiple group memberships to contribute to (vs. interfere with) the subjective cohesiveness of the individual's self-concept. Further, being able to effectively articulate and express the self – both to others and within one's own mind – and to access the support of others by virtue of their shared group membership, should all in turn contribute to enhanced individual well-being and thus mediate effects of multiple group memberships on the self.

To explore the relationship between multiple group membership and well-being in these terms, we conducted two correlational survey studies. The first study looked specifically at the impact of SIC on the relationship between multiple identities and well-being, while the second study investigated the significance of identity stigma and compatibility in this relationship. Across both studies, and consistent with previous literature, we expected that belonging to multiple groups would generally be associated with increased well-being. However, we also expected that this relationship would be amplified when those multiple group memberships were relatively distinctive (i.e., high SIC), and attenuated when group memberships were instead highly overlapping. This hypothesis was tested in Study 1.

Thinking about the relationship between multiple group membership and well-being in other terms than those connected to SIC, we expected that the perception of one's identities as either devalued or valued within one's social environment would moderate links between multiple group memberships and well-being. Specifically, the translation of (multiple) group memberships to positive well-being should be more straightforward when the individual identities are valued, or when devalued identities are not visible to others. In the

context of multiple, visibly devalued (i.e., stigmatized) identities, relationships between multiple group membership and wellbeing would be negative. This hypothesis was tested in Study 2. Further, we also tested whether identity compatibility moderated the relationship between multiple group membership and wellbeing. That is, in line with past research (Brook et al., 2008), we believed that multiple identities would facilitate individual well-being, but only when those identities were compatible. This hypothesis was also tested in Study 2.

Finally, we explored plausible mediators of multiple group membership effects on well-being. In line with the above discussion, we explored perceived identity expression (i.e., the successful articulation of the self to others) and social support as potential mediators in both studies. At the outset of Study 2, and to be as comprehensive as possible, we supplemented the measure of social support to also include perceptions of social inclusion by others.

# STUDY 1

We conducted a correlational survey study designed to examine the relationship between multiple group membership and wellbeing. In addition to assessing the number of groups to which individuals belonged, we also assessed the perceived overlap vs. distinctiveness of the most important groups. Consistent with models of SIC (Roccas and Brewer, 2002), we hypothesized that multiple group membership would be positively associated with well-being when the key component identities are seen to be relatively distinct (high SIC) as opposed to overlapping (low SIC).

We were also interested in the mechanisms through which these effects might occur. Specifically, we reasoned that membership in multiple groups might, in fact, lay the foundations for a more distinctive sense of self and more practice enacting and expressing this self to others within one's social environment. That is, if the individual perceives his or her multiple social identities as discrete and separate (high SIC), then this should translate into a clearer idea of exactly what these identities represent (both individually and in combination), ultimately facilitating their accurate and effective expression. By contrast, if the individual's social identities are highly overlapping and indistinct (low SIC), communicating their precise content and meaning effectively to others may be relatively difficult. Accordingly, we measured perceived identity expression (the perceived ease and freedom to express and enact one's identities to others). Finally, other research has highlighted the role of multiple group membership in providing individuals with access to multiple actual, or expected, bases of social support and thus well-being (Haslam et al., 2008). Here, we believe that belonging to multiple distinct groups (high SIC) translates into a corresponding number of distinct channels through which unique group-based social support may be accessed. Belonging to multiple overlapping groups, however, would likely offer relatively few distinct points of social support, ultimately comprising a narrow and redundant social support base. On this rationale, we also included a measure of social support as a mediator between SIC and individual well-being.

# Method Participants

The research was conducted via an online survey advertised with flyers at various public locations (e.g., public transport, libraries, and universities), on social networking sites (linkedIn, facebook), as well as by email to personal and professional contacts. In response to this advertising, a sample of 131 adults was recruited. Of these, 19 cases had missing data and were therefore excluded from the analyses. The final sample of 112 participants included 23 males and 89 females. The majority of these (31.9%, n = 36) were aged between 18 and 25 years old, with a total of 68.2% (n = 77) of the sample being 18–35 years old. A total of 19 different nationalities were included in the sample, with the majority 75.3% (n = 85) being from Western countries such as Australia (31.9%, n = 36), the UK (30.1%, n = 34), or the US (13.3%, n = 15). The most common occupation was university student (58.6%, n = 65) followed by academic (15.3%, n = 17). The sample included 13 different ethnicities, but the vast majority of participants identified as White (75.9%, n = 85).

# Survey and Measures

The survey comprised measures relating to identity and wellbeing.<sup>1</sup> Specifically, the survey first asked participants to list as many group memberships that they could think of ('In the text box below, list as many groups that you can think of that are relevant to your daily life'). From the resulting list, they were then prompted to choose the four groups that they felt were the most important and that best defined them. We decided to ask participants to choose four group memberships (as opposed to more or less), mainly for practical reasons to do with survey length, but also because four groups allow for multiple (specifically six) comparisons between individual group memberships (see Brewer and Pierce, 2005, for a similar method). Participants then responded to a series of items focusing on various aspects of social identity (complexity, expression, and centrality) in relation to each of the four groups they had chosen (detailed below). Participants finally completed more general measures of psychological well-being.

### **Identity measures**

Social identity complexity was defined in terms of both the number of group memberships and the perceived extent of overlap between their most important (i.e., top four) groups (Roccas and Brewer (2002). A larger number of non-overlapping identities indicated higher identity complexity, whereas fewer and more overlapping identities indicated lower complexity. Thus, after listing any number of identities which defined them, each participant was asked to rate the degree of overlap between each possible pairing of his or her four most important identities (i.e., 'of people who belong to e.g., the group American, how many also belong to the group Christian?') on a 10-point Likert scale (1 = very few, 10 = nearly all). The average overlap score for all identity pairings was then calculated to obtain an overall measure of identity overlap.

<sup>1</sup>All measures and manipulations in these studies are reported in this paper.

We also measured a number of features of each of these identities (all measured on 5-point Likert scales with 1 = strongly disagree, 5 = strongly agree). First, identity centrality was measured with two items created for the study ('The group [X] is an important reflection of who I am,' 'In general, belonging to [Group X] is an important part of my selfimage'; α = 0.84). Next, identity value was rated both in terms of value to the self ('To what extent do you consider your membership with [Group X] as generally positive or negative?') and perceived value in the eyes of others ('To what extent do you think your membership with [Group X] is considered positively or negatively by others in the community/society in which you live?'). These measures were included to give an indication of the nature of the identities selected by participants.

Finally, after rating each of their chosen four identities on each of these dimensions, a number of more general questions about the self and identity were asked that did not refer to the specific groups. These items were again measured on 5-point Likert scales ranging from 1 (strongly disagree) to 5 (strongly agree). First, three items were developed for this study to measure perceived identity expression. These items focused on the person's perceived freedom to express their own identities, and the degree to which others accurately recognized how they saw themselves as a result ('In general, I feel free to fully express myself and my identity to the people around me,' 'Other people don't see me the way I want to be seen' (reversed), and 'Sometimes I feel like other people are trying to put me in a box that doesn't fit' (reversed; scale reliability α = 0.69).

#### **Well-being measures**

Psychological well-being was assessed through the General Wellbeing Index (GWBI) (Hopton et al., 1995) (scale reliability α = 0.89) using five multiple choice answer options (e.g., Q: 'In general, do you feel disheartened or sad?' A: 'All of the time,' 'Most of the time,' 'From time to time,' 'Very occasionally,' 'Not at all'). A three-item measure of perceived access to social support was developed for the study and measured on 5-point Likert scales (scale reliability α = 0.82), 'To what extent do you feel that you have family or friends so close to you that you can count on them if you have serious problems?', 'How much concern/interest do people show in what you are doing?', 'How difficult would it be for you to get practical help from neighbors if you should need it?'.

## Results

#### Descriptive Findings

On average, respondents listed a total of approximately seven (M = 7.05, SD = 3.18) groups that they believed defined them in some way. The nature of the groups reported varied widely among participants and included gender, sexuality, hobby, profession, ethnicity, etc. The perceived overlap between participants' most important group memberships was generally low (M = 3.93, SD = 1.95 on a 10-point scale), indicating relatively high SIC. The identity and well-being variables yielded average scores significantly higher than the scale midpoint of 3.00 (see **Table 1**). Thus, participants were generally thinking about important, positively valued group memberships.

Next, mean correlations for the main identity and well-being variables were calculated to provide a preliminary assessment of any statistically significant relationships (see **Table 1**). Aside from a significant correlation between identity overlap and identity value (others), indicating that people with more overlap between identities (low complexity) perceived their group memberships as less socially valued, there were no significant correlations between identity overlap and any of the well-being measures. Nor were there any significant correlations between identity quantity and any other variable (see **Table 1**). To account for any potential variation within each mean identity measure, we also examined the association between the identity and wellbeing variables at the level of each individual identity that participants listed (i.e., their four most important identities). Here, we found no significant correlation between any measure of identity overlap, identity centrality, or identity visibility, and any of the outcome variables. Identity 2 and 4, however, did correlate significantly, but moderately, with social support, social inclusion, and well-being (see **Table 2**).

#### Regression Analyses

Regression analyses were conducted to further ascertain the nature of the relationship between the variables, and to test the specific hypothesis that identity complexity (the combination of multiple, distinctive groups) has consequences for individual well-being. To test this hypothesis, well-being was used as the


<sup>∗</sup>p < 0.05, ∗∗p < 0.01. †Mean departs from scale midpoint (identity overlap = 5.5; all others = 3) significantly at p < 0.001. Identity overlap was measured on a 10-point Likert scale; all other measures were taken on 5-point scales.


TABLE 2 | Correlations between identity and well-being measures.

<sup>∗</sup>Correlation is significant at the 0.05 level (two-tailed).

main dependent variable (DV), while identity overlap, identity quantity, and their interaction were entered as the independent variables (IV). Prior to the analysis, identity overlap and identity quantity were mean centered and an Overlap X Quantity interaction variable was computed by multiplying the centered scores. Regression analyses were conducted in which the main effect terms were entered at the first step, followed by the interaction term at Step 2.

#### **Well-being**

A regression analysis assessing the impact of identity overlap and identity quantity on well-being revealed no significant main effects (β = 0.06, p = 0.57, 95% CI low = −0.05, high = 0.11; β = 0.11, p = 0.25, 95% CI low = −0.03, high = 0.07, respectively). Inclusion of the interaction term, however, increased the overall variance explained, R 2 change = 0.05, p = 0.02, and the interaction itself was significant at this step, β = −0.22, p = 0.02, 95% CI low = −0.05, high = −0.00.

In order to deconstruct the interaction, the effect of identity quantity was examined at high (+1 SD) and low (−1 SD) identity overlap (see **Figure 2**). This revealed a significant main effect for identity quantity at low identity overlap, β = 0.31, p = 0.02, 95% CI low = 0.05, high = 0.39, while no such effect was evident at high identity overlap, β = −0.09, p = 0.47, 95% CI low = −0.14, high = 0.09. Thus, it would appear that multiple group memberships are associated with enhanced well-being only at low levels of identity overlap (i.e., high complexity).

#### **Perceived identity expression**

This analysis was repeated on the measure of perceived identity expression. The analysis revealed a significant main effect of identity overlap, β = 0.20, p = 0.03, 95% CI low = 0.01, high = 0.21, and a significant Overlap × Quantity interaction, β = −0.42, p < 0.001, 95% CI low = −0.08, high = −0.03.

The main effect of identity quantity was significant at both low and high levels of identity overlap (**Figure 3**). At low identity overlap, identity quantity was positively associated with perceived identity expression, β = 0.28, p = 0.01, 95% CI low = 0.01, high = 0.08, whereas at high identity overlap, this relationship was negative, β = −0.47, p = 0.001, 95% CI low = −0.11, high = −0.02. Thus, having many different and distinct ways of identifying oneself seemed to facilitate the individual's perceived ability to freely and clearly express their identities. Identifying with many overlapping social categories, however, appeared to inhibit perceived identity expression. The particular nature of the interaction should also be noted, though. Specifically, it would appear that people with few, but highly overlapping identities felt freer to express themselves than people with few, but nonoverlapping identities, t(30) = −2.89, p = 0.01. This effect appeared to reverse as number of identities increased such that people with many distinct identities felt freer to express these identities than people with many overlapping ones. Although this trend is apparent the effect was not statistically significant, t(39) = 1.45, p = 0.15.

#### **Social support**

The analysis performed on the measure of social support, also generated effects comparable to those found for well-being. Again, although there were no significant main effects of either identity overlap or identity quantity, the Overlap × Quantity interaction was again significant, β = −0.21, p = 0.03, 95% CI low = −0.06, high = −0.01. Further analysis revealed that identity quantity was positively associated with social support at low, β = 0.30, p = 0.01, 95% CI low = 0.00, high = 0.06, but not high, β = −0.08, p = 0.59, 95% CI low = −0.04, high = 0.00, levels of identity overlap (see **Figure 4**).

We also assessed the extent to which the identity value variables (perceived by the self and others) moderated the relationship effects of identity quantity, identity overlap, and their interaction on well-being. These analyses revealed no significant effects, all ps > 0.50.

#### Mediation Analysis

Given the fact that our theoretical framework placed wellbeing as the ultimate outcome variable, we considered whether

the effects of identity complexity on the DV were mediated by perceived identity expression and social support. To test the mediating role of perceived identity expression, this was included in the regression equation predicting wellbeing along with identity overlap, identity quantity, and the Overlap × Quantity interaction. In this analysis, the previously significant Overlap × Quantity interaction became non-significant, β = −0.08, p = 0.44. This was replaced by a significant main effect for perceived identity expression on well-being, β = 0.34, p < 0.001, suggesting that perceived identity expression mediated the effect of the Overlap × Quantity interaction on well-being.

The same analysis was repeated with social support included as a possible mediator. Similarly, the previously significant interaction became non-significant, β = −0.15, p = 0.10. Again, this interaction was replaced by a significant effect of the mediator (social support) alone, β = 0.35, p < 0.001, suggesting that social support also mediated some of the effects of the Overlap × Quantity interaction on well-being. This pattern of dual mediation via perceived identity expression and social support was confirmed in a bootstrapping analysis testing the significance of the indirect paths (Preacher and Hayes, 2008). Path analyses of the relationship are presented in **Table 3**, and the indirect effects of Overlap × Quantity on well-being via perceived identity expression and social support are shown in **Table 4**. Both indirect effects are significant at p < 0.05 (95% CI) (see **Table 4**). These relationships are depicted graphically in **Figure 5**.

# Discussion

The aim of Study 1 was to examine the relationship between multiple group memberships and well-being. In line with a growing body of research (Haslam et al., 2008; Binning et al., 2009; Jones and Jetten, 2011; Jetten et al., 2012), our results demonstrated that multiple group memberships contribute to well-being. But, importantly, this contribution

TABLE 3 | Impact of Quantity × Overlap on mediator variables and of mediator variables on well-being.


TABLE 4 | Indirect effects of Identity Quantity on Well-being via mediators and at different levels of Identity Overlap.


5000 bootstrap samples, <sup>∗</sup> significant at p < 0.05.

was dependent on more than the sheer number of group memberships the individual has access to. Consistent with theorizing about SIC (Roccas and Brewer, 2002), the effect of multiple group memberships was contingent on perceived identity overlap. Specifically, it was multiple non-overlapping (i.e., distinctive) group memberships that contributed most positively to individual well-being.

Moreover, this effect was mediated by the individual's perceived access to social support and the reported ease of self-expression. Consistent with the ideas presented in the introduction, it seems likely that belonging to social categories that are well defined and distinctive facilitates both the individual's ability to articulate to their social world who they are, as well as their access to the social support that similar others might be able to provide. On the other hand, if social categories are highly overlapping, and thus consist of largely the same groups and people, both the ability to express the self clearly, and the ability to access multiple distinct sources of social support might be compromised. Beyond this, the psychological benefits of clear perceived identity expression can be explained in terms of self-verification processes. Past research suggests that people generally desire to be seen by others as they see themselves (i.e., self-verification), and that self-verification is psychologically a positive experience (Swann et al., 1989). Effective perceived identity expression is likely to be instrumental in attaining self-verification, and the process of enacting the self successfully to multiple others could help explain the relationship between perceived identity expression and well-being. The fact that perceived identity expression is facilitated by belonging to multiple distinctive groups, and undermined by membership in multiple overlapping groups, is an interesting extension of past work on these ideas. Specifically, these findings highlight the fact that multiple identities are not necessarily beneficial and in some instances maladaptive (i.e., when identities are non-distinct, possessing fewer identities appears to be better in terms of self-expression and through this well-being).

# STUDY 2

Study 1 highlights the importance of the distinctiveness of the groups to which the individual belongs for unlocking well-being benefits of multiple group memberships. Given our overarching aim to map out the factors that underpin the relationship between multiple group membership and well-being, however, we designed a second study looking at other potential mediators and moderators. In the introduction we highlighted additional features of multiple group membership that might also frame the relationship between these and well-being. In particular, we suggested how concerns around stigma can complicate the individual's ability to enact group memberships and reap the associated benefits, and how stigma attached to one group membership might create incompatibilities with others. Thus, we proposed that belonging to multiple groups might not be so beneficial for well-being when the component groups are perceived to be devalued in the eyes of others (i.e., stigmatized: Schmitt et al., 2014; see also Brook et al., 2008).

Although we did ask about perceived social value in Study 1, it should be noted that most reported identities were perceived as valuable, both to the self and to others. In addition, the format of the questionnaire in Study 1, with its focus on uniqueness via questions about identity overlap, is likely to have prioritized concerns around the distinctiveness of the identities that comprise the self-concept more than questions of social stigma. In light of this, the aim of Study 2 was to shift the focus from questions about identity distinctiveness to the social meaning of identity in terms of social stigma, and to explore the consequences of this for relationships between multiple group memberships and well-being.

Given the above rationale, we again examined the relationship between multiple identities and well-being in Study 2. Here, however, our questions focused on the broader context of group memberships, rather than just their distinctiveness. Specifically, instead of asking people to compare their various group memberships in terms of overlap (which does not itself communicate social value), in this study we focused our questions on the visibility, value, and (in)compatibility between identities. Our reasoning was that assessing these features of identities, alone and in combination, should call to mind a different set of concerns than questions about overlap/distinctiveness. Specifically, questions about compatibility highlight the normative and socially determined aspects of multiple groups within the wider social environment (Brook et al., 2008). Similar social evaluative concerns are also activated by questions

about the visibility of one's identities to others, questions that were also not asked in Study 1. Indeed, past research suggests that belonging to visibly stigmatized groups (e.g., physically disabled, ethnic minority) as opposed to concealable ones (e.g., homosexuality) can have different consequences for individual well-being (Quinn, 2006; Quinn and Chaudoir, 2009; Quinn and Earnshaw, 2013). Thus, in this study, we re-framed our identity questionnaire in ways that allowed us to investigate additional features of identity that might play a role in determining links between multiple group memberships and well-being. Namely, the value of one's identities in the eyes of others (i.e., visible vs. invisible identity value) and the degree to which those identities are compatible or not.

Consistent with past research, we expected that the wellbeing consequences of (multiple) group membership would be dependent on (1) the social value of these groups (i.e., whether some of these were stigmatized), (2) the degree to which these group memberships were obvious to others (visible vs. invisible), and (3) whether the group memberships and their associated identities were compatible. Specifically, we expected that multiple group membership would be positively related to well-being only when the component identities were visibly socially valued rather than visibly devalued. Membership in multiple visibly stigmatized groups was thus expected to undermine individual well-being. We further reasoned that belonging to multiple, stigmatized groups should be associated with a sense of incompatibility within one's matrix of group memberships. This sense of incompatibility should, in turn, relate negatively to well-being. In addition, and similar to Study 1, we expected that possessing visibly stigmatized identities would strongly determine interactions with others and therefore make it difficult to express the self fully (i.e., in terms of the other identities that one might possess) and to access social support from others. Feeling unable to express one's identities, and having lower perceived social support, were thus expected to mediate well-being. Finally, we also reasoned that one way in which identity stigma may negatively impact on the relationship between multiple group membership and well-being is through social exclusion. Specifically, being allowed access to the groups to which one belongs may become difficult if one or more of those groups is visibly stigmatized (Ellemers and Barreto, 2006). In order to test these ideas, we included a two-item measure of social inclusion. Thus, in this study, we test three possible mediators of well-being effects: Self-expression, social support, and social inclusion.

# Method

#### Participants

The survey was conducted online and advertised in an identical fashion to Study 1. A sample of 144 adults was recruited. Of these, 40 cases were missing data and were therefore not included. The final sample of 104 participants included 17 males and 86 females. The majority of these (55.6%, n = 58) were between 20 and 30 years old. A total of 20 different nationalities were included in the sample, with the majority 67.3% (n = 70) being from Australia (40.4%, n = 42) and the UK (26.9%, n = 28). The most common occupation was university student (63.5%, n = 66) followed by 'other' (13.5%, n = 14) and academic (11.5%, n = 12). The sample included 11 different ethnicities, but the vast majority of participants identified as White (76%, n = 70).

### Survey and Measures

Similar to Study 1, participants were instructed to first list as many of their social identities that they could think of. They then chose the four most important identities before responding to a series of items focusing on various aspects of social identity, this time in terms of compatibility, visibility, and value. Participants then completed measures of psychological well-being.<sup>2</sup>

#### **Identity measures**

Identity quantity was measured in an identical fashion to Study 1. Identity visibility was measured for the top four identities with a single item, created for the study ('To what extent do you feel that your membership with the category [X] is generally obvious to others?' 1 = Not at all, 5 = Very much so). Similarly, the measure of identity value was gauged with a single item, 'To what extent do you think your membership with [X] is considered positively or negatively by others in the community/society in which you live?' The item was measured using a 5-point Likert scale with 1 = Generally negatively, 5 = Generally positively). We measured identity centrality using the same two-item measure as in Study 1 (α = 0.84).

In Study 1, participants were asked to compare important groups in terms of their overlap. In this study, those comparisons between important groups were instead based on perceptions of the compatibility between groups, something we argue is likely to have shifted the way identities were framed for the questions that followed, away from a focus on individual uniqueness (Study 1) and toward one of the social and normative meaning of identities. Specifically, the degree of perceived compatibility between each possible pairing of participants' four most important identities (i.e., 'Thinking about [group X] and [group X], how easy or difficult is it to belong to these two groups/social categories at the same time?') was measured on 5-point Likert scales (1 = very difficult, 5 = very easy). The average compatibility score for all identity pairings was then calculated to obtain an overall measure of identity compatibility.

After rating each of their chosen identities on these dimensions, a number of more general questions about the self and identity were asked that did not refer to the specific groups. The Study 1 measures of perceived identity expression (α = 0.79) and perceived access to social support (α = 0.82) were used. Finally, we also included a two-item measure of perceived social inclusion that was developed for the study (scale reliability α = 0.93) ('Generally, I feel included by my peers in the community,' 'Generally, I feel accepted by my peers in the community'). Psychological well-being was assessed using the same GWBI (Hopton et al., 1995) as in Study 1.

<sup>2</sup>All measures and manipulations in these studies are reported in this paper.

# Results

#### Descriptive Findings

fpsyg-08-01038 June 23, 2017 Time: 14:50 # 11

On average, respondents listed a total of just over seven (M = 7.36, SD = 3.38) social identities that they believed defined them in some way. Similar to Study 1, participants reported a wide variety of groups, including gender, religion, profession, ethnicity, etc. The perceived compatibility of participants' identities was generally high (M = 4.10, SD = 0.82), with the most compatible and incompatible identities exemplified by match-ups including female and mother, or working class and Ph.D.-student, respectively. Similarly, the other identity and outcome variables yielded average scores significantly higher than the scale midpoint of 3.00, except for perceived identity expression [t = −0.38 (103), p = 0.70; see **Table 5**]. Examples of highly visible identities, included those based on gender and ethnicity, while invisible identities related to, for instance, profession or politics. Minority ethnicity was also attributed low value, whereas university graduate, for example, was considered high value. In general, participants listed socially valued, highly visible, and compatible identities.

Next, mean correlations for the main identity and outcome variables were calculated (see **Table 5**). Significant positive correlations were evident between identity quantity and identity value (β = 0.23, p < 0.05), social support (β = 0.26, p < 0.05), and social inclusion (β = 0.26, p < 0.05). Similarly, identity compatibility (β = 0.37, p < 0.05) and identity value (β = 0.28, p < 0.05) were both positively associated with social inclusion. Perceived identity expression also correlated positively with both social support (β = 0.23, p < 0.05), and social inclusion (β = 0.23, p < 0.05). Interestingly, identity visibility and value correlated positively (β = 0.76, p < 0.01). While this association may at face value indicate that these constructs are in some way dependent, we argue that the high correlation is due to the notion that highvalue identities are more likely to be rendered visible by the individual. That is, Christians may signal their religious identity openly (e.g., attending church service, wearing a cross, etc.) in a majority Christian society, but not in one critical of that particular faith. Overall, the participants who listed greater numbers of identities generally perceived their identities to be of high value, felt socially included, and had greater access to social support. Further, those who described their identities as highly compatible and of high value felt more socially included (see **Table 5**). However, and as in Study 1, identity quantity per se was not correlated with well-being. Similar to Study 1, we also examined the association between the identity and well-being variables at the level of each individual identity that participants listed (i.e., their four most important identities). These correlations largely mirrored those described above, with all but the last level of identity compatibility and first level of identity value correlating positively with social inclusion. There was also an inverse relationship between identity visibility and perceived identity expression at identity level 3 and 4 (see **Table 6**).

#### Regression Analyses

Regression analyses were undertaken to further determine the specific nature of the relationships among the variables. Wellbeing was used as the main DV, with social inclusion, social support, and perceived identity expression as secondary DVs. The IVs identity quantity, identity visibility, identity value, and identity compatibility and their interaction were of principal interest in accordance with the stated hypotheses. Regression analyses for all DVs were conducted in which the main effect terms were entered at the first step, followed by the interaction term at Step 2.

#### **Well-being**

There were no significant main effects of identity quantity (β = −0.03, p = 0.76, 95% CI low = −0.05, high = 0.03), identity visibility (β = 0.13, p = 0.19, 95% CI low = −0.08, high = 0.25), identity value (β = 0.05, p = 0.63, 95% CI low = −0.15, high = 0.27), or identity compatibility (β = 0.07, p = 0.43, 95% CI low = −0.10, high = 0.23) on well-being. Nor were there any interactions among these variables (all p > 0.05).

#### **Social inclusion**

For social inclusion, main effects were evident for identity quantity (β = 0.19, p = 0.04, 95% CI low = 0.00, high = 0.10) and identity value (β = 0.23, p = 0.02, 95% CI low = 0.02, high = 0.60), as well as for identity compatibility (β = 0.31, p = 0.001, 95% CI low = 0.17, high = 0.61). Further, in line with our hypothesis that the relationship between multiple group membership and well-being was moderated by identity visibility and identity value, we tested a three-way interaction of identity quantity, identity value, and identity visibility. The interaction variable was significantly related to social inclusion (β = 0.15,


<sup>∗</sup>p < 0.05, ∗∗p < 0.01. †Mean departs from scale midpoint (3) significantly at p < 0.001.


TABLE 6 | Correlations between identity and well-being measures at the level of individual identities.

<sup>∗</sup>Correlation is significant at the 0.05 level (two-tailed). ∗∗Correlation is significant at the 0.01 level (two-tailed).

p = 0.001, 95% CI low = 0.01, high = 0.29). To deconstruct the interaction, the correlation between identity quantity and social inclusion was examined at low and high identity value and visibility. When identity value was low (i.e., participants were reporting stigmatized identities), there was a significant positive relationship between identity quantity and social inclusion, but only when identity visibility was also low (β = 0.20, p = 0.00, 95% CI low = 0.01, high = 0.09) rather than high (β = −0.08, p = 0.19, 95% CI low = −0.07, high = 0.02). When identity value was high (i.e., non-stigmatized/positive), there were only weak (non-significant) positive relationships between identity quantity and social inclusion regardless of the level of identity visibility (Low: β = 0.05, p = 0.22, 95% CI low = −0.01, high = 0.04; High: β = 0.08, p = 0.07, 95% CI low = −0.01, high = 0.06; see **Table 7** and **Figure 6**). Thus, multiple group memberships contributed to a sense of social inclusion most for individuals with stigmatized but invisible identities.

#### **Social support**

The analysis on social support revealed main effects of identity quantity (β = 0.22, p = 0.03, 95% CI low = 0.01, high = 0.10), but not identity visibility (β = −0.01, p = 0.89, 95% CI low = −0.19, high = 0.16), identity value (β = 0.07, p = 0.50, 95% CI low = −0.19, high = 0.16), or identity compatibility (β = 0.09, p = 0.29, 95% CI low = −0.08, high = 0.20). However, identity quantity, visibility, and value again interacted to impact significantly on social support (β = 0.14, p < 0.001, 95% CI low = 0.00, high = 0.08). Deconstructing the interaction, we examined the relationship between identity quantity and social support at low and high identity value and in turn at low and high levels of identity visibility. This revealed several significant main effects. Replicating the previous pattern, when identity value was low (stigmatized identities), there was a significant positive relationship between identity quantity and social support when identity visibility was also low (β = 0.14, p = 0.01, 95% CI low = 0.01, high = 0.09), but not when identity visibility was high (β = 0.02, p = 0.55, 95% CI low = −0.05, high = 0.01). When identity value was instead high, there was a significant positive relationship between identity quantity and social support when identity visibility was also high (β = 0.12, p = 0.00, 95% CI low = 0.00, high = 0.08), but not when visibility was low (β = −0.06, p = 0.28, 95% CI low = −0.01, high = 0.03) (see **Figure 7**). The results thus indicated a positive correlation between identity quantity and social support for people with stigmatized, but invisible, identities, and for people with valued and visible identities.

#### **Perceived Identity expression**

The analysis on perceived identity expression, revealed only a significant three-way interaction between identity quantity, identity value, and identity visibility (β = 0.13, p < 0.01, 95% CI low = 0.01, high = 0.07). When identity value was low



(i.e., stigmatized identities), identity quantity was negatively correlated with perceived identity expression when identity visibility was also high (β = −0.20, p = 0.00, 95% CI low = −0.06, high = −0.00), but not when visibility was low (β = −0.00, p = 0.20, 95% CI low = −0.03, high = 0.02) (see **Table 7** and **Figure 8**). Further, when identity value was high, identity quantity correlated positively with perceived identity expression when identity visibility was also high (β = 0.10, p = 0.03, 95% CI low = 0.00, high = 0.04), but not when visibility was low (β = 0.05, p = 0.21, 95% CI low = −0.01, high = 0.01). As such, multiple identities inhibited social expression when identities were stigmatized and visible. When identities were socially valued and visible, however, the quantity of identities facilitated expression.

# FIGURE 8 | The association between identity quantity and ease of identity expression at high and low visibility and value.

#### **Identity compatibility**

Given the lacking contribution of identity compatibility as a moderator variable in the above analyses, we explored the possibility that perceived compatibility might be an outcome of the features of identities – that is, that possessing stigmatized identities might give rise to a sense of identity incompatibility, mediating the effects of stigma on well-being. Thus, we theorized that stigmatized identities may be more difficult to square with other identities in the self-concept, thus creating a sense of incompatibility. While there were no significant main effects of identity quantity, identity value, or identity visibility, the threeway interaction was again significant, β = 0.12, p = 0.00, 95% CI low = 0.00, high = 0.10. When identity value was low (i.e., stigmatized identities), identity quantity was positively related to identity compatibility when identity visibility was also low (β = 0.18, p = 0.00, 95% CI low = 0.00, high = 0.13). However, this relationship reversed in the context of low value, but high visibility identities (β = −0.10, p = 0.06, 95% CI low = −0.10, high = −0.00) (see **Table 7** and **Figure 9**). Thus, having relatively many identities facilitated a sense of identity compatibility when identities were stigmatized but invisible to others. The opposite was true, when stigmatized identities were visible, although this relationship was weaker and only marginally significant.

While the above analyses are interesting and in line with predictions, several somewhat surprising patterns should be noted. Specifically, it would appear that people who had relatively few identities were worse off in terms of social support, perceived identity expression, and compatibility when these identities were devalued and invisible than if they were devalued and visible (see **Figures 7–9**). The comparable effects on these variables appeared to reverse, however, as the number of identities increased, such that people with many identities had access to more social support and found it easier to self-express if those identities were devalued and invisible rather than devalued and visible. Speculatively, this suggests that certain forms of stigma can also be self-protective (e.g., Crocker and Major, 1989).

#### Mediation Analysis: Social Inclusion, Social Support, Perceived Identity Expression

Despite the absence of any association between the IVs and the key DV (well-being), we nonetheless explored the patterns

on the expected mediator variables of social inclusion, social support, and perceived identity expression. Past research has advocated testing for mediation in the absence of direct effects by bootstrapping on the grounds that the causal steps approach is considerably low in power, and thus relatively unlikely to detect an effect. In addition, when research is guided by a theoretical model, it is preferable to test the degree to which that model – as originally conceived – fits the data. Although it may seem counter-intuitive to do this when there are no discernible direct effects on the dependent measure, the absence of these also does not preclude validity of the original model (Hayes, 2009, 2012; Hayes and Preacher, 2014). Because PROCESS does not allow multiple parallel mediators in models that involve interactive IVs, we tested each mediator separately. Thus, we report the results of five mediation models. As hypothesized, the results of three mediation models run using PROCESS (Hayes, 2012; Models 3 and 12), indicated indirect effects of the interaction variable on well-being via social support, social inclusion, and perceived identity expression, respectively. Further, in another two models, we found that perceived identity expression and identity compatibility, respectively mediated the relationship between the interaction variable and social inclusion. We explain this with the idea that belonging to potentially incompatible groups (e.g., gay and Muslim) will likely make it difficult for the individual to express the associated identities equally and freely, and obstruct his or her inclusion in either or both of these groups (Jaspal and Cinnirella, 2010). We have integrated each of our mediation models into a single figure for a complete representation (**Figure 10**), and presented the appropriate statistics in **Tables 7–10**.

These mediation effects were confirmed in bootstrap analyses testing the significance of the indirect paths. Path analyses and the indirect effect of identity quantity × identity value × identity visibility on well-being via perceived identity expression, social inclusion, and social support are presented in **Tables 7**, **8**. Path analyses and the indirect effect of the interaction variable via perceived identity expression and identity compatibility on social inclusion are shown in **Tables 9**, **10**, respectively. The indirect relationships for well-being are significant at p < 0.05 (social support, boot 95% CI low = 0.01, high = 0.01; social inclusion, boot 95% CI low = 0.00, high = 0.09; perceived identity

TABLE 8 | Indirect effects of identity quantity on well-being via mediators, and at different levels of social acceptance and identity visibility.


<sup>∗</sup>p < 0.05.

TABLE 9 | Impact of three-way interaction on mediators, and mediators on social inclusion.


expression, boot 95% CI low = 0.00, high = 0.06) as are those for social inclusion (perceived identity expression, boot 95% CI low = 0.0049, high = 0.0964; identity compatibility, boot 95% CI low = 0.0025, high = 0.0838).

Finally, we regressed well-being onto all of the IVs and their interaction, with the mediator variables simultaneously included as predictors of well-being. As expected, we found a significant association between the outcome variable and social support (β = 0.27, p = 0.005) and social inclusion (β = 0.32, p = 0.003). Perceived identity expression was marginally



significantly associated with well-being (β = 0.18, p = 0.08), while the relationship between identity compatibility and wellbeing was non-significant. Identity compatibility did, however, significantly mediate social inclusion (β = 0.31, p = 0.001), as did perceived identity expression (β = 0.26, p = 0.004).

# Discussion

As a preamble to our discussion of the results of Study 2, we feel the need to restate the exploratory nature of the present research. Specifically, we point out the obvious limitation of conducting three-way interactions with our relatively small sample size, and thus recommend caution in the interpretation of our results. In comparison to Study 1, which foregrounded questions about the overlap between, and therefore distinctiveness of, the individual's configuration of group memberships, Study 2 focused respondents more on the socially prescribed features of group memberships (i.e., the degree to which it is hard to belong to two groups simultaneously). This, we argue, is likely to have brought any concerns around stigma to the fore when individuals contemplated multiple group memberships and their self-concept in relation to these. In line with the assumed focus on stigma in this study, the results revealed consistent interactions among the number of groups to which an individual belonged and the social value and visibility of those groups. Multiple group memberships appeared to be a resource especially for individuals who belonged to groups that were relatively devalued (i.e., stigmatized) but otherwise invisible to others. Under these conditions (i.e., invisible stigma), belonging to multiple groups enhanced the ability to express the self, increased perceived access to social support as well as the individual's sense of social inclusion. In comparison, when group memberships were stigmatized and visible, the benefits of multiple group memberships for social inclusion and social support did not accrue, and multiple group membership undermined the perceived ability to express one's identities. A parallel pattern was observed on perceived identity compatibility. Conversely, in the context of valued identities, the benefits of multiple group memberships seemed more straightforward, and if anything were enhanced by the visibility of those valued identities.

Although there were no direct effects of the IVs on our ultimate DV, well-being, analysis of the indirect paths revealed a similar pattern of mediation to Study 1. Again, the extent to which multiple group memberships contributed to perceived identity expression and social support indirectly determined their benefits for well-being. Elaborating this picture, identity compatibility and perceived identity expression mediated feelings of social inclusion, which in turn also mediated well-being effects.

While the results are broadly consistent with expectations, certain unexpected and counter-intuitive patterns should be noted. Specifically, for people with few and stigmatized group memberships, feelings of compatibility, self-expression, and social support were greater if the associated identities were visible rather than invisible (**Figures 7–9**). Given their unexpected nature, it would be unwise to over-interpret these patterns. Nonetheless, in terms of social support and perceived identity expression, this could speculatively be due to the social isolation arising as a consequence of belonging to few and stigmatized identities which are also invisible. Under these circumstances individuals might feel especially alone. The ability and willingness to reach out and connect with similar others for support, might be hindered when this involves exposing devalued and concealed identities to others, especially when one has little else to draw on for one's self –concept (Barreto and Ellemers, 2003). By contrast, belonging to a devalued but visible group, represents a situation in which the individual has no control over the prominence of his or her stigmatized identity. In the context of a self-concept comprising only few group memberships, this may be a relatively good thing, as self-expression and accessing social support (perhaps more from similar others; Frable et al., 1998) is likely to be relatively clear-cut. Indeed, some past research suggests that stigma can sometimes be selfprotective (Crocker and Major, 1989), and that one way in which individuals might deal with devaluation from others is through defining themselves more strongly in terms of that stigmatized identity (Schmitt et al., 2003). In this way, being defined, and defining one's self, in terms of simple yet stigmatized identities might not always be psychologically costly. Similarly, with respect to identity compatibility, belonging to few groups that are highly visible and stigmatized leaves the individual with little other choice than to accept their obvious membership in these groups, and somehow combine the associated identities into a congruous self-concept. If these identities are invisible, however, the individual may not feel the same outside pressure or need to merge their identities into a cohesive self, leaving the incompatibility unresolved. Of course, the correlational and self-reported nature of our study also leaves open the possibility that those who are most vulnerable – for example, those with a very singular and stigmatized sense of self – might also withdraw from the identities involved, and from questions about them. This could inflate the apparent value of being visibly stigmatized and relatively isolated because it is only those who have come to terms with this situation who are also willing to report their experiences. Similarly, those identities that are visible yet stigmatized (e.g., ethnicity in our sample), differ on many dimensions to those identities that are stigmatized yet invisible (e.g., mental illness (specifically depression) in our sample). It is also plausible that some of these unexpected findings stem from the specifics of the identities people were contemplating. Our approach, which collapses across specific group memberships to consider broad features of relevance to stigma and identity, necessarily loses some of this important detail.

# GENERAL DISCUSSION

fpsyg-08-01038 June 23, 2017 Time: 14:50 # 16

The present research focused on the relationship between multiple group membership and well-being. Although past findings suggest a positive relationship between these constructs, the focus of our two studies was to explore in more detail the complexity of identities and how this might shape the benefits of these when they are combined. Specifically, in two studies, we explored how the benefits of multiple group membership might depend on the extent to which individual component identities are distinct (Study 1), and on how concerns arising from social stigma, and associated identity incompatibilities, may disable the benefits of multiple group memberships (Study 2).

Study 1 showed that the advantages of multiple group membership for well-being were moderated by SIC (Roccas and Brewer, 2002). Membership in multiple groups promoted individual well-being most when (the most central) individual group memberships were perceived to be non-overlapping and distinct (i.e., high SIC). The effect of multiple group membership in combination with SIC was mediated through perceived identity expression and social support. That is, individuals with multiple group memberships characterized by distinctiveness also felt more confident in how they could express their self to others and perceived more sources of social support, both of which contributed positively to well-being.

In Study 2, we aimed to activate a different interpretive framework for individuals as they contemplated their multiple group memberships and what these might mean for their self. Rather than highlighting questions about overlap and the distinctiveness of one's identity combinations, the questions in our second study highlighted issues of social stigma, including the value and visibility of one's identities to others, as well as the perceived conflict between one's identity combinations. In this study, individuals benefited most from multiple group membership when their (most central) identities were stigmatized, but invisible, or valued and visible to others. Further, the results indicated that any well-being consequences of identities were indirect and mediated through perceived identity expression and social support (as in Study 1), as well as social inclusion and, to a lesser extent, identity compatibility.

Integrating the results from Studies 1 and 2, these findings add to previous discussion and general theory of the benefits of multiple group membership for individual well-being (Roccas and Brewer, 2002; Iyer et al., 2009; Jones and Jetten, 2011). As outlined in the Introduction, past research has tended to focus on the number of multiple identities held in the selfconcept as a central determinant of well-being (Jetten et al., 2010; Jones and Jetten, 2011). While other studies have suggested the significance of the specific features and characteristics of social identities in terms of their well-being benefits, these have seldom been incorporated in research on the benefits of multiple group memberships. The mechanisms by which the previously observed positive relationship between multiple group membership and well-being is facilitated, or undermined, are thus somewhat unclear. Our research, sheds some light on the specific nature of this association. While Study 1 demonstrates the general value of multiple distinctive identities, Study 2 highlights how the benefits of multiple identities are contingent on socially prescribed identity characteristics such as stigma and incompatibility. In so doing, the results generated by these two studies together identify a set of factors that, in different situations and capacities, may both facilitate and impede the previously demonstrated benefits of belonging to multiple groups.

Considering the patterns across studies – patterns that were simultaneously convergent (with respect to the underlying processes) and divergent (with respect to which factors moderated the impact of multiple group memberships) – it could be suggested that each study foregrounded different frameworks that guided how individuals interpreted the meaning of their group memberships. Study 1 focuses purely on the descriptive facets of group memberships and their boundaries – that is, who is and who is not a member of various groups? Within this framework, it would appear that the more and the more distinct the component identities, the better off the individual is in terms of well-being, while the more and the more overlapping the identities, the worse off the individual is. Study 2, on the other hand, emphasizes the social meaning of group memberships (rather than their boundaries and membership overlap) by looking at how identities are viewed by others and whether it is simple or complicated to belong to multiple groups at once. Once the social meaning of these group memberships is accounted for in this way, the nature of the relationship changes and the benefits of group membership becomes more defined by the social context in which the individual exists and the social perception of the given group memberships. That is, in this framework the well-being consequences of multiple group membership is more about what the current social environment permits based on the socially determined value of the specific groups to which the individual belongs. Thus, beyond that of the individual's own interpretation and acknowledgment of the boundaries of his or her group memberships (i.e., distinctiveness vs. overlap), the present broader social environment also features as both a potential obstacle, inhibitor (in terms of compatibility), and facilitator of the benefits of group membership.

# Limitations and Future Directions

While these two studies have highlighted different ways in which multiple group memberships might affect well-being, our results are by no means exhaustive. Importantly, the findings we have offered are novel, and therefore tentative, especially given the relatively small samples we have drawn on to test our predictions. While we believe that there are sensible theoretical foundations for the relationships we describe, the estimates of these effects from the current data are likely to be imprecise and potentially unstable. As such, further empirical investigation is needed. Our more modest hope is that by discussing the different ways in which multiple groups might contribute to individual well-being, and providing demonstrations of divergent possibilities, future

researchers will take up these ideas and elaborate on the critical questions of when multiple groups are a benefit, and when they are a burden, to the individual.

Our research designs also preclude asking more complicated questions about the place of multiple group memberships within individual selves. For example, our studies did not allow for any extensive insight into the internal negotiation of multiple identities in participants. In the context of many vs. few identities, it would be interesting to look at how individuals reason and negotiate concealable vs. obvious stigmatized identities or overlapping vs. distinct identities, and whether this may have an impact on the well-being effects of multiple group memberships. This could allow for a greater understanding of our somewhat counter-intuitive findings where people with few and highly overlapping or visibly stigmatized identities somehow were better off than people with few and distinct or valued identities.

Further, the method used to measure identity centrality, value, and visibility was based on averaging the individual scores for the four identities that participants were asked about. Acknowledging the fact that people might belong to groups that differ greatly in terms of these characteristics, this method might be somewhat problematic. For example, the effects of belonging to a heavily stigmatized category might not register proportionately if the individual rating of the associated identity is counterweighed by three positive or even neutral identities. Especially in terms of identity visibility, averaging presents a special problem by not accounting for exactly which identities are visible or invisible. As our data indicates, this matters a great deal when it comes to identity value. In other words, using averages to estimate overall identity value, visibility, and importance, might muddle the true effects of these variables. However, we also reasoned that interpreting the data at an identity level would be similarly complicated as we effectively would be comparing across participants, and thus across varied identities. Considering the fact that we found significant correlations in spite of these limitations, one could speculate that our results would only strengthen if these methodological issues were accounted for. This, of course, remains to be empirically tested.

We also acknowledge two limitations related to the measure of identity overlap. First, as outlined in the methods section, in line with Roccas and Brewer (2002), we asked participants to gauge the overlap between each pair of chosen identities. However, in order to keep the survey as short as possible and minimize the risk of participant fatigue, we chose to instruct participants only to estimate the overlap between, for example, Groups 1 and 2 and not vice versa. As such, this might only tell part of the story if the degree of overlap is dependent on which group is compared to the other. That is, most (religious) Italians might be Catholic, but most Catholics are not Italian. Again, exploring the details of individual identities and their combinations is something for future research. However, as a broad starting point, these studies suggest value in pursuing these questions further.

Another limitation relates to the correlational method employed that limits our ability to make statements about the causal directions of the observed relationships. Thus, while we have defined particular variables as antecedents, mediators, and outcomes on a theoretical basis, there is no empirical evidence of causality in the relationships found. Future study should therefore attempt experimental longitudinal research designs to tease out causal connections between multiple group membership and well-being variables. This could be done, for example, by replicating and elaborating on the experimental design of Jones and Jetten (2011), which activated different numbers of groups before assessing physical resilience. Adapting this paradigm, future studies could manipulate the number and meaning of salient identities to explore their combined effects on indicators of well-being.

Finally, it might be productive for future research to move beyond broad measures of well-being and to delve more deeply into the processes behind those feelings. In that regard, the use of more objective outcome variables, such as for example physiological measures of well-being (e.g., indexed by stress responses and resilience, see Blascovich et al., 2011), might be useful to supplement findings based on self-report scales. For example, past research has indexed physiological arousal in terms of the extent to which the participant perceives something as a threat or a challenge, and linked such responses to mental wellbeing (e.g., loneliness, Cacioppo et al., 2002; Hawkley et al., 2003) and self-esteem (Seery et al., 2004). Indeed, there is some evidence that these kinds of physiological reactions are causally influenced by multiple group memberships (Jones and Jetten, 2011).

Other questions arising from our research concern the connection between the three central identity features discussed above. For example, how do SIC and identity compatibility relate to one another? The somewhat paradoxical notion that multiple overlapping identities (generally a bad thing) are likely to also be compatible (generally a good thing) whereas many disparate identities are more likely to give rise to incompatibility, seems theoretically reasonable. Exactly how the descriptive features of groups relates to the subjective experience of group membership is, however, likely to be complex – especially when considering counter examples of highly overlapping, but nonetheless incompatible identities, including for example, women in the workforce. That is, there are many women who have careers, and these two categories (women and workers) thus overlap considerably. However, they are also perceived as somewhat incompatible (Burgess and Borgida, 1999; Schneider and Northcraft, 1999; Fiske et al., 2002). In light of these apparent contradictions, it seems pertinent to look at exactly how different, but potentially related, identity structures interact and affect individual well-being, and specifically to attend to the difference between what is (i.e., who factually belongs to this group and that group) and what should be (i.e., whether belonging to these two groups is socially permitted and subjectively experienced as such).

# CONCLUSION

The results reported here demonstrate that the relationship between multiple group memberships and well-being is neither straightforward nor linear. Instead, we have presented evidence showing that the quantity of identities available to a person does

predict well-being, but that this link is dependent on the specific features of identities being combined. When the individuals' most important group memberships signal distinctiveness, membership in multiple groups seems to contribute positively to well-being. When the individual's most important group memberships are overlaid with concerns around stigma, membership in multiple groups might instead detract from sources of well-being. Interestingly, and despite these divergent patterns, the pathways between multiple group memberships and well-being were consistently found to be mediated through individual perceived identity expression and access to social support, as well as related processes like social inclusion. In some ways, these processes underpinning the relationship between multiple group membership and well-being might be the most important findings of this study. First, they link the individual perception of the categories to which one belongs (i.e., identity distinctiveness) with the outward enactment of the associated identities (perceived identity expression and social support), and consequent benefits (well-being). Second, they highlight the importance of the socially anchored meaning of group memberships (stigmatization and its visibility) to the way in which we are able to present ourselves socially (expression), engage with others productively (social support), and integrate multiple identities into a coherent and supportive self-concept (identity compatibility). Identifying these processes opens up the possibility of a deeper understanding of how, and under which conditions, multiple group memberships support and protect vs. undermine and fragment the self, and of the costs and benefits this may have for personal well-being.

# REFERENCES


# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Economic and Social Research Council and the Psychology Ethics Committee of the University of Exeter with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Psychology Ethics Committee of the University of Exeter.

# DATA AVAILABILITY

The research materials supporting this publication can be accessed by contacting Dr. Anders Larrabee Sønderlund at sonderlund@gmail.com.

# AUTHOR CONTRIBUTIONS

AS, TM, and MR all made substantial contributions to the conception, design, analysis, and interpretation of the work. The work was initially drafted by AS and then critically revised for important intellectual content by TM and MR. AS, TM, and MR all approved the version of the manuscript to be published, and all three authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.


inclusion and performance. Neth. J. Psychol. 62, 51–57. doi: 10.1007/bf0306 1051


contact, distinctiveness threat, and outgroup attitudes. Pers. Soc. Psychol. Bull. 35, 1085–1098. doi: 10.1177/0146167209337037


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer TK declared a shared affiliation, though no other collaboration, with one the authors MR to the handling Editor, who ensured that the process nevertheless met the standards of a fair and objective review.

Copyright © 2017 Sønderlund, Morton and Ryan. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Women (Do Not) Belong Here: Gender-Work Identity Conflict among Female Police Officers

#### Jenny Veldman\*, Loes Meeussen, Colette Van Laar and Karen Phalet

Center for Social and Cultural Psychology, Department of Psychology, KU Leuven, Leuven, Belgium

The current paper examines antecedents and consequences of perceiving conflict between gender and work identities in male-dominated professions. In a study among 657 employees working in 85 teams in the police force, we investigated the effect of being different from team members in terms of gender on employees' perception that their team members see their gender identity as conflicting with their work identity. As expected in the police force as a male-dominated field, the results showed that gender-dissimilarity in the team was related to perceived gender-work identity conflict for women, and not for men. In turn, perceiving gender-work identity conflict was related to lower team identification for men and women. Although lowering team identification might enable employees to cope with conflicting social identities and hence protect the self, this may also have its costs, as lower team identification predicted higher turnover intentions, more burn-out symptoms, less extra role behavior, lower job satisfaction, lower work motivation, and lower perceived performance. Additionally, for women, experiencing support from their team members and team leader showed a trend to mitigate the relationship between gender-dissimilarity and perceived gender-work identity conflict, and a positive diversity climate was marginally related to less perceived gender-work identity conflict. The results show the importance of the team context in shaping a climate of (in)compatible identities for numerically underrepresented and historically undervalued social group members in order to hinder or protect their work outcomes.

#### Edited by:

Clara Kulich, University of Geneva, Switzerland

#### Reviewed by:

Catherine Verniers, Université Paris Descartes, France Fabrice Gabarrot, Université de Bourgogne Franche Comté, France

> \*Correspondence: Jenny Veldman jenny.veldman@kuleuven.be

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

> Received: 27 October 2016 Accepted: 18 January 2017 Published: 06 February 2017

#### Citation:

Veldman J, Meeussen L, Van Laar C and Phalet K (2017) Women (Do Not) Belong Here: Gender-Work Identity Conflict among Female Police Officers. Front. Psychol. 8:130. doi: 10.3389/fpsyg.2017.00130 Keywords: conflicting identities, gender diversity, team identification, support, diversity climate, well-being, work motivation

Women (do not) belong here: Gender-work identity conflict among female police officers

"Although I generally feel good within our team, the feeling that as a woman you have to prove yourself even more to get the same appreciation as a police detective prevails."

Female police officer participant

Nowadays, people often work in teams and teams shape people's work experiences (Jackson, 1996), such as the extent to which they feel valued in their work team. Being a member of a social group that is a numerical minority can lead people to feel less valued in their work team or in the organization as a whole because of their social identity (Inzlicht and Good, 2006). The quote above from a female participant illustrates this can indeed be the case for women working in an

organization such as the police force, which has a relatively short history of female employees and where women are still strongly underrepresented worldwide (Bureau of Justice Statistics, 2010; Statistics Belgium, 2010; Europol, 2013). She feels less valued in her team due to her gender and feels like she has to prove herself even more than her male colleagues to be equally valued. Put differently, she feels that her team members believe her gender to conflict with her work.

In the present research, we argue that the team context plays an important role in shaping employees' perceptions of such gender-work identity conflict. We take an individual-within-theteam perspective on gender diversity and argue that differing from team members in terms of gender in this masculine environment affects women's and not men's perception of gender-work identity conflict. Additionally, we argue that employees who experience that their team members see their gender and work identities as conflicting can cope with this by attaching the self psychologically less to the team. Although reducing their team identification might enable employees to cope with conflicting identities and hence protect the self, this may have its costs for important work outcomes related to wellbeing, motivation, and performance at work (see, e.g., Ellemers et al., 1998; Ouwerkerk et al., 1999; Van Laar and Derks, 2003). Experiencing support from team members and from the team leader, and perceiving a positive diversity climate are examined as team contextual supportive factors that can buffer identity conflict for women.

# Gender-Dissimilarity in the Team and Gender-Work Identity Conflict

Work organizations and teams are increasingly diverse with also traditionally underrepresented groups finding their way into work fields in which they were less represented (Ely and Thomas, 2001). Research has primarily focused on effects of diversity at the group level, for example examining whether more diverse teams are more innovative and effective than less diverse teams (e.g., McLeod et al., 1996; Horwitz and Horwitz, 2007; Van Dijk et al., 2012). Other research has taken an individual-within-the-group perspective on diversity and how this affects individual outcomes (Tsui et al., 1992; Liao et al., 2004). This is important as individuals often struggle with being different from their team members, for example from genderdissimilarity (Guillaume et al., 2012). Gender-dissimilarity is the difference between a focal group member and his or her group members with respect to gender. It could also be seen as a reflection of how prototypical a group member is within a group in terms of his or her gender (Oakes et al., 1998). In the current research we took such an individual-withinthe-team perspective on gender diversity and examined how this related to employees' meta-perception of how their team members view their gender and work identities (Frey and Tropp, 2006). We investigated whether being dissimilar from team members in terms of gender was related to employees perceiving more that their team members see their gender as conflicting with the work they do (i.e., perceived gender-work identity conflict).

We build on several theoretical frameworks for this research question. First, following the social identity approach (Tajfel and Turner, 1979; Turner et al., 1987), people categorize themselves and others into in- and out-groups based on observable similarities and differences (such as gender). Being dissimilar from others makes this category more salient (Wilder, 1984) and increases individuals' tendency to expect that others will view them in terms of their group membership (Frey and Tropp, 2006). Thus, being dissimilar from other team members in terms of gender makes employees more aware of their gender and increases their expectation that they are viewed by other team members in terms of their gender. Adding to relational demography literature which is not conclusive on whether being dissimilar is more consequential for certain groups (see, e.g., Chattopadhyay et al., 2015), we argue that dissimilarity negatively impacts experiences in a team only when this dissimilarity is based on a social identity that is stigmatized within the given context. Indeed, the social identity that is made salient when one is dissimilar in a team is not neutral: being a woman in a traditionally male context tends to be associated with lower value to that identity (Branscombe and Ellemers, 1998; Inzlicht and Good, 2006; Van Laar et al., 2010). Women were not allowed into most European police forces until a few decades ago, and still are in a vast numerical minority. Also, in this setting masculine characteristics are strongly valued (Somvadee and Morash, 2008; Archbold et al., 2010). Attributes typically associated with being female are seen as not fitting or as incongruent with the attributes associated with being a police officer (Heilman, 1983; Eagly and Karau, 2002; Lyness and Heilman, 2006). That is, the prototypical police officer does not match the prototypical woman. This creates a perceived incongruity or lack of fit between being a woman and the work identity of a police officer. Related to this, negative stereotypes about women in the police force prevail (Somvadee and Morash, 2008; Archbold et al., 2010). Thus, being dissimilar from team members in terms of gender makes one's gender salient and increases the expectation that other team members view you in terms of your gender. For women in the police force it also makes salient a social identity that is associated with negative stereotypes and lower value, and that is seen as incongruent with working in the police force. Therefore, we hypothesize that being gender-dissimilar from the team relates to a stronger perception that their team members see their gender as conflicting with their work identity for female police officers. Because of the masculine history and values of the organizational context, we expect that male officers will not have higher perceived gender-work identity conflict even when their gender differs more from that of their team members.

Hypothesis 1: Gender-dissimilarity is positively related to perceived gender-work identity conflict for female, but not male police officers.

# Coping with Gender-Work Identity Conflict and Work-Related Costs

People are not passive recipients of their environment when they experience that their social identity is devalued in a context, but they try to cope with it Van Laar et al. (2010). While

employees might have relatively little influence on their team members' perception of the conflict between their gender and work identities, they can cope with the conflicting identities by reducing their ties with the team. More specifically, perceived gender-work identity conflict is a cognitive concept that implies that the prototype of your gender group is not compatible with the prototype of your work identity (e.g., the prototypical police officer does not match the prototypical woman), hence leading to more marginal team membership (Oakes et al., 1998; Ellemers and Jetten, 2013). Level of team identification, on the other hand, refers to the affective ties of an individual to the team, which is more under an individual's own control (Ellemers et al., 1999; Leach et al., 2008). Researchers have primarily examined group contexts in which more marginal (i.e., less valued) group members are motivated to become more prototypical and core group members (e.g., Levine and Moreland, 1994). However, this moving toward the group is not always the preferred trajectory (Ellemers and Jetten, 2013). In an environment in which the team communicates that you are a less valued team member as they perceive your gender and work identities as conflicting, attaching the self psychologically less to the group can be an important route to self-actualization, identity, and value (Ellemers and Jetten, 2013). Therefore, we hypothesize that employees will identify less strongly with their work team the more they perceive that their team members see their gender and work identities as conflicting. As outlined above, we expect women, but not men to experience more perceived gender-work identity conflict in face of gender-dissimilarity. However, to the extent that men do experience this conflict, we expect that they too will identify less with the team. In other words, we expected both male and female employees to endorse this coping strategy in the face of perceived gender-work identity conflict.

Hypothesis 2: Perceived gender-work identity conflict is negatively related to team identification.

While reducing identification with the team when perceiving more gender-work identity conflict among team members might enable employees to better cope with conflicting social identities, this has costs for their outcomes at work. Previous research has shown that members who identify less strongly with their group are less willing to contribute to collective goals (Ellemers et al., 1998; Tyler and Blader, 2000; Meeussen and Van Dijk, 2016), are less productive (Worchel et al., 1998; Meeussen et al., 2014a), and are less willing to put effort in the group beyond what is expected of them (Ouwerkerk et al., 1999; Van Knippenberg, 2000). We examined not only work motivation and performance at work, but also work-related well-being. While individuals can identify less with the team in response to gender-work identity conflict to protect the self (or more general well-being), this may come at the cost of less well-being at work. We hypothesize that lower team identification is linked to important work-related outcomes: lower work motivation, lower work satisfaction, lower perceived performance, less extra role behavior, more burn-out symptoms, and higher turnover intentions at work.

Hypothesis 3: Team identification is positively related to work motivation, work satisfaction, perceived performance, and extra role behavior, and negatively related to burn-out symptoms and turnover intentions.

Additionally, we examined the indirect relationship between perceived gender-work identity conflict and the work-related outcomes via lower team identification. We expect that a climate in which employees experience that their team members see their gender as conflicting with their work identity does not always directly translate into lower outcomes. Employees have different possibilities to respond in such a climate – for instance, increasing efforts at work to challenge the idea that their gender group may not fit with this work (Alter et al., 2010) or distancing the self from one's gender group (Derks et al., 2011a). The current research examines lowering identification with the team as an important, but understudied, coping mechanism (Ellemers and Jetten, 2013). We expect that, to the extent that officers indeed do this, gender-work identity conflict will relate to lower work outcomes.

Hypothesis 4: Perceived gender-work identity conflict is indirectly negatively related to the work-related outcomes via lower team identification.

# Team Contextual Supportive Factors Buffering Gender-Work Identity Conflict

Lastly, we argue that supportive factors in the team context can reduce the negative impact of being gender-dissimilar from other team members for female police officers. First, experienced social support from team members and the team leader is argued to be important. Social support can include communication of emotional concern or comfort, affirmation, being able to turn to others for guidance and assistance when needed, and the provision of information (Wills, 1985; London et al., 2011; Richman et al., 2011). Such social support can reduce the impact of a stressor and the perceived stressfulness of an event or experience, hence offering a "stress-buffering" effect (Lazarus and Folkman, 1984; Cohen and Wills, 1985). Research on members of negatively stereotyped groups has shown that experiencing social support, or perceiving that it is available, is related to better achievement outcomes and engagement in work and education (Eccles, 1994; Walton and Cohen, 2007; Hartman and Hartman, 2008; Richman et al., 2011; Baysu et al., 2014). Another important team contextual supportive factor is a positive diversity climate. A positive diversity climate is the extent to which minority groups perceive the environment to be open toward their social group (Purdie-Vaughns et al., 2008; Gonzalez and Denisi, 2009; Plaut, 2010). Research has shown that perceiving a positive diversity climate is related to, amongst others, organizational commitment, organizational identification, feeling included, and lower turnover intentions for members of stigmatized groups (e.g., Gonzalez and Denisi, 2009; Meeussen et al., 2014b). Because of these key positive effects of experienced support and a perceived positive diversity climate for stigmatized group members' outcomes in work and education, these team contextual supportive factors are examined as possible buffers for gender-work identity conflict for women in the police force context. We expect that the effect of genderdissimilarity on perceptions of gender-work identity conflict for women is reduced when they experience support from their team members, when they experience support from their team

leader, and when they perceive a positive diversity climate in their team. Given that we do not expect men to experience identity conflict when they are gender-dissimilar in their team (as their gender identity is not devalued within this context), we also do not expect contextual supportive factors to moderate this.

Hypothesis 5: Team contextual supportive factors (experienced support from team members, experienced support from the team leader, and perceived positive diversity climate) reduce the relationship between gender-dissimilarity and perceived genderwork identity conflict for female officers.

These hypotheses were tested in a large cross-sectional study among men and women working at the police force in a Western European country.

# MATERIALS AND METHODS

# Participants and Procedure

The sample consisted of 789 employees of a police force in a Western European country.<sup>1</sup> As the current study focuses on effects of gender-dissimilarity in teams and on team leader support, we excluded 132 participants who were team leaders themselves, resulting in a final sample of 657.<sup>2</sup> The mean age of the participants was 43 years old (SD = 9.72, range: 21–64) and 60% were male (38% female, 2% unknown). For 54% of participants secondary education was their highest level of education, and 43% had a college or university degree (3% unknown). On average, participants had been working in the police force for 17 years (SD = 11.46), within their current department for 10 years (SD = 9.02), and in their current position for 9 years (SD = 8.12). Participants either had an executive position (55%) or a logistics or administrative position (42%; 3% unknown).

After the director of the organization had given consent for this research on "Diversity in the workplace," team leaders of 122 teams distributed across all regions of the country were invited for participation. 15% (n = 18) of the team leaders could not be reached and 2% (n = 2) declined. Of the remaining 102 teams, 84% (n = 85) participated. The average size of a team was 12 people (SD = 6.81) and ranged from 3 to 33 people plus their leaders.<sup>3</sup> Team leaders received and distributed the questionnaires among their members. Employees who consented to participate completed the survey individually during working hours, which took approximately 30 min. After completing the survey, employees returned their completed survey in a sealed envelope to their team leaders. Team leaders collected the envelopes from all team members and mailed them back to the researchers. The survey consisted of a wide range of questions on perceptions of diversity. The measures relevant for the current manuscript are detailed below. **Table 1** presents an overview of correlations between all variables.

## Measures

Unless otherwise indicated, items were answered on a five-point Likert scale from (1) strongly disagree to (5) strongly agree. Measures are scored such that higher scores indicate stronger scores on the concept.

#### Gender-Dissimilarity

Gender-dissimilarity is the difference between a focal team member and his or her team members with respect to gender (Guillaume et al., 2012). Gender-dissimilarity was measured by calculating the Euclidean Distance between each respondent and his or her other team members (see for recommendations Harrison and Klein, 2007). For each individual team member the Euclidean distance was calculated by dividing the number of group members with a different gender by group size and then taking the square root of this fraction (Tsui et al., 1992; see also Jansen et al., 2016). For example, in a team with three men and two women, the Euclidean distance for the men equals <sup>√</sup> (2/5) <sup>=</sup> 0.63 and for women equals <sup>√</sup> (3/5) = 0.77. The Euclidean distance can range from 0 (all team members have the same gender as the focal team member) to nearly 1 (all team members have a different gender group than the focal team member). Women had a higher gender-dissimilarity compared to men, but the range and standard deviation were similar. For men, the mean Euclidean distance was 0.42 (SD = 0.24, range: 0.00–0.94); for women, 0.62 (SD = 0.21, range: 0.00–0.96).

#### Team Members' Support

Experienced support from other team members was measured with "In my team, I can count on my colleagues when I experience difficulties at work" (M = 3.74, SD = 1.02; adapted from Van Veldhoven et al., 2002).

#### Team Leader Support

Experienced support from the team leader was measured with "In my team, I can count on my team leader when I experience difficulties at work" (M = 3.43, SD = 1.26) (adapted from Van Veldhoven et al., 2002).

#### Positive Diversity Climate

Positive diversity climate was operationalized in the current research as the perceived openness of the team to differences (items based on Luijters et al., 2008; Nakui et al., 2011). This was measured with eight items: e.g., "The ideas of colleagues who differ from each other complement each other" and

<sup>1</sup>Due to agreements with the organization about anonymity, we do not state in which country this research was conducted.

<sup>2</sup>We only included participants in the analyses of whom we could be sure that they did not have a leadership position in the team. Because of anonymity concerns, we could not ask the team leader to identify him/herself directly. We therefore identified the team leader using a bottom–up procedure consisting of several steps. Participants were asked whether they had a leadership position; if only one member of the team indicated they did, then this person was marked as the leader of the team. If two people indicated they did, then this mostly implied that one of them was the leader of the work team and the other had a leadership position higher up in the organization (which could be concluded by comparing the indicated number of people they led and its match with the number of participants from that team, and looking at the indicated number of levels and departments they led). They were both excluded from the current sample.

<sup>3</sup>Team size estimates are based on the number of completed questionnaires in a team. This could be a slight underestimation for some teams, in case a team member was absent during the week of data collection.

#### TABLE 1 | Correlations between all measures.

fpsyg-08-00130 February 4, 2017 Time: 11:3 # 5


<sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; −1 = male, 1 = female.

"The differences between team members make this a valuable collaboration" (α = 0.70, M = 3.22, SD = 0.65).

#### Gender-Work Identity Conflict

The perception of gender-work identity conflict among team members was measured with "To what extent do you think that other colleagues consider your function as a police officer to be compatible with your gender?" [(1) not at all to (5) very much; M = 3.87, SD = 1.25]. The responses were reverse coded so that higher scores indicate more perceived gender-work identity conflict.

#### Team Identification

Participants' identification with their team was measured with seven items (taken from Ellemers et al., 1999; Roccas et al., 2008). Example items are "I identify with the other members of my team" and "I feel strongly affiliated with this team" (α = 0.73, M = 3.51, SD = 0.66).

#### Work-Related Outcomes

Six measures assessed participants' outcomes at work. First, work motivation was measured with the overall item "How motivated are you for your job?" [(1) not at all to (5) very much; M = 3.85, SD = 0.93]. Additionally, we measured extra role behavior with "I volunteer for things at the police force without being asked to" (M = 3.63, SD = 1.03, adapted from Van Veldhoven et al., 2002), and turnover intentions with two items: e.g., "I sometimes think of searching for work outside the police force" (r = 0.50, M = 2.26, SD = 1.20; adapted from Van Veldhoven et al., 2002). We measured perceived performance with two items: e.g., "How well did you execute the responsibilities of your job in the past month?" (r = 0.71, M = 4.18, SD = 0.67; based on Abramis, 1994). We measured employees' job satisfaction with "Overall, how satisfied are you with your job?" [(1) not at all to (5) very much; M = 3.79, SD = 0.85] (based on the validated single-item measure developed by Dolbier et al., 2005; for a meta-analysis validating the use of single-item job satisfaction measures see also Wanous et al., 1997). Lastly, we measured employees' burnout symptoms with three items (α = 0.83, M = 2.32, SD = 0.99): e.g., "I feel empty at the end of the day" (adapted from Van Veldhoven et al., 2002).

# RESULTS

The data were analyzed using two-level multilevel analyses with a random intercept model in Mplus 5 (Muthén and Muthén, 1998). While all variables are individual-level measures, multilevel analyses allow us to control for the nested structure of data with employees nested in the 85 teams. We used maximum likelihood (ML) estimations, preferable when using 30 groups or more (Browne and Draper, 2000; Maas and Hox, 2004). Continuous independent variables were grand-mean centered. Standardized estimated effects are reported. We estimated a multilevel structural equation model to test the hypotheses that gender-dissimilarity predicts perceived gender-work identity conflict for female officers (Hypothesis 1), that perceived gender-work identity conflict predicts lower identification with the team (Hypothesis 2), that team identification predicts work-related outcomes (Hypothesis 3), and that perceived gender-work identity conflict is indirectly related to lower work-related outcomes via lower team identification (Hypothesis 4). This model is depicted in **Figure 1** and showed good model fit (CFI = 0.96, TLI = 0.91, RMSEA = 0.06, SRMR = 0.03, Bentler, 1990).

First, we looked into the relation between gender-dissimilarity and perceived gender-work identity conflict. Results showed a marginally significant main effect of gender-dissimilarity (β = 0.09, p = 0.06) on perceived gender-work identity conflict, such that employees who differed more from their team members

in terms of gender perceived their team members to see more conflict between their gender and work identities. This marginal main effect was qualified by an interaction with gender: in line with Hypothesis 1, there was a significant interaction effect between gender and gender-dissimilarity on perceived genderwork identity conflict, β = 0.09, p = 0.03. For female police officers, gender-dissimilarity was related to higher perceived gender-work identity conflict (β = 0.17, p = 0.01), while for male police officers their gender-dissimilarity was unrelated to perceived gender-work identity conflict (β = −0.01, p = 0.83).<sup>4</sup> There was no main effect for gender (β = 0.01, p = 0.89).

Confirming Hypothesis 2, the results showed that the more employees perceived that their team members see their gender and work identity as conflicting, the less they identified with their team, β = −0.10, p = 0.01.<sup>5</sup> In turn, as expected in Hypothesis 3, team identification was significantly related to the six work-related outcomes: identifying with the work team related to lower turnover intentions (β = −0.35, p < 0.001), fewer burn-out symptoms (β = −0.27, p < 0.001), more extra role behavior (β = 0.23, p < 0.001), higher job satisfaction (β = 0.51, p < 0.001), higher work motivation (β = 0.54, p < 0.001), and higher perceived performance (β = 0.29, p < 0.001). Consistent with Hypothesis 4, perceived gender-work identity conflict was indirectly (and not directly, all ps > 0.16) related to work-related outcomes via team identification, and these effects were consistent and significant across the different outcome measures (turnover intentions: β = 0.04, p = 0.02; burn-out symptoms: β = 0.03, p = 0.02; extra role behavior: β = −0.02, p = 0.02; job satisfaction: β = −0.05, p = 0.01; work motivation: β = −0.05, p = 0.01; perceived performance: β = −0.03, p = 0.02).

Combined, the results show that being different from team members in terms of gender related to a stronger perception that their team members see their gender as conflicting with their work identity for female, but not male police officers. This perceived gender-work identity conflict was related to lower team identification, which in turn related to higher turnover intentions, more burn-out symptoms, less extra role behavior, lower job satisfaction, lower work motivation, and lower perceived performance.

Investigating Hypothesis 5 that team contextual supportive factors (experienced team members' support, experienced team leader support, and perceived positive diversity climate) reduce the relationship between gender-dissimilarity and perceived gender-work identity conflict for female officers, we also added a three-way interaction between the team contextual supportive factors, participants' gender, and gender-dissimilarity in the model. These interactions were not significant (Experienced team members' support: β = −0.07, p = 0.12; Experienced team leader support: β = −0.07, p = 0.16; Positive diversity climate: β = −0.07, p = 0.17). Still, given that our results showed, as expected, an interaction between participants' gender and gender-dissimilarity on perceived gender-work identity conflict, indicating only an effect for women, we performed tentative additional analyses looking into the role of contextual support factors for women only, as they are the ones who experience identity conflict when they are dissimilar (see **Table 2**).

<sup>4</sup>The effect of gender-dissimilarity on perceived gender-work identity conflict for female officers remained when controlling for their position (executive vs. administrative), β = 0.23, p = 0.002. For male officers the relationship remained non-significant, β = −0.01, p = 0.87. Additionally, the interaction between genderdissimilarity and gender also remained after controlling for participants' position (executive vs. administrative), β = 0.12, p = 0.016.

<sup>5</sup>Consistent with expectations, this relationship was not moderated by gender, β = −0.04, p = 0.64.


TABLE 2 | Standardized estimated effects (Maximum Likelihood) and standard errors on perceived gender-work identity conflict for female police officers.

†p < 0.10; <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

We first examined whether experienced support from team members could mitigate the negative effect of genderdissimilarity on perceived gender-work identity conflict for women. There was a marginally significant negative relationship indicating that female police officers who experienced support from team members reported lower perceived gender-work identity conflict, β = −0.12, p = 0.090). This main effect was qualified by a marginally significant interaction between gender-dissimilarity and experienced team members' support, β = −0.12, p = 0.082. An inspection of the simple slopes (see **Figure 2**) revealed that gender-dissimilarity was only related to higher perceived gender-work identity conflict when women experienced low support from their team members (p < 0.001), and not when experiencing high support from team members (p = 0.32). Thus, gender-dissimilarity was not related to perceived gender-work identity conflict anymore when experiencing high support from team members.

Second, there was no relationship between experienced team leader support and perceived gender-work identity conflict. However, there was a marginally significant interaction between gender-dissimilarity and experienced team leader support, β = −0.13, p = 0.095. Again, an inspection of the simple slopes (see **Figure 3**) revealed that gender-dissimilarity was related to higher gender-work identity conflict when women experienced low support from their team leader (p = 0.002), and that genderdissimilarity was not related to perceived gender-work identity conflict anymore when experiencing high support from the team leader (p = 0.56).

There was no significant interaction between gender-work identity conflict and positive diversity climate. There was, however, a marginal negative relationship between positive diversity climate and perceived gender-work identity conflict indicating that female police officers perceived less gender-work identity conflict when perceiving a positive diversity climate in the team, independent of their gender-dissimilarity in the team, β = −0.14, p = 0.079.<sup>6</sup>

<sup>6</sup>For male officers the relationship between gender-dissimilarity and perceived gender-work identity conflict did not depend on experienced team members' support (β = 0.02, p = 0.68), experienced team leader support (β = 0.01, p = 0.86), and positive diversity climate (β = 0.05, p = 0.34). Additionally, for male police officers experienced team members' support (β = −0.09, p = 0.12), experienced team leader support (β = −0.06, p = 0.26), and positive diversity climate (β = −0.03, p = 0.61) were not significantly related to perceived gender-work identity conflict.

Thus, Hypothesis 5 could not be confirmed since there were no significant three-way interactions between participants' gender, gender-dissimilarity and the team contextual supportive factors. Yet, our additional analyses do suggest that there may be potential in team contextual support factors for women, who are most vulnerable to identity conflict in teams where they are dissimilar. Results indicated that there was a trend for support from team members and team leaders to buffer perceived genderwork identity conflict when they are different from their team members in terms of gender.

# DISCUSSION

The present study demonstrated the importance of dissimilarity from one's work-team in shaping perceptions of conflicting identities for numerically underrepresented and historically undervalued social group members. This was shown among work teams in the police force in a Western European country – a setting where women are a numerical minority overall, where masculine characteristics are strongly valued, and where negative stereotypes about women prevail (Somvadee and Morash, 2008; Archbold et al., 2010). Additionally, we showed that perceiving conflicting identities in a team was related to lower identification with the team, which in turn related to lower important workrelated outcomes.

# Gender-Dissimilarity Is Related to Gender-Work Identity Conflict for Female Police Officers

We demonstrated that for female police officers, being different from team members in terms of gender was related to perceiving more that their team members see their gender as conflicting with their work identity. The advantage of the current method in which we took an individual-within-the-team perspective on diversity is that we did not have to make a clear-cut distinction between majority and minority members, but could look for each individual how dissimilar they were compared to their team members (Jansen et al., 2016). So, although women are in a minority in the police force overall, and previous research has shown that compared to male officers female officers feel less valued in the police force (e.g., Archbold et al., 2010), the current research showed that this does not necessarily represent the experience of all women in the police force. Indeed, our data suggest that when they worked in more gender-balanced teams women perceived that their team members saw their gender as less conflicting with their work identity. Put differently, they did not feel less valued in their work team because of their gender.

For male police officers we did not find a relationship between gender-dissimilarity and perceived gender-work identity conflict. Previous research investigating gender differences in effects of gender-dissimilarity in teams or organizations on work outcomes has been inconclusive (Chattopadhyay et al., 2015), showing sometimes that effects may be stronger for women (e.g., Gonzalez and Denisi, 2009), and other times that gender-dissimilarity may actually be more consequential for men, who tend to be in the majority in work environments and are less used to genderdissimilarity (e.g., Tsui et al., 1992), or that there is no difference in the effects for men and women (Jansen et al., 2016). Advancing relational demography literature, the current findings indicate that whether gender-dissimilarity negatively impacts experiences in a team depends not only on being dissimilar, but also on what is communicated about this social identity. That is, being dissimilar from other team members in terms of one's social identity makes the social identity salient (Tajfel and Turner, 1979; Turner et al., 1987), and the value related to this salient identity then determines whether dissimilarity has negative effects: in

traditionally male-dominated organizations like the police force (Somvadee and Morash, 2008; Archbold et al., 2010) dissimilarity makes an undervalued and negatively stereotyped social identity salient for women but not for men. The current research context provided an excellent opportunity to investigate this, as it allowed us to examine relationships in a masculine work environment in which women are both a numerical minority overall and in which negative stereotypes about women prevail (Somvadee and Morash, 2008; Archbold et al., 2010), but also one in which the range and variance in gender-dissimilarity was similar for male and female police officers. The reason this was the case was that employees with an administrative (vs. an executive) position were relatively more strongly represented in the current sample (42%) compared to the population of the police force (27%). Thus, our sample was not entirely representative for the police force overall, but this overrepresentation of administrative teams provided a more conservative test of our hypotheses where women are more represented and less negatively stereotyped. Future research could examine whether the current findings replicate in feminine contexts (e.g., nursing or teaching) where men are underrepresented. Do men in such contexts experience the same identity conflict in teams in which they are genderdissimilar, and does it have similar consequences? In line with our interpretation of the findings in the current research, we would expect that this depends on whether the male identity is undervalued and negatively stereotyped in that organizational context.

# Coping with Gender-Work Identity Conflict and Work-Related Costs

To cope with experiences of conflict between identities, attaching the self psychologically less to the group can be used to protect self-identity and value (Ellemers and Jetten, 2013). In line with this, the present study found that employees identify less strongly with the team they work in when they perceive their team members to see their gender and work identities as more conflicting. Researchers have primarily examined group contexts in which less valued group members are motivated to become more prototypical and core group members (e.g., Levine and Moreland, 1994). However, moving toward the groups is not always the preferred trajectory (Ellemers and Jetten, 2013). Until recently, psychologists assumed that the need for belonging is better satisfied by being more typical in a group, as a core position is associated with greater acceptance and inclusion than a more marginal position (Baumeister and Leary, 1995). However, people are likely to hold multiple group memberships simultaneously that can satisfy their need for belonging (Ellemers and Jetten, 2013), and can hence identify less with certain groups they are part of to protect their self-identity and other social identities.

While this lower group identification might enable individuals to cope with conflicting social identities, this had its costs, since lower team identification related to lower work satisfaction, lower perceived performance at work, lower work motivation, less extra role behavior, higher turnover intentions, and more burnout symptoms at work. This also implies that individuals may eventually start searching for work teams or occupations in which they feel more valued as a team member. This is consistent with a view of resilience in individuals to social identity threats. That is, individuals are quite able to cope with identity threat, finding ways to maintain general well-being, but this coping also has indirect and often hidden costs for the self or group (Derks et al., 2009, 2015; Van Laar et al., 2010; Ståhl et al., 2012), as evidenced also by the findings presented here.

In line with the view that people can be resilient when faced with identity threat, we found that perceived genderwork identity conflict was indirectly (via team identification), but not directly related to the work outcomes. Thus, only to the extent that individuals identify less with their team as a way of coping with identity conflict this is related to lower work-related outcomes. This shows that a climate in which employees experience that team members see conflict between one's gender and work identities does not necessarily directly translate into lower outcomes, but that people can cope with this by lowering their identification with the team, which translates into lower work outcomes. In real life, individuals may use other strategies as well to cope with the experience of conflicting identities. For example, trying to convince the other team members that they should not consider their gender to be conflicting with their work in the team by emphasizing their masculinity or actively distancing themselves from other women (e.g., by emphasizing that they are different from other women at work, Derks et al., 2011a,b, 2016). In fact, Derks et al. (2011b) indeed found that women in the police force show this when reminded of their devalued status. So, individuals cannot only cope with perceived conflict between gender and work identities by distancing themselves from the work team to increase identity fit, but they can also distance from their gender group or try to resemble the other gender group more. Additionally, people could also reduce identification with their gender group. There are, however, reasons to believe that this is not very likely to happen. Gender is inherited and one of the most salient categories used to categorize oneself and others into in- and outgroups (particularly within organizational contexts, Hogg and Terry, 2000), and gender especially becomes salient when being gender-dissimilar from others (Tajfel and Turner, 1979; Turner et al., 1987). This makes it difficult to actually discard the identity when the team context makes this devalued social identity (in the current masculine context) salient, and when others repeatedly address one consistent with that identity (Branscombe and Ellemers, 1998). In line with this, recent research indeed found that underrepresented groups in outgroup domains identified more strongly with the outgroup and showed reduced concern for the ingroup, but did not identify less with the ingroup (Kulich et al., 2015). Investigating moderators influencing which coping mechanism is used when could be an interesting avenue for future research. For example, the current research focused on employees, showing that they can cope with conflicting identities by reducing their identification with the team. Previous research on self-group distancing as an identity management strategy, however, focused on employees with a leadership position (e.g., Derks et al., 2011a,b) and

they might be less inclined to distance themselves from their achieved high status professional position. The status of one's professional position could thus be an interesting moderator triggering different coping mechanisms. Additionally, future research could provide more insight in the differential (protective and harmful) effects of these different coping mechanisms on work outcomes. More generally, we hope that future research using longitudinal designs will be able to provide more insight in the dynamic interaction between the team context and negatively stereotyped group members' coping with conflicting identities.

# The Role of Team Contextual Supportive Factors

The current findings indicate that conflict between gender and work identities has consequences for the person who experiences this identity conflict, as well as for teams and organizations they are working in. Therefore, we also examined whether such experiences of identity conflict when being in a genderdissimilar team for female police officers could be mitigated by team contextual supportive factors. The results did not provide conclusive answers to this question. Although we did not find the expected interaction with participants' gender, additional tentative analyses among female officers, who are most vulnerable to identity conflict in gender-dissimilar teams, did suggest that there may be potential in team contextual supportive factors. There was a trend for support from team members and team leaders to buffer identity conflict when they are different from their team members in terms of gender. This raises questions for future research. Perhaps being dissimilar from team members is not so easily overcome, particularly in the case of women in a highly male-dominated and masculine organization such as the police force. Team dynamics are embedded in the wider organizational structure and climate, which might constrain the impact of the supportive climate within the team. Thus, experiencing support and perceiving a positive diversity climate in the direct work environment might not be enough in such highly masculine organizations and are to be supplemented by more structural change in the broader work environment. Additionally, in future research it might be valuable to make a distinction between more emotional (e.g., communication of emotional concerns or comfort) and more instrumental (e.g., receiving guidance and assistance, provision of information) forms of social support (Wills, 1985). Certain types of social support might be better able to provide a buffer for experienced identity conflict than others.

# Limitations

A limitation of the current research is that our data was cross-sectional, and therefore we are not able to draw causal conclusions about the relations that were examined. For example, we examined perceived gender-work identity conflict in the team predicting individual team members' identification with the team, but cannot conclude from the current study that perceiving conflicting identities causes lower team identification. Building on previous research, we know that perceiving that your team members see conflict between your gender and work identities indicates that one is a less valued team member because of one's gender, and when people do not feel valued in a team, they may attach psychologically less to the team as an important route to self-actualization, identity, and value (Ellemers and Jetten, 2013). Because of this, it seems likely that perceiving conflict between identities leads to individuals themselves reducing their ties with the team. Yet, it is possible that on top of that, other group members may perceive more conflict between women and a work identity when they perceive women to have lower identification at work – hence the two may reinforce each other over time. Similarly, in the current research we expected that team identification would predict important work-related outcomes, which is in line with previous research that longitudinally and experimentally showed that team identification leads to better work-related outcomes (e.g., Worchel et al., 1998; Meeussen and Van Dijk, 2016). Still, it is possible that lower outcomes at work also lead to lower identification with the team. Follow up research could examine these relations in experimental or longitudinal designs to investigate causality.

# CONCLUSION

The current research showed the importance of the team context in shaping numerically underrepresented and historically undervalued social group members' perception that their team members see their identities as conflicting. Findings indicated that as a woman being gender-dissimilar from team members in a masculine organizational context can trigger such perceptions of conflict between gender and work identities. This is important, because perceiving conflicting identities had negative implications for such individuals' identification with the team, and consequently for their work outcomes. Understanding, how the team context can shape a climate of compatible identities for underrepresented group members in order to protect their work outcomes is an important question for future research. The current findings provided some indications that experiencing support from team members and team leaders might be able to shape such a climate.

# ETHICS STATEMENT

Before starting the questionnaire, participants agreed to an informed consent. They were informed that the researchers were interested in their experiences with, and perceptions of diversity in the police force. They were informed that participation was voluntary and could be discontinued at any moment during the study; and that their responses were anonymous and would be treated confidentially. Moreover, they were provided with room for questions and comments as well as contact information of the researchers. Additionally, they received a full debriefing at the end of the study.

Ethical approval was not required as per the local legislation at the time of the study commencement, and no ethics committee existed yet at the university for this type of research (only for clinical trials and medical research). However, this study was setup in consultation with the police force, and the director of the organization approved the procedure of the study.

# AUTHOR CONTRIBUTIONS

fpsyg-08-00130 February 4, 2017 Time: 11:3 # 11

All authors contributed to the interpretation of the results and the writing of the manuscript. JV, LM, and CVL contributed to the development of hypotheses. LM and KP contributed to the development of the project. JV conducted the statistical analyses.

# REFERENCES


# FUNDING

This research was supported by an Odysseus grant to CVL from the Research Foundation of Flanders (FWO) grant number G.O.E66.14N and by the doctoral grants for Loes Meeussen and Ellen Delvaux to KP and Batja Mesquita.

# ACKNOWLEDGMENT

We would like to thank Ellen Delvaux and Norbert Vanbeselaere for their contributions to the development of this project.



theoretical perspectives. Pers. Soc. Psychol. Rev. 12, 280–306. doi: 10.1177/1088868308319225


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Veldman, Meeussen, Van Laar and Phalet. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Stereotype Threat and Perceptions of Family-Friendly Policies among Female Employees

#### Courtney von Hippel<sup>1</sup> \*, Elise K. Kalokerinos<sup>2</sup> and Hannes Zacher<sup>3</sup>

<sup>1</sup> School of Psychology, University of Queensland, Brisbane, QLD, Australia, <sup>2</sup> Faculty of Psychology and Educational Sciences, KU Leuven, Leuven, Belgium, <sup>3</sup> Institute of Psychology, University of Leipzig, Leipzig, Germany

In their efforts to recruit and retain female employees, organizations often attempt to make their workplaces "family-friendly." Yet there is little research on how women view family-friendly policies, particularly women who experience gender-based stereotype threat, or the concern of being viewed through the lens of gender stereotypes at work. Pilot research with female managers (N = 169) showed that women who experienced stereotype threat perceived more negative career consequences for utilizing familyfriendly policies. We then conducted two studies to further probe this relationship. Study 1 replicated the relationship between stereotype threat and the perceived consequences of utilizing family-friendly policies among women who recently returned to work after the birth of a child (N = 65). In Study 2 (N = 473), female employees who reported feelings of stereotype threat perceived more negative consequences of utilizing family-friendly policies, but they also reported greater intentions to use these policies. Our findings suggest that female employees are susceptible to stereotype threat, which in turn is associated with more negative views of family-friendly policies. Thus, the mere provision of such policies may not create the kind of family-friendly workplaces that organizations are attempting to provide.

#### Edited by:

Clara Kulich, University of Geneva, Switzerland

> Reviewed by: Loes Meeussen, KU Leuven, Belgium Thekla Morgenroth, University of Exeter, UK

## \*Correspondence:

Courtney von Hippel c.vonhippel@uq.edu.au

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

Received: 22 September 2016 Accepted: 16 December 2016 Published: 05 January 2017

#### Citation:

von Hippel C, Kalokerinos EK and Zacher H (2017) Stereotype Threat and Perceptions of Family-Friendly Policies among Female Employees. Front. Psychol. 7:2043. doi: 10.3389/fpsyg.2016.02043 Keywords: family-friendly policies, gender, stereotype threat, work-life balance, work-family interface

# INTRODUCTION

Concerns about gender equity, recruitment, and retention push firms to provide flexible scheduling and alternative work arrangements, broadly known as work-life balance practices or familyfriendly policies (Beauregard and Henry, 2009). Family-friendly practices include a large range of programs that often focus on issues of flexible scheduling (e.g., flextime, compressed work week, telework) and support for family care (e.g., parental leave, compassionate leave, on-site childcare). Any policy or program designed to help employees balance work and family (or nonwork) constitutes a family-friendly policy. Thus, stress management programs, eldercare support, and providing a lactation room for mothers are all examples of family-friendly policies. Familyfriendly policies give organizations an advantage when they compete for employees, with these policies shown to positively impact recruitment and retention (see Beauregard and Henry, 2009 for a review of family-friendly policies and their impact on organizational performance). They are also aligned with pressure on firms to enact policies that help women work while raising a family, which have been prioritized by governments as fertility rates have fallen below replacement levels in most industrialized nations (Lesthaeghe, 2014).

At first blush, family-friendly policies appear to be an ideal solution to gender-equity concerns in the workplace. For example, these policies can enable women to continue in caregiving roles without being forced to side-track or derail their career. Nevertheless, unexpected trade-offs between well-intended human resource management practices and important employee outcomes, such as occupational well-being are not uncommon (Grant et al., 2007; Kaiser et al., 2013). Family-friendly policies may be one such practice, providing opportunities for women to better balance work and family, but at a potential cost if those who utilize these policies worry there will be negative career consequences of doing so. In this article, we report pilot data and two studies designed to investigate how experiences of stereotype threat relate to female workers' perceptions of the career consequences of utilizing family-friendly policies.

# STEREOTYPE THREAT AND FAMILY-FRIENDLY POLICIES

Women working in the corporate world often experience stereotype threat (e.g., von Hippel et al., 2015; Hoyt and Murphy, 2016), which is the concern that they are being evaluated through the lens of negative gender stereotypes (Steele, 1997). Women in corporate environments are often in positions in which their job requires characteristics that are inconsistent with their gender identity (Heilman and Okimoto, 2007). For example, a senior manager should be analytical, independent, and assertive, as these (stereotypically masculine) traits are associated with managerial success. In contrast, stereotypic female traits – being gentle, warm, and tender – are seen as inconsistent with the traits required for success in most businesses (e.g., Koenig et al., 2011). Thus, women are often required to manage their contrasting "female" and "work" identities in order to emphasize their role as skilled employees, particularly when such skills are counterstereotypic for women. Actor, writer, and producer Rashida Jones encapsulates this struggle to manage her multiple identities: "I want to be a boss and also be vulnerable. I want to be outspoken and respected, but also sexy and beautiful" (New York Times, 2016 October 18). Efforts to manage these competing identities can result in stereotype threat, as women become acutely sensitive to the possibility that they are being stereotyped (von Hippel et al., 2011a). Importantly, it is not necessary to actually be stereotyped by others to experience stereotype threat, nor must people believe the stereotype is true of their group or themselves. Rather, people need only worry that they may be stereotyped for stereotype threat effects to emerge (Steele, 1997).

In Steele's (1997) original theoretical description of stereotype threat, he discussed two types of consequences: acute performance deficits and attitudinal consequences. A large body of research has now demonstrated the performance impairing effects of stereotype threat in the laboratory (for a meta-analysis, see Nguyen and Ryan, 2008), but far less research has examined the attitudinal consequences of stereotype threat (Kalokerinos et al., 2014; Kulik et al., 2016). For example, when people experience stereotype threat they report lowered aspirations, view the stereotyped domain as less important to their self-concept, indicate less interest in participating in the stereotyped domain, and attempt to disassociate themselves from the stereotyped domain (Steele and Aronson, 1995; Davies et al., 2005; Murphy et al., 2007). Although research examining the effects of stereotype threat on organizational outcomes such as job attitudes and disengagement is still in its infancy, the studies that do exist are consistent with Steele's argument that stereotype threat can have long-term consequences on domain-relevant attitudes (e.g., von Hippel et al., 2013).

The finding that women are susceptible to stereotype threat in the workplace is not surprising, given the incongruity described above between the female gender role (e.g., being nurturing, kind, sensitive) and the masculine traits required for success in many organizations (e.g., being achievementoriented, competitive, dominant; Eagly and Karau, 2002; Schein, 2007). Indeed, experimental work in organizational settings has demonstrated that stereotype threat causes female employees to respond to these competing role demands by separating their work and gender identities, apparently as a way to manage internal conflict between these competing identities (von Hippel et al., 2011b).

In addition to such intra-personal responses to the role conflict induced by stereotype threat, women may also engage in various inter-personal strategies to manage stereotype threat and minimize the associated negative perceptions of others. For example, members of stereotyped groups have been shown to claim disinterest in activities that are stereotypic of their group (Steele and Aronson, 1995) and to claim competence in counterstereotypic domains (von Hippel et al., 2005). People in such circumstances also assert that stereotypic qualities of their group are not self-descriptive (Pronin et al., 2004). It seems possible that these processes of distancing oneself from the stereotype may play a role in women's responses to the opportunities provided by family-friendly policies.

By virtue of the fact that family-friendly policies are generally directed at and adopted by women (Sabattini and Crosby, 2009), they have the potential to cast their recipients as stereotypically female and in need of help. When women, particularly those who experience stereotype threat, are already trying to manage the impressions held by their colleagues, they may view such implications of family-friendly policies as undermining their efforts to distance themselves from the stereotype of their group. Because women are stereotyped as less committed to their careers and more focused on their families (Correll et al., 2007), they may worry that if they avail themselves of family-friendly policies they will only be confirming the very stereotype they are endeavoring to refute. Thus, to the degree that women experience stereotype threat, they may believe there are negative career consequences from using family-friendly policies, and may be reticent to utilize the very policies that have been designed to help them. Moreover, because susceptibility to stereotype threat is an indication that women feel they are undervalued in their firm by virtue of their female-stereotypic qualities, these women may be concerned that utilizing family-friendly policies may be seen as confirming these gender stereotypes, and thus exacerbate this problem. Thus, we hypothesize that the women who experience stereotype threat

will be more likely to perceive negative career consequences of using family-friendly policies.

In this paper, we present studies using experimental (pilot study) and cross-sectional designs (Studies 1 and 2) to explore this possibility. We predict that stereotype threat will be associated with perceptions that using family-friendly policies has negative career consequences. As a consequence, we also hypothesize that stereotype threat will be negatively associated with interest in using family-friendly policies.

This research contributes to the literatures on the workfamily interface and stereotype threat by highlighting the potential unintended negative consequences of family-friendly policies (i.e., perceiving policy utilization as potentially harmful to one's career). The work-family literature has not yet examined the role of stereotype threat in understanding the effects of organizational interventions on employee outcomes. Moreover, while research on stereotype threat in organizations is growing, few studies have examined outcomes other than job performance (Kalokerinos et al., 2014; Walton et al., 2015).

# PILOT STUDY

In von Hippel et al. (2011b), we manipulated an antecedent of stereotype threat to investigate the causal role of stereotype threat in identity separation among female managers. The proportion of women in a performance context has been shown to affect feelings of stereotype threat and concomitant performance (Sekaquaptewa and Thompson, 2002; Inzlicht and Ben-Zeev, 2000; Murphy et al., 2007). Thus, in that study, we manipulated the accessibility of male-dominance in the workplace by either reminding or not reminding female managers that most partners in their firm are male. By increasing the accessibility of the gender imbalance in their organization, this reminder was intended to induce stereotype threat.

The corporation in which we conducted this study was interested in their employees' perceptions of their familyfriendly programs, which focused on flexible scheduling (e.g., flexible work options, such as flextime) and alternative career paths (e.g., job sharing, telework). Thus, in collaboration with the organization we added items to the survey to assess the perceived career consequences of utilizing familyfriendly policies, as well as perceived organizational support for family-friendly policies. Although these items were originally developed in collaboration with the organization, they are nevertheless of potential interest in the context of stereotype threat. When women experience stereotype threat at work they may worry that utilizing family-friendly policies will reinforce or confirm the very stereotypes they are endeavoring to refute. As a consequence, women who experience stereotype threat may be particularly likely to believe that utilizing family-friendly policies is harmful to their careers (Hypothesis 1). We further predict that a reminder about the male dominance in the upper echelons of the firm leads to feelings of stereotype threat, which in turn causes women

to believe utilizing family-friendly work arrangements are harmful to their careers (Hypothesis 2). Finally, we explore the possibility that perceived organizational support mediates the relationship between the gender imbalance reminder and beliefs that utilizing family-friendly policies has negative career consequences.

# Method

## Participants and Design

As noted above, these pilot data were part of a larger data collection effort, and the remaining data have been previously published (von Hippel et al., 2011b). Participants were female senior managers working in the Australian office of an international accounting and consulting firm. Initially, 188 participants opened the survey, and of these participants, 169 completed the measures of stereotype threat and evaluations of family-friendly policies. Around half of the participants (42.3%) reported having children. To help ensure anonymity of participants, age was assessed categorically: 13.2% of participants were 25 to 29 years of age, 35.9% of participants were 30 to 34, 26.9% were 35 to 39, 18.0% were 40 to 44, 4.2% were 45 to 49, and 1.8% were 50 or older. The average tenure for women in the organization was 6.9 years (SD = 4.7).

Participants were randomly assigned to one of two experimental conditions when they accessed the online survey. The stereotype threat manipulation was adapted from Murphy et al. (2007), with approximately half of the participants (N = 81) reminded of the percentage of females in the firm (coded as 1). These participations read an introduction saying "Thank you for choosing to complete this questionnaire seeking to understand why only 10% of the partners at [firm name] are females. The participants (N = 88) in the control condition (coded as −1) read an introduction saying "Thank you for choosing to complete this questionnaire investigating employees' self-concepts, goals, and experiences at work."

## Measures

Responses were provided on 7-point scales, ranging from 1 = strongly disagree to 7 = strongly agree. Because the organization had a well-known family-friendly policy program in place, we referred to this specific program and did not need to define the construct for participants.

#### **Stereotype threat**

To check that the manipulation successfully elicited feelings of stereotype threat, we used a single item from von Hippel et al. (2011a): "Some of my colleagues feel that I have less managerial ability because I'm a woman." This item is intended to tap participants' concern that others may discount their ability due to their gender, and was itself adapted from Steele and Aronson's (1995) three-item scale measuring stereotype threat. After answering this question, participants completed measures unrelated to the current manuscript that were previously reported in von Hippel et al. (2011b).

#### **Perceived negative career consequences of utilizing family-friendly policies**

Perceived negative career consequences of utilizing familyfriendly policies were assessed with the single item "Women who accept family-friendly policies limit their career opportunities."

#### **Perceived organizational support for family-friendly policies**

Perceived organizational support for family-friendly policies was assessed with the single item "Family-friendly policies are not supported in day-to-day practice," which was adapted from Eisenberger et al.'s (1990) scale on perceived organizational support. This item was reverse-coded such that higher numbers indicated greater perceived organizational support.

# Results and Discussion

fpsyg-07-02043 December 28, 2016 Time: 14:54 # 4

**Table 1** provides the descriptive statistics and correlations. An independent groups t-test indicated that those in the gender imbalance condition scored significantly higher on the stereotype threat measure (M = 3.48, SD = 1.98) than those in the control condition (M = 2.70, SD = 1.72), t(167) = 2.74, p = 0.007, Cohen's d = 0.42. Measured stereotype threat was positively associated with perceived negative career consequences of utilizing family-friendly policies, supporting Hypothesis 1 (see **Table 1**).

An independent groups t-test revealed no difference in the perceived negative career consequences of utilizing familyfriendly policies between the gender imbalance (coded as 1; M = 5.22, SD = 1.59) and control (coded as −1; M = 4.84, SD = 1.85) conditions, t(166.22) = 1.44, p = 0.152, Cohen's d = 0.22<sup>1</sup> .

An independent groups t-test revealed that participants in the gender imbalance condition perceived lower organizational support for family-friendly policies (M = 3.16, SD = 1.85) than those in the control condition (M = 4.06, SD = 2.04), t(166.97) = 3.00, p = 0.003, Cohen's d = 0.46.<sup>2</sup> This finding

#### suggests a direct effect of gender imbalance on perceived organizational support for family-friendly policies. Across the whole sample (M = 3.63, SD = 1.99), stereotype threat was negatively associated with perceived organizational support for these policies (r = −0.28, p = 0.001).

To examine the hypothesis that the gender imbalance reminder would indirectly affect the perceived career consequences of family-friendly policies via stereotype threat, we conducted bootstrapped mediation analyses with 10,000 resamples (using model 4 of the PROCESS macro; Hayes, 2008). This analysis revealed that the indirect effect of the gender imbalance reminder through stereotype threat was significant (see **Figure 1**; IE = 0.07, SE = 0.04, 95% CI: 0.02, 0.18)<sup>3</sup> . Thus, Hypothesis 2 was supported.

Exploratory analyses examining perceptions of organizational support for family-friendly policies were also conducted. Bootstrapped mediation analyses testing the indirect effect of the gender imbalance reminder on the perceived organizational support for family-friendly policies via stereotype threat revealed that the indirect effects of the gender imbalance reminder through stereotype threat was significant (IE = −0.09, SE = 0.05, 95% CI: −0.23, −0.02)<sup>4</sup> . Interestingly, there was a significant indirect effect of the gender imbalance manipulation on perceived consequences through perceived organizational support (IE = 0.13, SE = 0.06, 95% CI: 0.04, 0.27).

The results of this pilot study demonstrate that a reminder about male dominance in the upper echelons of the firm induced stereotype threat (as previously reported in von Hippel et al., 2011b), which in turn led women to believe utilizing family-friendly work arrangements would be harmful to their careers. Exploratory analyses suggest that one explanation for these findings may be that a small proportion of women in the upper echelons of the organization leads to perceptions that the organization does not really support family-friendly policies in day-to-day practice, which in turn is associated with perceived negative career consequences of policy utilization. Interestingly, although the policies offered by the organization are couched as "family-friendly," as can be seen in the

#### TABLE 1 | Descriptive statistics and correlations for pilot study.


FFP, family-friendly policies. Parental status is coded as −1 = no children, 1 = children. All correlations use pairwise deletion (N varies between 167 and 169). <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

<sup>1</sup>Assumptions regarding homogeneity of variance were not met [Levene's F(197) = 4.09, p = 0.045], and so we report a conservative t-statistic (which does not use the pooled statistic for the error term), and adjust the degrees of freedom using the Welch–Satterthwaite method. These results remained non-significant when controlling for parental status (−1 = no children, 1 = children) in an ANCOVA, F(1,164) = 2.15, p = 0.145, η 2 <sup>p</sup> = 0.01. Finally, no interaction emerged between the gender imbalance manipulation and parental status in predicting perceived consequences of utilizing family-friendly policies, F(15,132) = 0.94, p = 0.525, η 2 <sup>p</sup> = 0.10.

<sup>2</sup>As per footnote 2, t-values and degrees of freedom were adjusted because assumptions regarding homogeneity of variance were not met. These effect remained significant when controlling for parental status (−1 = no children, 1 = children) in an ANCOVA, F(1,164) = 8.97, p = 0.003, η 2 <sup>p</sup> = 0.05.

<sup>3</sup>We also ran this model including parental status and the parental status × gender imbalance manipulation as covariates, and found the same indirect effect of the gender imbalance reminder through stereotype threat (IE = 0.09, SE = 0.04, 95% CI: 0.02, 0.20).

<sup>4</sup>This indirect effect remained when controlling for parental status and the parental status × gender imbalance manipulation interaction (IE = −11, SE = 0.06, 95% CI: −0.25, −0.03).

footnotes the results remained unchanged whether women had children living at home or not. This finding suggests that women may contend with the stereotype that they must manage work and family, irrespective of whether they have children. Such a possibility is unsurprising given that women do more housework than men (Bianchi et al., 2012) and are more likely to care for elderly parents (Van Houtven et al., 2013). Nevertheless, this pilot study does not distinguish between women who have older, largely independent children living at home and women with young children who need substantial care. It is possible the relationship between stereotype threat and perceived negative career consequences of policy utilization would be more pronounced among women who have infants and toddlers. Study 1 examines this possibility by sampling female employees who recently returned to work after the birth of a child, while also addressing the limitation of relying on a single item measure of the key outcome variable.

# STUDY 1

Stereotypes about working mothers are even more pervasive than those of working women. Working mothers are thought to spend less time at work and consequently are seen as less productive compared to their male counterparts (Heilman and Okimoto, 2008; Wallace and Young, 2008). Mothers are thought to be conserving energy for their family responsibilities, or are perceived as having less energy to expend at work after meeting their domestic responsibilities (Voydanoff, 2004). These stereotypes can make it even more challenging for women to manage their multiple identities (Hodges and Park, 2013), where the expectations of being a good mother conflict with those of being a good employee (Wallace and Young, 2008). Thus, Study 1 sought to provide a conceptual replication of the relationship between stereotype threat and perceived negative career consequences of utilizing family-friendly policies among a sample of mothers who recently returned to work after the birth of a child. Study 1 also used an expanded measure of the perceived negative career consequences of utilizing familyfriendly policies to provide a more rigorous test of these ideas.

# Method

#### Participants and Design

Participants were 65 working mothers who recently returned to work after the birth of a child (Mage = 35.08, SDage = 4.82). They were recruited through a newsletter distributed by a local center for mothers and babies in Australia. Women who had recently had a baby were asked to participate in an online survey, at the end of which they could provide an email address to be entered into a prize draw to win a \$100 gift card. Participants worked in a range of industries, including healthcare and social assistance (35.4%), professional, scientific, and technical (27.7%), information occupations (7.7%), and retail (6.2%). The majority of respondents (70.7%) worked in organizations that offer family-friendly policies, although 18.5% reported their organizations did not offer family-friendly policies. The remaining participants (9.2%) were not sure whether their organization offered family-friendly policies (1 person did not answer this question). The average tenure with their organization was 6.81 years (SD = 5.52), and the average number of children was 1.68 (SD = 0.83).

#### Measures

Before completing the survey items, participants were provided with the following explanation: "Family-friendly policies consist of programs, resources and procedures that organizations have in place to help employees balance work and family responsibilities. Family-friendly policies can include things like flexible working hours and job sharing. We're interested in your opinions about family-friendly policies more generally, not a particular program." In addition to the measures below, participants responded to items unrelated to the central goals of this manuscript. For transparency, the additional constructs measured in Studies 1 and 2 (and their correlations with stereotype threat) are provided in an Appendix. All data are available upon request.

#### **Stereotype threat**

Stereotype threat was measured on a 7-point scale (1 = strongly disagree, 7 = strongly agree) using four items from the scale developed by von Hippel et al. (2011a; α = 0.96; e.g., "Some of my colleagues feel I'm not as committed to my career because I'm a woman") and adapted from Steele and Aronson (1995).

#### **Perceived negative career consequences of utilizing family-friendly policies**

Perceived negative career consequences of utilizing familyfriendly policies were assessed on a 7-point scale (1 = strongly disagree, 7 = strongly agree) using four items (α = 0.93) developed for this study based on previous research. One item ("Using family-friendly programs would harm my status at work") was adapted from Butler et al. (2004). Two items ("Using family-friendly programs would hurt my career progress"; "Using family-friendly programs would suggest that I'm not as serious about my career as employees who don't use these programs") were adapted from Thompson et al. (1999). A final item ("Using family-friendly programs would result in negative outcomes for me at work") was developed for this study.

# Results and Discussion

fpsyg-07-02043 December 28, 2016 Time: 14:54 # 6

Consistent with the pilot data, experiences of stereotype threat were associated with more perceived negative career consequences of utilizing family-friendly policies (r = 0.28, p = 0.029). This effect remained significant when controlling for number of children (β = 0.31, p = 0.016).

This study replicated the results in the pilot study showing a positive link between stereotype threat and perceived negative career consequences of utilizing family-friendly policies. This study focused exclusively on working mothers, a sample in which family-friendly policies are likely to be of central importance. Family-friendly policies often target this population, and thus the association between stereotype threat and perceived negative career consequences of utilizing familyfriendly policies in this sample is indicative of the potential applied importance of our findings. A limitation of this study is its cross-sectional and correlational design, which does not allow conclusions about causality. Additionally, although women who experience stereotype threat perceive more negative career consequences from utilizing family-friendly policies, these data do not speak to women's reticence to actually use the policies. Study 2 addressed this unanswered question.

# STUDY 2

Women who experience stereotype threat are likely to take steps to distance themselves from the stereotype (Steele and Aronson, 1995; von Hippel et al., 2005). This process of distancing oneself from the stereotype may manifest itself in women being less likely to avail themselves of familyfriendly work policies. Thus, we hypothesized that to the degree that women experience stereotype threat, they will report less interest in utilizing the very policies that have been designed to help them (Hypothesis 3). In addition, we aimed to replicate the relationship between stereotype threat and perceived consequences of family-friendly policies (Hypothesis 1).

# Method

#### Participants and Design

Participants were 473 working women (Mage = 30.01, SDage = 10.57) in Australia recruited using a convenience sampling approach, including advertising in a university staff newsletter, and forwarding the survey to organizational contacts. In return for their participation, participants were given the chance to win one of several \$100 gift cards. Nineteen percent (19.1%) of participants had children living at home, and 60.4% of participants intended to have their first child, or more children, in the future. Of the women who were currently childless, 68.8% reported that they wanted to have children in the future. Participants worked in a number of different industries, most frequently education (25.8%), science and technology (15.8%), mining (7.2%), and health and community services (6.8%). Their average tenure in their current organization was 3.48 years (SD = 4.16).

#### Measures

#### **Stereotype threat**

Stereotype threat (α = 0.87) was assessed using the 4-item measure from Study 1.

### **Perceived negative career consequences of utilizing family-friendly policies**

Perceived negative career consequences of utilizing familyfriendly policies were assessed using a single item from Thompson et al. (1999) that was used in Study 1: "Using family-friendly programs would suggest that I'm not as serious about my career as employees who don't use these programs." <sup>5</sup> Participants responded on a 7-point scale, ranging from 1 = strongly disagree to 7 = strongly agree.

#### **Interest in utilizing family-friendly policies**

To measure interest in utilizing family-friendly policies, participants were presented with a list of eight common family-friendly programs and asked how likely they would be to use each program if it were available in their workplace (α = 0.78). Participants responded on a 7-point scale ranging from 1 = very unlikely to 7 = very likely. The list of common family-friendly programs was adapted from a measure developed by Hammer et al. (2005). The programs included were alternative work arrangements, flexible work hours, job-sharing, telecommuting, unpaid leave, personal time off/paid leave, on-site support groups, and work and family seminars.

# Results and Discussion

Descriptive statistics and correlations are presented in **Table 2**. An independent-groups t-test showed no significant difference in stereotype threat between women who did not have children (M = 3.00, SD = 1.42) and women who had children (M = 2.91, SD = 1.58), t(454) = 0.51, p = 0.613, Cohen's d = 0.06. A second independent groups t-test showed that there was also no significant difference in stereotype threat between women who were planning to have their first child, or more children, in the future (M = 3.07, SD = 1.49), and women who were not planning to have children, or to have more children (M = 2.87, SD = 1.40), t(448) = 1.42, p = 0.157, Cohen's d = 0.14.

In support of Hypothesis 1 and consistent with Study 1, stereotype threat was positively correlated with perceived

<sup>5</sup>A second item from Thompson et al. (1999) was included in the survey and can be seen in the Appendix. This item is not about family-friendly policies and thus is not included in these analyses. Nonetheless, the results remain unchanged when both items are included in the analyses.

#### TABLE 2 | Descriptive statistics and correlations for Study 2

fpsyg-07-02043 December 28, 2016 Time: 14:54 # 7


FFP, family-friendly policies. All correlations use pairwise deletion (N varies between 384 and 460). Children living at home and intention to have children in the future are coded as −1 = no, 1 = yes. <sup>∗</sup>p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001.

negative career consequences from using family-friendly policies (see **Table 2**) 6,7 .

Unexpectedly, there was a positive correlation between stereotype threat and interest in utilizing family-friendly policies, such that women who experienced stereotype threat at work indicated more interest in utilizing familyfriendly policies (see **Table 2**), despite their beliefs that doing so would hurt their career. Thus, Hypothesis 3 was not supported. The perceived negative career consequences of utilizing family-friendly policies variable was uncorrelated with interest in utilization of family-friendly policies (see **Table 2**).<sup>8</sup>

To explore whether the relationship between stereotype threat and interest in utilization was mediated by perceived

<sup>8</sup>To control for parental status we conducted a hierarchical multiple regression analysis in which we entered parental status at the first step, and stereotype threat at the second step, to predict perceived interest in utilization. We found that the stereotype threat – utilization relationship was also present when controlling for parental status (β = 0.13, p = 0.005). We conducted a second such analysis including intention to have (more) children in the future, instead of parental status. We also found that the stereotype threat-utilization relationship remained significant in this analysis (β = 0.12, p = 0.011). Finally, and consistent with the prior results, there was no interaction between stereotype threat and parental status (β = 0.03, p = 0.533), or stereotype threat and intention to have children (β = 0.05, p = 0.556), in interest in utilizing family-friendly policies.

consequences, we conducted bootstrapped mediation analyses with 10,000 resamples (using model 4 of the PROCESS macro; Hayes, 2008). There was no significant indirect effect of stereotype threat on interest in utilization through perceived negative career consequences of utilizing family-friendly policies (IE = −0.01, SE = 0.01, 95% CI: −0.04, 0.01).

This study replicated our previous findings that stereotype threat is positively related to perceived negative career consequences associated with using family-friendly policies. Unexpectedly, however, stereotype threat was positively associated with women's intentions to utilize work-family practices. Family-friendly policies may be perceived as a double-edged sword among women who experience stereotype threat – women perceive the policies as harmful to their career but feel they have no choice but to use them. Importantly, however, we did not measure actual utilization, nor do we know whether participants had access to family-friendly policies. Nonetheless, these data suggest that women who experience stereotype threat feel the benefits of these policies outweigh the potential career costs associated with using them. In an ideal world, family-friendly policies would be an unmitigated plus, but the unfortunate reality appears to be that among women who experience stereotype threat these policies (are perceived to) incur a cost.

# GENERAL DISCUSSION

Employees in organizations that offer family-friendly policies and programs hold more positive job attitudes and are less likely to withdraw from work (Kossek and Ozeki, 1998; Anderson et al., 2002; Kossek et al., 2014). The current research, however, suggests that these benefits may not be realized among women who experience stereotype threat. Across two studies (and a pilot study) with working women, we found that stereotype threat was associated with the perception that taking advantage of family-friendly policies would have negative career consequences. Interestingly, controlling for whether women had children living at home (pilot study and Study 2), or whether they intended to have children in the future (Study 2) did not change any of the results, suggesting that issues surrounding family-friendly policies are relevant to all working women, not just current or future mothers.

<sup>6</sup>As noted in the Appendix, this study also measured gender identification, and thus we are able to explore whether, consistent with previous laboratory research, gender identification moderates these relationships. The interaction between gender identification and stereotype threat does not predict the perceived consequences of family-friendly policies (β = 0.03, p = 0.556) or interest in utilizing family-friendly policies (β = 0.05, p = 275). McClelland and Judd (1993) note that power to detect an interaction in a field study is much lower than in traditional lab studies, and thus this null effect should be interpreted with caution. <sup>7</sup>To control for parental status we conducted a hierarchical multiple regression analysis in which we entered parental status (−1 = no children, 1 = children) at the first step, and stereotype threat at the second step, to predict perceived negative career consequences. This analysis showed that the stereotype threat – career consequences relationship was also present when controlling for parental status (β = 0.42, p < 0.001). We conducted a second hierarchical multiple regression analysis in which we entered intention to have (more) children (−1 = no intention to have (more) children, 1 = intent to have (more) children in the future) at step one, and stereotype threat at the second step, to predict negative career consequences. The stereotype threat – career consequences relationship remained present when controlling for intention to have (more) children in the future (β = 0.41, p < 0.001). Finally, we also conducted two hierarchical multiple regression analyses in which stereotype threat and parental status, or stereotype threat and intention to have (more) children were entered at the first step, and their interaction at the second step. Consistent with the pilot study, there was no significant interaction between stereotype threat and parental status (β = 0.00, p = 0.998), or between stereotype threat and intentions to have children (β = 0.04, p = 0.556), in predicting perceived career consequences.

# The Disconnect between Perceived Career Consequences and Interest in Policy Utilization

Ironically, Study 2 suggests that, despite perceptions of negative career consequences of utilizing these policies, women who experienced stereotype threat were more interested in using them. Although the current studies do not provide data that help us understand this disconnect, there are several potential explanations. For example, perhaps women experiencing stereotype threat feel the benefits of the policies outweigh the costs associated with using them. If so, familyfriendly policies may be perceived as a double-edged sword, whereby women perceive the policies as costly to their career, but feel that the benefits they bring are necessary.

In our previous work, stereotype threat has been associated with work disengagement, including intentions to quit and to retire (e.g., von Hippel et al., 2013). In this sample, the increased interest in using family-friendly policies in the presence of perceived costs may be another signal of disengagement from work. Women who feel stereotype threat may believe they have poor prospects in their career and so disengage from work, taking advantage of family-friendly policies to make life easier in other domains.

# Family-Friendly Policies Can Be Problematic

Our work suggests that, like affirmative action strategies, familyfriendly policies may even the playing field in principle, but have negative consequences in practice (Heilman et al., 1992). Previous work examining affirmative action policies found that people perceive the recipients of such policies as in need of help, less committed, and as stereotypic of women in general (Heilman et al., 1992). Our work suggests that working women may be aware of such perceptions, and rightly perceive familyfriendly policies as having negative career consequences. Thus, it is particularly important that we better understand what relates to perceptions of negative career consequences of utilizing familyfriendly policies, both among people who use them and among those who do not. Additionally, future research should consider how employees perceive their co-workers who make use of family-friendly policies.

# Limitations and Future Directions

As with any research it is necessary to interpret these findings in light of their limitations. Although this research focused on "family-friendly policies," whether women had children or not did not impact the results. It is possible that all women, irrespective of parent status, must contend with the stereotype that they need to manage work and family. Such a possibility is consistent with research showing that women do more housework than men and are also more likely to provide support for elderly parents (Bianchi et al., 2012; Van Houtven et al., 2013). Future research should attempt to disentangle "family-friendly policies" from other workplace flexible practices.

The field settings of our studies are both a strength and a weakness. On the strength side, participants were working women (many of whom were also mothers) and thus familyfriendly policies are consequential in their lives. But working women are busy people, and so in an effort to maximize participation rates we kept the surveys as short as possible. As a consequence, we did not examine various mediators and moderators that might facilitate understanding of the current findings. For example, do women who experience stereotype threat worry that utilizing these policies will signal they are prioritizing family over work? Will the relationship between stereotype threat and perceived career consequences be attenuated by individual (e.g., self-esteem) or work-related resources (e.g., hierarchical position)? The pilot study provides suggestive evidence that perceived organizational support mediates the relationship between gender imbalance at the top of the organizational hierarchy and perceived negative career consequences of policy utilization. This finding is particularly troubling in light of research demonstrating that perceptions of support for such policies can be more important than availability of the policies themselves (Behson, 2005; Wayne et al., 2013). Indeed, organizations find that even when they have desirable family-friendly policies in place, employees often do not have access to these policies (Shrm Survey Findings, 2015). Due to the organization-specific nature of perceived organizational support (e.g., participants working in organizations without family-friendly policies cannot answer a question about perceived support in day-to-day practice) we were unable to pursue this line of inquiry in Studies 1 and 2 (which relied on employees from numerous organization). Thus, further research is required to have confidence in this mediating mechanism.

Although the pilot study manipulated the salience of gender imbalance in the organization to elicit stereotype threat, the remaining studies relied on correlational and cross-sectional designs. Thus, it is possible that stereotype threat is not a causal mechanism in these correlational studies. For example, women who use family-friendly policies may believe utilizing these policies sends the message that they are unable to balance family and work, and thus, as a consequence, might be more susceptible to experiences of stereotype threat. Additionally, an unmeasured variable may account for these relationships. For example, an unwelcoming organizational climate is likely to lead to feelings of stereotype threat as well as beliefs that utilizing family-friendly policies have negative career consequences. This possibility is consistent with Steele's (1997) theorizing – an unwelcoming climate will lead to feelings of stereotype threat because it suggests to women that they have "poor prospects" to advance their career and causes women to feel a "lack of belonging." Longitudinal or experimental research designs are required to better understand these relationships.

Traditional gender roles involve women in the role of primary caregiver, and thus women are more susceptible to identity conflict from work than their male counterparts (Hodges and Park, 2013; Williams et al., 2016). For example, the role of "good" mother and wife is very different to that of "good" father and husband (Nomaguchi et al., 2005). The expectations of fatherhood and the demands of work roles often coincide, whereas the expectations of motherhood and work usually conflict (Milkie and Peltola, 1999; Okimoto and Heilman, 2012).

For these reasons, our research focused exclusively on women, but future research would benefit from understanding men's perceptions of these issues. Family-friendly policies are important for men, and research needs to address the potential workfamily conflict that men experience. Research suggests men face a "flexibility stigma" whereby utilizing family-friendly policies calls into question their devotion to the job (e.g., Rudman and Mescher, 2013). This stigma might be particularly strong in organizations where there are few women because there is less likely to be a norm of acceptability around using family-friendly policies. More generally, family-friendly policies oriented toward women and not men make it more difficult to change the norm of women as primary caregivers. Finally, future research should examine whether these results generalize to other groups who are susceptible to stereotype threat in the workplace (e.g., older employees; ethnic minority group members).

# CONCLUSION

Demographic, economic, and egalitarian pressures have coalesced to bring family-friendly policies to the center of many organizations' staffing practices. Our research suggests that female employees are susceptible to stereotype threat, which in turn is associated with more negative views of familyfriendly policies. These results highlight the difficulties faced by companies who offer family-friendly policies, and clarifies the need for organizations to better communicate and promote their policies. Identity safe workplaces are necessary to reduce experiences of stereotype threat, which in turn should reduce perceptions that family-friendly policy utilization has negative career consequences. Although these policies are designed to help employees, there may be perceived costs of utilization in organizational climates where women feel stereotyped. In

## REFERENCES


summary, these findings suggest that the mere provision of family-friendly policies is unlikely to create the kind of familyfriendly workplaces that organizations are attempting to provide their employees.

# ETHICS STATEMENT

This study was approved by University of Queensland's School of Psychology Ethics Review Panel (Studies 2 and 3) and University of New South Wales School of Psychology Ethics Review (Study 1). Participants read an information sheet and then chose to continue to the survey or exit the survey (by closing their web browser).

# AUTHOR CONTRIBUTIONS

CvH and EK designed the studies and coordinated data collection. EK analyzed the data. CvH, EK, and HZ discussed the results and wrote the paper. All authors approved the final submission.

# ACKNOWLEDGMENT

We thank Arianne Zouroudis for collecting the pilot data.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpsyg. 2016.02043/full#supplementary-material



Survey\_Findings\_Strategic-Benefits-Flexible-Work-Arrangements.pdf [accessed June 16, 2016]


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer LM declared a shared affiliation, though no other collaboration, with the author EK to the handling Editor, who ensured that the process nevertheless met the standards of a fair and objective review.

Copyright © 2017 von Hippel, Kalokerinos and Zacher. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Similarly Torn, Differentially Shorn? The Experience and Management of Conflict between Multiple Roles, Relationships, and Social Categories

Janelle M. Jones<sup>1</sup> \* and Michaela Hynie<sup>2</sup>

<sup>1</sup> Department of Biological and Experimental Psychology, School of Biological and Chemical Sciences, Queen Mary University of London, London, United Kingdom, <sup>2</sup> Department of Psychology, York University, Toronto, ON, Canada

In three studies we examined the experience and management of conflict between different types of multiple identities. Participants described a conflict between pairs of role, relational, or social identities before rating the experience (i.e., magnitude, stress, and growth) and management of conflict on a newly developed scale assessing four strategies: reconciliation, where identities are integrated, realignment, where one identity is chosen over another, retreat, where both identities are avoided, and reflection, where fit (with others, situation) determines identity selection. In general, the types of identities mattered for conflict management but not its experience: Magnitude and growth did not differ, however, stress was greater for role identity conflicts (Study 3 only) and participants endorsed the use of more realignment for role conflicts (Study 2) and more retreat for relational conflicts (Study 3) relative to other types of identity conflicts. Furthermore, findings suggested that the perceived flexibility of identities, not their importance or valence, were associated with realignment and retreat for roles and with retreat for relationships. Experiencing conflicts between multiple identities leaves people similarly torn, but multiple roles and relationships may be differentially shorn to manage conflict.

Keywords: multiple identities, identity conflict, conflict management strategies, roles, relationships, social categories

# INTRODUCTION

Student. Friend. British. These are some of the roles, relationships and social categories from which individuals can derive a sense of identity – knowledge of who they are, how they should act, and/or their place in the world (Tajfel, 1981; Tajfel and Turner, 1986; Turner et al., 1994; Deaux et al., 1995). These multiple identities are important ways that individuals come to understand themselves and others. Indeed, a considerable body of theoretical and empirical work on multiple identities has shed light on the ways that individuals perceive multiply categorizable targets and respond to (mis-) categorizations by others, as well as the implications of multiple identities for individuals' cognition, well-being, and interactions (see Kang and Bodenhausen, 2015 for a comprehensive review). However, our understanding of how individuals experience and manage multiple identities within themselves is still in its infancy. Existing work has focused on the tension between multiple cultural identities (i.e., biculturalism; Benet-Martínez et al., 2002;

#### Edited by:

Natasza Kosakowska-Berezecka, University of Gdansk, Poland

#### Reviewed by:

Sarah Gaither, Duke University, United States Małgorzata Anna Gocłowska, University of Amsterdam, Netherlands

> \*Correspondence: Janelle M. Jones j.jones@qmul.ac.uk

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

Received: 11 November 2016 Accepted: 19 September 2017 Published: 05 October 2017

#### Citation:

Jones JM and Hynie M (2017) Similarly Torn, Differentially Shorn? The Experience and Management of Conflict between Multiple Roles, Relationships, and Social Categories. Front. Psychol. 8:1732. doi: 10.3389/fpsyg.2017.01732

Benet-Martínez and Haritatos, 2005) or multiple racial identities (Shih and Sanchez, 2005; Sanchez et al., 2009). Yet, as highlighted above, social categories such as culture and race reflect only one possible type of identity. The extent to which other types of identities such as roles and/or relationships result in tension requires further scrutiny. Moreover, given people's increasing awareness and adoption of multiple identities across contexts and time (e.g., Hodges and Park, 2013; Goclowska and Crisp, 2014), the nature of any tension experienced and how it is managed for different types of identities are important empirical questions.

To begin to address these issues, the present research conducted three studies on the experience and management of identity conflict, a perceived intrapersonal tension, discrepancy or interference, between at least two identities. In particular we focused on identity conflicts between pairs of roles, relationships, and social categories – key dimensions on which individuals base their identities (Baumeister et al., 1985; Deaux et al., 1995; Benet-Martínez et al., 2002; Settles, 2004; Benet-Martínez and Haritatos, 2005; Hirsh and Kang, 2016). We assessed the experience of conflict between multiple role, relational, and social identities by focusing on perceptions of conflict magnitude and associated stress and growth. Drawing from psychological, sociological and cultural theories on multiple identities (Baumeister et al., 1985; Berry, 1997, 2005; Roccas and Brewer, 2002; Burke, 2003, 2006; Goclowska and Crisp, 2014; Amiot et al., 2015) we developed a scale to assess the management of conflict between multiple identities. We also considered whether the characteristics associated with role, relational, and social identities, namely flexibility, valence, and importance (Deaux et al., 1995; Settles, 2004; Brook et al., 2008; Rabinovich and Morton, 2016) might influence the experience and/or management of conflict. This work makes an important contribution to the literature on multiple identities as, to our knowledge, it is the first to investigate whether the types of identities matter for the experience and management of conflict. Critically, this work is also the first to develop a scale to measure identity conflict management. Below we briefly review the literature on conflict and self-organization as it relates to multiple identities to build a case for examining the experience and management of conflict between multiple identities.

# Conflict between Multiple Identities: Experiences and Consequences

According to Baumeister et al. (1985), identity conflict arises when more than one identity is elicited in a given situation and these identities dictate different commitments, or sets of norms, values, motives and goals for the individual (see also Settles, 2004; Brook et al., 2008; Hirsh and Kang, 2016). Under these circumstances, individuals must contend with different ways of defining or expressing themselves, a decision that may be further complicated by the awareness that choosing one identity can come at the expense of another identity. For instance, identity conflict can occur when an individual adopts a new identity that conflicts with an old identity such as a female student in the sciences (e.g., Settles, 2004). Here prevailing norms that science is associated with men may dictate that a scientist identity contradicts a female identity (but cf. Goclowska and Crisp, 2014). Competing demands and expectations associated with two identities such as being a student and an athlete or being a parent and a professional can also promote conflict (e.g., Settles et al., 2002; Hodges and Park, 2013). Here, individuals may find it challenging to shift between two of their existing identities, leading to difficulties in the performance of one or both of these identities. Specific contexts can also cue multiple identities and highlight conflict. For instance, two existing and incompatible identities might be elicited for a lesbian-Christian attending a Sunday church sermon that condemns homosexuality; see Mahaffy, 1996; Borgman, 2009). What is striking about these examples is both the diversity of the identities that can be involved in conflict and the mundane nature of the catalysts of conflicts – Whether roles, relationships, social categories, or a combination, more than one identity can potentially be brought to mind at different times and in different contexts, potentially promoting difficulties for individuals in everyday life.

Despite having a sense of what conflict is, and when and where conflict might arise, less is known about how conflict is experienced and what it is about the identities involved that might constrain the experience of conflict. We believe that insight into these issues can be gained by considering who experiences conflict. The classic example is found in Benet- Martínez's work on biculturals, individuals who straddle at least two different cultures (e.g., Asian-Americans; Benet-Martínez et al., 2002; Benet-Martínez and Haritatos, 2005). Here, individuals can differ in their bicultural identity integration (BII), the extent to which their two cultural identities are associated with cultural distance, whether cultures are (in)compatible with each other, and cultural conflict, whether cultures are (not) in tension with each other. High BII is associated with seeing cultural identities as compatible whereas low BII is associated with seeing cultural identities as conflictual. In addition to highlighting that some individuals may be more or less likely to experience conflict, the perceived discrepancy between identities is an important aspect of how (and whether) conflict is experienced, ultimately influencing individuals' scores on BII. Indeed Cheng and Lee (2013) have demonstrated that the measure of BII can shift, increasing after recalling positive bicultural experiences and decreasing after recalling negative bi-cultural experiences. This finding is important because it suggests that individuals' experience of their multiple identities can shape their evaluations of the perceived discrepancy between these identities.

This is not to say that perceiving a discrepancy between multiple identities will always lead to conflict. We know that identities derived from roles, relationships, and social groups are perceived as distinct from each other and differ on key dimensions such as whether identities are ascribed vs. achieved (i.e., automatic rather than chosen group memberships) and/or central vs. peripheral (i.e., core and important rather than marginal and unimportant; Deaux et al., 1995; see also McConnell and Strain, 2007; McConnell, 2011). In fact these differences between the types of identities may shape perceptions of their importance, valence, and flexibility – characteristics that might influence whether perceived discrepancies between multiple identities promote conflict or not. Indirect evidence

that characteristics matter comes from several sources: Settles (2004) demonstrated that the importance of multiple identities moderated the magnitude of the conflict experienced. Brook et al. (2008) found that negative self-discrepant emotions mediated the relationship between conflicting multiple identities and lower psychological well-being. And Rabinovich and Morton (2016) examined individuals' perceptions of self-flexibility as related to the experience of conflict between multiple roles. Here when the self was perceived as inflexible (i.e., as stable and fixed) experiencing conflict between multiple roles was seen as a problem. If we extend this idea about flexibility to the elements of the self rather than the overall self-concept, it is equally plausible that the perceived flexibility of identities themselves might influence the experience of conflict. Overall then, there is some suggestion that the types of identities, and their associated characteristics, might influence the experience of conflict. However, what individuals do to manage the conflict experienced between their multiple identities, and whether the types of identities and/or their characteristics are related to conflict management is less clear. To begin to address these questions we draw from existing theoretical models and empirical research to propose specific identity conflict management strategies.

#### Strategies for Managing Conflict between Multiple Identities

Identity theorists in psychological, sociological, and cultural traditions have proposed various ways that individuals might organize their multiple identities within the self or manage these identities at times of conflict and transition. For instance, in their model of social identity complexity, Roccas and Brewer (2002) contend that social identities differ in their degree of overlap in one of four ways: Intersection, where a compound identity (e.g., female-scientist) is used to define one's ingroup, dominance, where one identity (e.g., scientist) serves as the superordinate category that defines the individual, compartmentalized, where each social identity is completely isolated from the others and only the relevant identity is elicited in a specific context (e.g., one's scientist identity would only be activated at work but not at home), and merged, where both social identities would be used concurrently to define the individual in a given context (e.g., identifying both with women and scientists). In a related vein, Amiot et al. (2007, 2015) consider how pre-existing and new identities are organized within the self. In addition to categorizing themselves in terms of the new group, individuals engage in compartmentalization, where they keep their new and existing group memberships separate, and integration where elements of the new and existing groups are brought together. Similarly, Goclowska and Crisp (2014) proposed that when faced with a new identity that is similar to an existing identity, there is no change. However, when a new identity differs from an existing identity, individuals may alternate between these identities, integrate aspects of both identities into their self, or include the other identity into their self. As we can see integration, alternation, and compartmentalization feature as consistent ways that individuals structure multiple identities within the self.

Proposed strategies for managing conflicting multiple identities are similar to those advanced to explain how multiple identities are organized with the self. Baumeister et al. (1985) suggested that individuals might choose one identity over the other, attempt to compartmentalize their identities, separating them fully such that conflict is not possible, or try to make a compromise between their identities such that one identity is temporarily allowed to take prominence over the other, to manage the intrapsychic discomfort experienced when faced with conflicting multiple identities. Burke (2003) also identified three ways in which identities could be managed when in conflict: Individuals might choose to withdraw from one of their identities by selecting the one identity deemed more important to themselves, compromise by avoiding the situations that led to conflict between their multiple identities, or try to balance or change the meanings of both identities so that their identities approached each other. More recently Hirsh and Kang (2016) proposed that individuals could suppress the problematic identity, enhance the dominant identity, avoid situations that elicit both identities, or engage in integration as ways to manage conflict between multiple identities. Taken together these models suggest that individuals use similar strategies when trying to make sense of who they are, integrating new identities with existing identities, and in order to reduce the intrapsychic discomfort that can arise when multiple identities are elicited and conflict with each other (Baumeister et al., 1985; Roccas and Brewer, 2002; Burke, 2003; Amiot et al., 2007, 2015; Goclowska and Crisp, 2014; Hirsh and Kang, 2016).

Synthesizing the foregoing literature we propose that individuals manage conflicting multiple identities in at least three ways: Reconciliation, where individuals try to balance their multiple identities by integrating aspects of both identities, Realignment, where individuals choose one identity over another identity by selecting one identity to enact, or by focusing on one identity more than the other, and Retreat, where individuals avoid both of their conflicting identities, effectively compartmentalizing them. In this case, individuals may withdraw from situations where both of their identities are elicited or they may ignore both of their identities as they are incompatible with each other. These proposed identity conflict management strategies are not exhaustive, rather they represent a starting point in light of the preceding literature. Importantly, these strategies are not conceptualized as individual differences. Instead we contend that they may be used alone or in different combinations to manage conflicting multiple identities.

Although the strategies detailed above remain largely untested in relation to the experience of identity conflict, similar strategies have been examined with respect to acculturation, where individuals attempt to make sense of new and existing cultural information in order to adapt to a new context (Berry, 1997, 2005). Here individuals may engage in assimilation, where they choose the new culture over their existing culture, separation, where the existing culture is maintained while avoiding the new culture, integration where the new culture is adopted and the existing culture maintained, and marginalization, where individuals dis-identify with both their existing and new cultures. The similarities between the proposed

strategies and acculturation strategies are easy to identify, with realignment reflecting aspects of assimilation, retreat reflecting aspects of marginalization and separation, and reconciliation reflecting aspects of integration. Importantly, Berry has found that integration is associated with positive adaptation to the environment whereas marginalization is associated with negative adaptation (assimilation and separation are intermediate), and argues that integration is associated with protective factors (i.e., societal and social support, flexibility in personality) relative to the other strategies, thereby encouraging its use (Berry et al., 2006; Sam and Berry, 2010). This finding suggests that all individuals might try some form of reconciliation to manage conflicting multiple identities given its benefits.

# RESEARCH OVERVIEW

The aims of the present research were threefold. First, we were interested in understanding the experience of conflict. This issue was approached in two ways: We know that conflict can occur between multiple identities, and arise for a variety of reasons, yet we do not have a sense of the types of multiple identity conflicts that typically occur. As such, we considered the nature and frequency of conflicts between multiple identities. We were also interested in understanding the experience of conflict between different types of multiple identities (e.g., pairs of roles, pairs of relationships). Here too, we have some idea that identity conflict is uncomfortable and associated with specific consequences but we do not know whether these experiences are similar across different types of identities. Second, we were interested in understanding how individuals manage conflicting multiple identities. Although existing models provide some indication of relevant identity conflict management strategies, to our knowledge, there has been little empirical research to address whether these strategies reflect the actual things that people do to manage identity conflict, or whether different types of multiple identities influence the conflict management strategies used. Third, we were interested in understanding the characteristics of multiple identities that might be related to the experience and/or management of conflict. Given that roles, relationships, and social groups differ from each other on dimensions that might be related to flexibility, importance, and valence, these characteristics might be associated with the experience and/or management of conflict. We examined these issues across three studies.

Study One provided a descriptive account of the frequency, experience and management of conflicts between multiple identities and was used to establish our initial hypothesis. Here participants described a current or recent identity conflict, rated its experience (i.e., magnitude, stress, and growth), and described the way(s) they managed this conflict. Studies 2 and 3 adopted an experimental approach to examine whether the types of multiple identities might influence the experience and management of identity conflict. In these studies participants were randomly assigned one of three conditions where they were asked to describe a conflict between two specific roles (i.e., student-employee), relationships (i.e., family-friends), or social categories (i.e., ethnicity-nationality) before rating their experience of conflict (i.e., magnitude, stress, and growth). We developed a measure of conflict management (Study 2) and examined whether types of identities in conflict influenced the use of these conflict management strategies (Studies 2 and 3). We also examined whether the characteristics of role, relational, and social identities in general (i.e., flexibility, importance, and valence) were associated with the experience and management of conflict (Study 3). Given the lack of prior empirical work on these specific relationships we developed and built on our hypotheses in stages in response to our findings. Hypotheses are presented where made below.

# STUDY 1: UNDERSTANDING THE FREQUENCY, EXPERIENCE AND MANAGEMENT OF IDENTITY CONFLICT

The aim of Study 1 was to gain insight into the frequency, experience, and management of different types of identity conflict. Given the exploratory nature of these questions, specific hypotheses were not made.

# Method

#### Participants, Procedure, and Measures

Participants were 116 undergraduate psychology students (Age: M = 20.70; SD = 5.43; Gender: Female: n = 101; Male: n = 15; Race/Ethnicity: White: n = 48; East Asian: n = 22; South Asian: n = 14; Black: n = 7; Jewish: n = 6; Middle-Eastern: n = 6; South East Asian: n = 5; Multi-Racial: n = 4; Hispanic: n = 3) recruited from a large, suburban university in Canada for an online study on "understanding identity." Ethical approval was obtained from the institutional research ethics committee. After completing informed consent, participants read definitions of identity, conflict, and identity conflict (see Appendix A) before selecting which of a pair of identities from a randomized list had created their most troubling conflict. The list included 11 pairs of identities derived from social psychological, developmental, and sociological literatures (e.g., ethnic-national identities, teenager-adult identities, studentemployee identities; family-friend identities), an "other" option where participants could list any other pair of identities that resulted in conflict for them, and a "no conflict" option to indicate that an identity conflict had never been experienced. Eighty-nine participants (77%) reported experiencing an identity conflict. Participants who reported experiencing conflict completed openended questions where they described what the conflict entailed. Participants reporting an on-going conflict then described what they did to manage their experience of conflict. Participants who reported a past conflict described what they had done to resolve their conflict<sup>1</sup> . Responses were content coded by three independent coders (SC, JJ, and MM). Consistency of coding was computed between each pair of coders using Cohen's Kappa

<sup>1</sup>Participants were also asked: when and/or where it occurred, who else was involved, how they felt when they experienced this conflict. For brevity, these descriptive data are not reported here.

where the percentage of agreement between pairs of coders ranged from 0.51 to 0.75. According to Landis and Koch (1977), this indicates moderate to substantial agreement between coders. Discrepancies were resolved by a final decision made by the first author (JJ).

Participants then rated their experience of conflict on three aspects: magnitude, stress, and growth. The measure of magnitude, the amount of conflict experience in general, was adapted from an existing measure of bi-cultural identity integration (BII: Benet-Martínez and Haritatos, 2005; Cheng and Lee, 2013). Items were re-worded to reflect identities in general (e.g., I feel conflicted between these two identities). The five item scale was unreliable (Cronbach's α = 0.63) and suggested improved reliability with the removal of one item: I feel as though these two identities are combined. This left us with a 4-item scale (Cronbach's α = 0.71; M = 5.13, SD = 1.08). Participants also rated four items to assess stress (i.e., how severe, difficult, challenging, troubling is/was this conflict; α = 0.91; M = 5.07, SD = 1.32), and three items to assess growth (i.e., how much have you learned, changed, and grown from experiencing this conflict; α = 0.83; M = 5.16, SD = 1.22). All ratings used scales from 1 (Strongly Disagree) to 7 (Strongly Agree) with higher scores indicating conflicts of greater magnitude, stress, and growth. Participants also completed demographic information (e.g., age, gender, and ethnicity) before receiving an online debriefing and our thanks.

## Results

Descriptive statistics (i.e., frequencies, means and standard deviations, correlations) were used to assess the selections, open-ended responses, and ratings provided by participants. Frequencies may be greater than a count of 89 where participants gave multiple responses.

#### Descriptions of Conflict

#### **Types of identity conflict**

The most frequently selected conflicts pertained to role identities (e.g., student-employee, student-athlete, employeeathlete; n = 37), transitional identities (i.e., teenager-adult; n = 20) and relational identities (e.g., couple-family, friendfamily, couple-friend; n = 17). Social identities did not figure prominently, with bicultural identity conflicts accounting for less than 6% of the conflicts experienced (i.e., ethnicity-nationality: n = 5), and only one report of another type of social identity conflict (i.e., sexuality-religion). Mixed conflicts, combinations of different types of identities such as relational-social (i.e., friend-racial/ethnic; couple-racial/ethnic; n = 5) and role-social (i.e., gender-employee; n = 2), were also relatively low in frequency. Two participants indicated that they had experienced another type of identity conflict. These included a role-relational conflict (i.e., student-friend) and what can be characterized as an existential conflict (i.e., religion and life itself).

#### **Reasons for identity conflict**

In describing what the conflict entailed or why it occurred, 69% of participants who experienced conflict described it as comprising one issue. Thirty-one percent of participants described their conflict as comprising two or more issues. Overall, these issues took the form of time management concerns (n = 28), trying to balance between two identities (n = 16), managing others' expectations or beliefs (n = 16), disapproval of one's relationships (e.g., romantic relationships, friendships) by others (n = 16), transitional concerns (e.g., different life stages; n = 12), fighting with (or taking on) responsibilities (n = 11), managing personal expectations or beliefs (n = 6), financial concerns (n = 3), or other concerns (e.g., being treated in a disrespectful way by co-workers, realization that friends are changing, trying to figure out who one is; n = 4).

#### **Managing identity conflict**

In describing what was done when conflict was experienced, 79% of participants who experienced conflict indicated that one response was elicited. The remaining twenty-one percent of participants who experienced conflict reported that two or more responses were elicited. Most often, participants tried to balance their identities (n = 20). Participants also report attempting to ignore or avoid the conflict (n = 12), engaging in other positive strategies (e.g., prayer, writing, relaxing; n = 12), focusing on (or choosing) one identity over the other (n = 10), breakingdown or becoming emotional (n = 9), rationalizing the conflict (n = 8), doing nothing (n = 8), accepting the conflict (i.e., 'just deal' or 'move on'; n = 7), or discussing the conflict with others (n = 6). Participants also reported engaging in other strategies (e.g., confronting others, adapting, becoming pensive, lying; n = 14) in an effort to manage identity conflict.

#### **Resolving identity conflict**

Twenty-one of the 89 participants who experienced conflict described this conflict as having been resolved. Thirteen of these participants reported engaging in one strategy, whereas eight participants reported engaging in two or more strategies to resolve this conflict. These responses differed from how participants reported managing their conflicts with the most common way to resolve conflict involving personal change. Individuals reported changing their attitudes or perspective (n = 7) or changing their lifestyle (n = 7) in order to resolve the conflict. Participants also reported re-evaluating the situation (n = 4), discussing the conflict (n = 4), introspection (n = 2), suppressing one identity (n = 1), or other methods (i.e., stopped trying to prove things to others; n = 1) as things they had done to resolve the conflict.

#### Experience of Conflict

For participants currently experiencing conflict, the magnitude, stress and growth associated with their identity conflict were rated as moderately high. Furthermore, Pearson's correlation indicated that the magnitude of the conflict was positively related to stress from conflict and showed a marginal positive relationship with growth from the conflict (see top diagonal, **Table 1**). For participants who had resolved their conflict, the magnitude and perceived stress associated with the conflict were rated as moderate and the growth associated with their conflict was rated as moderately high. There were no relationships between the magnitude, stress, and growth associated with

TABLE 1 | Means, standard deviations and correlations for the experience of identity conflict (current conflict – top diagonal, n = 68; resolved conflict – bottom diagonal, n = 21; Study 1).


<sup>I</sup>p < 0.07, <sup>∗</sup>p < 0.05; ∗∗p < 0.01, ∗∗∗p < 0.001.

conflict for these individuals (see bottom diagonal, **Table 1**; also see **Table 1** for means and standard deviations).

# Discussion

Study 1 provides initial insights into the frequency, experience and management of conflict between multiple identities. First, our findings indicate that conflicts between role, transitional, and relational identities were most frequently reported relative to other types of identity conflicts (i.e., mixed types, social categories). This makes sense given that roles, relationships, and life-stages reflect prominent self-aspects in the minds of young adults (Arnett, 2000; McConnell, 2011). Surprisingly, conflicts between social categories or those which included social categories were not frequently reported although research on biculturals, multi-racial identities, and women in the sciences indicates that social categories such as race, ethnicity, nationality, and gender are often implicated in the experience of conflict (e.g., Settles, 2004; Benet-Martínez and Haritatos, 2005; Shih and Sanchez, 2005; Brook et al., 2008). It could be that individuals do not spontaneously think about themselves in these terms, leading social identity conflicts to be under-reported relative to conflicts involving other types of identities. Second, in rating the experience of conflict, on-going conflicts were rated moderately high on magnitude, stress, and growth. Here, magnitude was positively related to stress and magnitude also showed a marginal positive relationship with growth. These correlations suggest that experiencing conflict may be associated with adaptation, which can contribute to well-being (Tedeschi and Calhoun, 2004). Resolved conflicts were rated as moderate in magnitude and stress and moderately high on growth, but here stress and growth were unrelated to magnitude or to each other. However, these latter findings should be interpreted with caution given the relatively small numbers of individuals who reported resolved conflicts. Third, we found qualitative evidence for the proposed conflict management strategies. In particular, individuals try to balance both identities, which maps onto the idea of reconciliation, they tend to ignore or avoid the identities in conflict, which maps onto the idea of retreat, and they report focusing on one identity, which maps onto the idea of realignment. However, without a measure of these conflict management strategies, it is difficult to determine their use in general, or specifically when faced with particular types of identity conflicts. Lastly, personal change was implicated in the resolution of identity conflict. Although personal change as a strategy was not anticipated, its mention exclusively among individuals who had already resolved their conflicts likely reflects the end of the conflict management process. To address questions of conflict prominence and whether there might be differences in the experience and management of conflict for different types of multiple identities, our next study took a closer look at specific identity conflicts and sought to develop and test a measure of identity conflict management.

# STUDY 2: DO THE TYPES OF IDENTITIES INFLUENCE THE EXPERIENCE AND MANAGEMENT OF CONFLICT?

The aim of this study was to examine whether the types of multiple identities influence the experience and management of conflict. To address the prominence of different types of identity conflict, participants were randomly assigned to one of three conditions where they were explicitly asked to describe a conflict between multiple role (i.e., student-employee), relational (i.e., family-friend), or social (i.e., ethnic-national) identities. Social identity conflict was used rather than a transitional identity conflict for two reasons: First, due to its developmental nature, transitional identity conflicts reflect permanent changes that have clear indicators of onset (e.g., age, physical changes) and expected behavior(s). Transitional conflicts may therefore differ from other types of identity conflicts in how they are experienced and managed. Second, as social identity conflicts do not appear to be spontaneously generated they were included to determine if participants might report experiencing them when explicitly asked. Bi-cultural conflicts were used to represent social identity conflicts because they reflect a real concern for many young adults in multicultural societies (see Benet-Martínez et al., 2002; Benet-Martínez and Haritatos, 2005). As in Study 1, the experience of conflict was assessed using measures of magnitude, stress, and growth. Conflict management was assessed using a new scale developed for the present study. In light of Study 1 where ratings of magnitude, stress and growth were moderate to high and all participants reported attempts to manage conflict, we expected that conflict magnitude, stress and growth would be positively related the conflict management strategies (Hypothesis 1). We also considered whether the types of identities might influence the experience and management of conflict. Given that people's role, relational, and social identities cluster along different dimensions (Deaux et al., 1995), it is plausible that the types of identities might also be associated different conflict experiences and management strategies. However, as there is no prior theoretical work or empirical evidence to suggest the direction of these effects specific hypotheses were not made about how these variables might be related to each other.

# Method

#### Participants, Procedure, and Measures

Participants were 366 undergraduates recruited for an online study on 'understanding identity' from a large suburban university in Canada (Age: M = 20.75; SD = 3.50; Gender: 64% Female; Race/Ethnicity: White: n = 151; South Asian: n = 61;

East Asian: n = 66; Multi-Racial: n = 20; Middle-Eastern: n = 12; Black: n = 24; Hispanic: n = 12; Native-Canadian: n = 1; Missing: n = 3). The study employed a 3 Type of Identities (Role, Relational, and Social) single factor between-subjects design. Participants received either course credit or \$5 Canadian dollars for completion of this study. Ethical approval was obtained through the institutional research ethics committee.

After completing informed consent, participants provided details about different aspects of themselves (i.e., age, gender, student, employee status, friend, family, race, ethnicity, religion, and relationship status) and rated the importance of these aspects (e.g., Being a student is an important part of my identity). Participants were then randomly assigned to describe a conflict between role (i.e., student-employee), relational (i.e., friend-family) or social (i.e., ethnicity-nationality) identities. If participants indicated that they had not experienced the identity conflict to which they had been assigned, they had the option of selecting an identity conflict from a list of conflicts, with one of these options allowing them to specify any other type of identity conflict or that they had never experienced conflict. These options were the same as those used in Study 1. One-hundred and fiftyone participants indicated that they experienced the conflict to which they had been assigned (role: n = 49; relational: n = 53; social: n = 49). Forty-six participants (12.6%) indicated that they had never experienced an identity conflict and were directed to the end of the questionnaire. One-hundred and sixty-nine participants indicated they had experienced a conflict other than the one to which they had been assigned. Of these participants, fifty-five subsequently selected one of the target role, relational, or social conflicts (student-employee: n = 26; friend-family: n = 18; ethnicity-nationality: n = 11). The remaining participants selected another type of conflict: relationships (couple-family: n = 16; friend-couple: n = 12), social (multi-racial: n = 2), transitional (teenager-adult: n = 37), mixed types of conflicts (i.e., role-relational – student-couple: n = 22, student-friend: n = 1; social-role – gender-student: n = 2; social-relational – couplerace: n = 8; friend-race: n = 4; couple-religion: n = 1), or wrote in their own conflicts (n = 10; e.g., the self and one identity; tri-identity conflicts).

Findings are presented for the 205 participants who reported experiencing the target conflicts between roles (i.e., student-employee), relationships (i.e., friend-family), and social categories (i.e., culture-nationality). Fifty-seven of these conflicts were described as on-going whereas 148 were described as recent or past. One person who had been assigned to the culturenationality identity condition did not complete the questionnaire and was omitted from further analyses. After describing their conflict, participants rated its experience on three dimensions from 1 Strongly Disagree to 5 Strongly Agree: magnitude (see Study 1: four items; Cronbach's α = 0.81), stress (seven items; i.e., how severe/difficult/challenging/troubling/bothersome/demanding/ stressful is this conflict; Cronbach's α = 0.94), and growth (see Study 1: three items; Cronbach's α = 0.84) as well as rating its management.

To assess conflict management a new measure was developed on the basis of the previous literature (e.g., Baumeister et al., 1985; Roccas and Brewer, 2002; Burke, 2003), participants' responses from Study 1, and in consultation with other identity researchers. To ensure representativeness across different types of conflicts, responses from all participants who reported an identity conflict (i.e., n = 319) were used in the factor analysis of an initial set of 32 items (principal axis, direct oblimin rotation, pattern matrix interpreted using a 0.50 loading cut-off). Fifteen items were distributed across the first five factors: Retreat, the avoidance and compartmentalization of conflicting identities (16.13% of the variance; four items; Cronbach'sα = 0.81), Reconciliation, trying to balance between, or integrate, conflicting identities (12.95% of the variance; four items; Cronbach'sα = 0.73), Realignment, choosing one identity over the other 11.62% of the variance; three items; Cronbach'sα = 0.74), Reflection, using the identity accommodates others or the situation (12.54% of the variance; two items; r = 0.55, p < 0.001), and Relinquishment, giving up one or both identities (4.96% of the variance; two items; r = 0.28, p < 0.001; see Appendix B for items and factor loadings)<sup>2</sup> .

# Results

Descriptive statistics (i.e., means and standard deviations, correlations) are reported in **Table 2**.

#### Relationships between the Experience and Management of Identity Conflicts

Partial correlations were used to examine the relationships between the experience and management of conflict controlling for conflict occurrence (on-going, recent/past), conflict choice (assigned, selected) and gender (male, female; see **Table 2**). In terms of the experience of conflict, magnitude, stress, and growth were all positively related to each other. In terms of the conflict management strategies, retreat was positively related to realignment and reflection. Reflection was positively related to reconciliation and realignment. Relinquishment was not related to any of the other management strategies. In support of Hypothesis 1, the experience of conflict was positively related to the management of conflict: As the magnitude of conflict increased, the use of retreat, realignment, and reflection strategies increased and reconciliation marginally increased. Similarly, as stress increased, the use of retreat, realignment, reflection strategies increased and reconciliation marginally increased. Finally, as growth increased, the use of the retreat, reconciliation, and reflection strategies increased.

#### Experiencing Conflict between Multiple Identities

A 3 Type of Identities (Role, Relational, and Social) single-factor between-subjects ANCOVA with conflict occurrence, conflict choice and gender as covariates was used to test whether the experience of conflict differed as a function of the identities in conflict. All pairwise comparisons were conducted with Least Significant Difference (LSD).

<sup>2</sup>Three additional factors emerged. However, as these factors did not comprise at least 2 items loading at 0.50 (factors 6 and 7 had no items loading at 0.5 or higher, factor 8 had 1 item loading at 0.5), they were not included in the present analysis. As principal axis factoring (PAF) was used, the variance accounted for was computed by re-running the PAF with the number of selected items (15), and dividing the rotated sums of squared loadings by the number of items in the solution.



<sup>I</sup>p < 0.07, <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Controlling for conflict occurrence, conflict choice, and gender.

#### **Conflict magnitude**

The magnitude of conflict did not differ as a function of the types of identities in conflict, F(2,199) = 1.93, p = 0.148. Participants experienced conflicts of similar magnitude regardless of the types of identities in conflict (Role: M = 2.84, SE = 0.098, Relational: M = 2.71, SE = 0.099, Social: M = 3.00, SE = 0.108; all ps > 0.051).

#### **Perceived stress**

Stress did not differ as a function of the types of identities in conflict, F(2,199) = 1.02, p = 0.364. Participants experienced similar levels of stress regardless of the types of identities in conflict (Role: M = 3.11, SE = 0.101, Relational: M = 2.90, SE = 0.103, Social: M = 3.00, SE = 0.112; all ps > 0.156).

#### **Perceived growth**

Growth did not differ as a function of the types of identities in conflict, F(2,199) = 2.46, p = 0.09. Although the overall effect was not significant, pairwise comparisons suggested that role conflicts were associated with less growth (M = 3.26, SE = 0.100) than were social conflicts (M = 3.57, SE = 0.110; p = 0.038). Relational conflicts (M = 3.50, SE = 0.101) did not differ in magnitude relative to role or social conflicts (both ps > 0.096).

#### Managing Conflict between Multiple Identities

A 3 Type of Identities (Role, Relational, and Social) × 4 Conflict Management (Reconciliation, Realignment, Retreat, and Reflection) mixed ANCOVA, with type of identities as the between factor, management strategy as the within factor, and conflict occurrence, conflict choice, and gender as covariates was used to test whether the management of conflict differed as a function of the identities in conflict. Relinquishment was not included as it was unrelated to the other conflict management strategies. All pairwise comparisons were conducted with LSD.

Although there was no main effect for type of identities, F(2,199) = 1.96, p = 0.14, a main effect for management strategy emerged, F(3,597) = 3.12, p = 0.025, η 2 <sup>p</sup> = 0.015: The use of reflection did not differ from the use of realignment (p = 0.134). The use of all other management strategies were significantly different from each other. Overall, participants used reconciliation more relative to all other management strategies (all ps < 0.006) and retreat less relative to all other management strategies (all ps < 0.002). This effect was qualified by the significant Type of Identities by Management Strategy interaction, F(6,597) = 2.33, p = 0.031, η 2 <sup>p</sup> = 0.023. To explore this interaction, four single factor between-subjects ANCOVAs, using each management strategy as a dependent variable, Type of Identities as the between subjects factor, and conflict occurrence, conflict choice, and gender as covariates, were conducted.

#### **Retreat**

There was no main effect for the Type of Identities on retreat, F(2,199) = 1.77, p = 0.17, η 2 <sup>p</sup> = 0.017. Participants were equally likely to engage in this strategy, regardless of the types of identities in conflict (Role: M = 3.12, SE = 0.095, Relational: M = 2.90, SE = 0.096, Social: M = 2.89, SE = 0.105; all ps > 0.102).

#### **Reconciliation**

There was no main effect for the Type of Identities on reconciliation, F(2,199) = 0.238, p = 0.788, η 2 <sup>p</sup> = 0.002. Participants were equally likely to engage in this strategy, regardless of the types of identities in conflict (Role: M = 3.52, SE = 0.079, Relational: M = 3.50, SE = 0.080, Social: M = 3.58, SE = 0.087; all ps > 0.498).

#### **Realignment**

A main effect for Type of Identities was found on realignment, F(2,199) = 5.79, p = 0.004, η 2 <sup>p</sup> = 0.055: Participants who described conflicts between role identities reported more realignment (M = 3.42, SE = 0.099), relative to participants who described conflicts between relational identities (M = 2.94, SE = 0.101; p = 0.001), and social identities (M = 3.10, SE = 0.110; p = 0.034). There were no differences in the use of realignment for conflicts between relational and social identities (p = 0.274).

#### **Reflection**

There was no main effect for the Type of Identities on reflection, F(2,199) = 0.549, p = 0.578, η 2 <sup>p</sup> = 0.005. Participants were equally likely to engage in this strategy, regardless of the types of identities in conflict (Role: M = 3.20, SE = 0.11, Relational: M = 3.24, SE = 0.111, Social: M = 3.37, SE = 0.121; all ps > 0.313).

We also looked at relinquishment as a separate dependent variable. Here a main effect for Type of Identities was found, F(2,199) = 5.47, p = 0.005, η 2 <sup>p</sup> = 0.052: Participants who described conflicts between social identities reported less relinquishment (M = 3.10, SE = 0.10) relative to participants who described conflicts between role identities (M = 3.39, SE = 0.09; p = 0.033) or relational identities (M = 3.54, SE = 0.092;

p = 0.001). There were no differences in the use of relinquishment for conflicts between role and relational identities (p = 0.255).

# Discussion

Study 2 provides important information about the experience and management of conflicts between different types of multiple identities. First, in considering our new measure of conflict management, we found that reconciliation was used most, and retreat was used least, relative to all other strategies. This suggests that people generally try to balance or integrate their multiple identities to manage identity conflict. Although relinquishment also emerged as a factor, as it was unrelated to the other management strategies it could be that relinquishment reflects a post-conflict outcome rather than a way of (actively) managing conflict. Some evidence to support this idea can be seen in additional analyses looking at individuals who reported that their conflicts were in the past (i.e., 7+ months ago). Here relinquishment was positively related to stress and growth and showed a marginal positive relationship with reconciliation. These relationships were not found for individuals who reported on-going or recent (i.e., 6 or less months) conflicts. Second, and in support of Hypothesis 1, the experience of conflict was positively related to its management: Although conflict magnitude, stress and growth were all positively related to retreat, realignment, and reflection, only growth was positively related to reconciliation. It could be that when contending with identity conflicts, individuals might gravitate toward strategies that enable them to avoid the conflicting identities or focus on one of these identities over the other. Yet when individuals have had some time to live with their conflicts, as might be suggested by experiencing growth, the extent to which they balance their conflicting identities might also increase. Third, the experience of conflict did not differ as a function of the types of identities in conflict. However, we did find evidence for the differential use of management strategies as a function of the types of identities in conflict. Here role identities were associated with greater use of realignment relative to relational or social identity conflicts. Reconciliation, reflection and retreat were not differentially used. Taken together, we have preliminary evidence that the types of identities matter for the management, but not necessarily for the experience, of identity conflict, with multiple roles lending themselves to more realignment relative to relationships or social categories. However, we do not know what it is about roles that might allow for the use of realignment relative to other types of identities. In addition to establishing the stability of our measure of identity conflict management, our subsequent study examined whether the perceived characteristics of individuals' identities in general might be associated with differences in the experience and management of conflict.

# STUDY 3: IDENTITY CHARACTERISTICS AND THE EXPERIENCE AND MANAGEMENT OF CONFLICT

The aim of Study 3 was twofold. In addition to examining the experience and management of conflict in another sample, we first sought to examine the stability of the measure of conflict management. Finding the same factors and similar reliabilities would go some way to confirming that these strategies reflect a meaningful measurement in the context of identity conflicts. Second, we sought to examine whether the characteristics of role, relational and social identities in general might be associated with the experience and management of conflict. We already know that the sources of identities (i.e., social categories, social groups, relationships, roles) are perceived as distinct from each other and differ on key dimensions including flexibility and importance (Deaux et al., 1995; McConnell and Strain, 2007; McConnell, 2011). And recent work has shown that the perceived importance and valence associated with multiple identities matters for the experience of conflict (Settles, 2004; Brook et al., 2008). When considered along with our initial evidence that role identities differ from relational and social identities in the use of realignment it could be that the general perceptions of different types of identities shape the specific strategies used in management of conflict. However, at least based on our initial findings, there is little evidence to suggest that these perceptions might influence the experience of conflict.

To examine these relationships, we had participants rate the perceived flexibility, valence, and importance of their identities in general before being randomly assigned to one of three conditions where they were explicitly asked to describe a conflict between multiple role, relational, or social identities, before rating their experience of conflict (magnitude, stress, and growth) and their use of conflict management strategies (retreat, reconciliation, realignment, and reflection). Given the findings from Study 2 we expected a positive relationship between the experience and management of conflict, with conflict magnitude, stress, and growth showing positive relationships with each of the conflict management strategies (Hypothesis 1). Given the findings from Study 2 we also expected that the types of identities in conflict would influence the management of conflict (i.e., reconciliation, realignment, retreat, and reflection) but not the experience of conflict (i.e., magnitude, stress, and growth). In particular, we expected that conflicts between multiple role identities would be associated with more realignment relative to conflicts between multiple relational or social identities (Hypothesis 2).

In extension of Study 2, we considered whether the characteristics of identities in general were associated with the experience and management of conflict. Given the findings from Study 2 and previous research which suggested that identities differ on dimensions related to flexibility, importance, and valence, and that these characteristics are related to the experience of conflict (e.g., Deaux et al., 1995; Settles, 2004; Brook et al., 2008; McConnell, 2011), we expected that the general perception that one's identities are more flexible and less important would be associated with conflicts of less magnitude, stress, and growth (Hypotheses 3a) and the use of realignment and retreat (Hypothesis 3b) particularly when these characteristics were associated with roles rather than relationships or social categories.

# Method

#### Participants, Procedure, and Measures

fpsyg-08-01732 October 4, 2017 Time: 16:39 # 10

Participants were 300 adults recruited for an online study on 'understanding identity' via Prolific Academic, a crowd-sourcing participant recruitment website. Seven participants identified as duplicates (the same IP address and location data for more than one entry) and three participants who selected but did not write about a conflict were excluded leaving us with a sample of 290 (Age: M = 24.76; SD = 6.41; Gender: 46% Female, 2% Agender; Race/Ethnicity: White: n = 203; East Asian: n = 27; Black: n = 14; Hispanic: n = 10; Multi-Racial: n = 9; South Asian: n = 5; Middle-Eastern: n = 1; Missing: n = 21; Nationality: American: n = 144, British: n = 88, Canadian: n = 18; Other: n = 40; Student: Yes: n = 219, No: n = 71; Employed: Yes: n = 170, No: n = 120). The study employed a 3 Type of Identities (Role, Relational, and Social) single factor betweensubjects design. Participants received two British pounds sterling for completion of this study. Ethical approval was obtained through the institutional research ethics committee.

After completing informed consent, participants provided demographic information (e.g., age, gender, and ethnicity) as well as details about different aspects of themselves (i.e., student, employee, friend, family, relationship, race, ethnicity, nationality, and religion) and the perceived flexibility (e.g., My student identity is: Very Inflexible – Very Flexible), importance (two items each: e.g., I identify with my family, My family is an important part of how I see myself) and valence (e.g., My ethnic identity is: Very Negative – Very Positive) of these identities.<sup>3</sup> Participants then read a description of an identity conflict (see Appendix A), indicated whether they had read the description (i.e., Y/N) and indicated whether, upon reading the description, "I understand the definition of identity conflict and how it might apply to me" on a scale from 1 Strongly Disagree to 5 Strongly Agree (M = 4.48, SD = 0.53). As in Study 2, participants were then randomly assigned to describe a conflict between their role (i.e., student-employee), relational (i.e., friend-family), or social (i.e., ethnicity-nationality) identities. If participants indicated that they had not experienced the identity conflict to which they had been assigned, they had the option of selecting an identity conflict from a list of conflicts, with one of these options allowing them to specify any other type of identity conflict or that they had never experienced conflict. One-hundred and sixtynine participants indicated that they experienced the conflict to which they had been assigned (role: n = 65; relational: n = 73; social: n = 31). Ninety-nine participants indicated that they had experienced a conflict other than the one to which they had been assigned. Of these participants, 19 selected one of the main role or relational conflicts of interest (student-employee: n = 11; friend-family: n = 9). No participants selected the main social identity conflict (i.e., ethnicity-nationality). The remaining 80 participants selected another type of conflict: roles (studentrecreational: n = 16), relationships (couple-family: n = 12), social (two races/ethnicities: n = 7; sexuality-religion: n = 3; religion-nationality: n = 1), transitional (two-life stages: n = 17), mixed types (social-role: gender-student: n = 6; social-relational: couple-race/ethnicity: n = 3; friend-race/ethnicity: n = 4), and other (n = 11; e.g., conflicts involving the self and one identity; conflicts between three identities). Twenty-two participants indicated that they never experienced conflict between any of their identities. Findings are presented for 188 participants who were assigned to, or subsequently selected, conflicts between roles (i.e., student-employee), relationships (i.e., friend-family), and social categories (i.e., ethnicity-nationality). Seventy-seven of these conflicts were current whereas 111 were resolved. Findings for the individuals who reported other types of identity conflicts (n = 80) are not reported.

After describing the conflict, participants rated their experience of conflict on three dimensions: magnitude (see Study 1: four items; Cronbach's α = 0.82, stress (see Study 2: seven items, Cronbach's α = 0.88), and growth (see Study 1: three items; Cronbach's α = 0.85). To assess the management of identity conflict, the 13 correlated items from Study 2 were used. Factor analysis (principal axis, direct oblimin rotation, pattern matrix interpreted using a 0.40 loading cut-off) suggested a four-factor solution: Reconciliation (19.84% of the variance; four items; Cronbach'sα = 0.77), Realignment (16.25% of the variance; three items; Cronbach'sα = 0.72), Retreat (16.09% of the variance; four items; Cronbach'sα = 0.76), and Reflection (12.02% of the variance; two items; r = 0.59, p < 0.001; see Appendix C for items and factor loadings). All ratings were completed on scales from 1 Strongly Disagree to 5 Strongly Agree.

For the identity characteristics, composites were created for student importance [two items; r(145) = 0.57, p < 0.001], employee importance [two items; r(116) = 0.75, p < 0.001], family importance [two items; r(188) = 0.84, p < 0.001], friend importance [two items; r(185) = 0.61, p < 0.001], ethnic importance [two items; r(186) = 0.70, p < 0.001], and national importance [two items; r(187) = 0.73, p < 0.001]. In line with the key conflicts of interest, these were then averaged to created composites for role importance [student, employee; r(86) = 0.25, p = 0.021], relational importance [family, friends; r(188) = 0.27, p < 0.001], social importance [ethnicity, nationality; r(188) = 0.52, p < 0.001], role valence [student, job; r(85) = 0.23, p = 0.033], relational valence [family, friends; r(187) = 0.22, p = 0.003], social valence [ethnicity, nationality; r(187) = 0.50, p < 0.001], relational flexibility [family, friends; r(187) = 0.19, p = 0.010] and social flexibility [ethnicity, nationality; r(187) = 0.35, p < 0.001]. As student flexibility and employee flexibility were unrelated to each other [r(146) = 0.014, p = 0.898], they were used as individual items.

## Results

#### Relationships between Characteristics and the Experience and Management of Conflict

Partial correlations were used to examine the relationships between identity characteristics, and the experience and management of conflict controlling for conflict occurrence, conflict choice, and gender (see **Table 3** for means and standard deviations; see **Table 4** for correlations). As in Study 1, conflict

<sup>3</sup>As not all participants responded to all aspects (e.g., unemployed participants did not complete job/employee ratings) the ns differ across the subsequent comparisons.



magnitude, stress, and growth were all positively related to each other. In support of Hypothesis 1, and similar to the results of Study 2, the experience of conflict was positively related to the management of conflict: As the magnitude of conflict increased, the use of retreat and reconciliation strategies increased. As stress increased, the use of retreat and realignment strategies increased. However, in contrast to Study 2, growth was unrelated to the conflict management strategies.

In partial support of Hypothesis 3a, identity characteristics were related to the experience of conflict. However, these relationships were in the opposite direction to our hypotheses: Student identity flexibility, a role, was positively related to stress and growth and marginally positively related to conflict magnitude. Unexpectedly, we also found that social identity flexibility was positively related to growth and that social identity importance was positively related to conflict magnitude, stress, and growth. In terms of valence, role identity valence was positively related to stress, relational identity valence was positively related to conflict magnitude, and social identity valence was positively related to conflict magnitude, stress, and growth. In partial support of Hypothesis 3b, student identity flexibility was positively related to retreat and realignment and relational identity flexibility was positively related to retreat. For importance, role, relational, and social identity importance were all positively related to reconciliation, with relational identity importance also showing a positive relationship with reflection. In terms of valence relational identity valence was positively related to reconciliation and reflection but negatively related to realignment. Role identity valence was marginally associated with reconciliation.

#### Experiencing Conflict between Multiple Identities

A 3 Type of Identities (Role, Relational, and Social) single-factor between-subjects ANCOVA with conflict occurrence, conflict


Controlling for conflict occurrence, conflict choice, and gender. <sup>I</sup>p < 0.07, <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Controlling for conflict occurrence, conflict choice, and gender.

choice, and gender as covariates was used to test whether the experience of conflict differed as a function of the identities in conflict. All pairwise comparisons were conducted with LSD.

#### **Conflict magnitude**

Similarly to Study 2, conflict magnitude did not differ as a function of the types of identities in conflict, F(2,180) = 2.55, p = 0.081. Although the overall effect was not significant, pairwise comparisons suggested that role conflicts were of greater magnitude (M = 3.25, SE = 0.099) than were social conflicts (M = 2.84, SE = 0.158; p = 0.030). Relational conflicts (M = 3.06, SE = 0.097) did not differ in magnitude relative to role or social conflicts (both ps > 0.177).

#### **Perceived stress**

Contrary to Study 2, stress differed as a function of the types of identities in conflict, F(2,181) = 6.95, p = 0.001, η 2 <sup>p</sup> = 0.071: Role conflicts were associated with more stress (M = 3.70, SE = 0.092) relative to relational conflicts (M = 3.32, SE = 0.09; p = 0.004) or social conflicts (M = 3.14, SE = 0.15; p = 0.001). The stress associated with relational and social identity conflicts were not different from each other (p = 0.30).

#### **Perceived growth**

Similarly to Study 2, growth did not differ as a function of the types of identities in conflict, F(2,181) = 0.62, p = 0.54. Participants experienced similar levels of growth regardless of the types of identities in conflict (Role: M = 3.62, SE = 0.11, Relational: M = 3.48, SE = 0.111, Social: M = 3.68, SE = 0.172; all ps > 0.339).

#### Managing Conflict between Multiple Identities

A 3 Type of Identities (Role, Relational, and Social) × 3 Conflict Management (Reconciliation, Realignment, Retreat, and Reflection) mixed ANCOVA, with type of identities as the between factor, management strategy as the within factor, and conflict occurrence, conflict choice, and gender as covariates was used to test whether the management of conflict differed as a function of the identities in conflict. All pairwise comparisons were conducted with LSD.

Although the main effects for management strategies and types of identities were not statistically significant, both Fs < 0.76, both ps > 0.52, in support of Hypothesis 2, and in line with Study 2, the Type of Identities by Management Strategy interaction was significant, F(6,543) = 2.18, p = 0.044, η 2 <sup>p</sup> = 0.023. To explore this interaction, four single factor between-subjects ANCOVAs, using each management strategy as a dependent variable, Type of Identities as the between subjects factor, and conflict occurrence, conflict choice and gender as covariates, were conducted.

#### **Retreat**

In contrast to Study 2, a main effect for Type of Identities was found on retreat, F(2,181) = 5.28, p = 0.006, η 2 <sup>p</sup> = 0.055: Participants who described conflicts between relational identities engaged in more retreat (M = 3.40, SE = 0.10), relative to participants who described conflicts between role identities (M = 3.07, SE = 0.10; p = 0.021), and social identities (M = 2.83, SE = 0.16; p = 0.003). There were no differences in the use of retreat for conflicts between role and social identities (p = 0.224).

#### **Reconciliation**

Similarly to Study 2, there was no main effect for the Type of Identities on reconciliation, F(2,181) = 0.40, p = 0.67, η 2 <sup>p</sup> = 0.00. Participants were equally likely to engage in this strategy, regardless of the types of identities in conflict (Role: M = 3.59, SE = 0.093, Relational: M = 3.59, SE = 0.090, Social: M = 3.74, SE = 0.148; all ps > 0.399).

#### **Realignment**

In contrast to Study 2, there was no main effect for the Type of Identities on realignment, F(2,181) = 1.27, p = 0.28, η 2 <sup>p</sup> = 0.014. Participants were equally likely to engage in this strategy, regardless of the types of identities in conflict (Role: M = 3.59, SE = 0.104, Relational: M = 3.40, SE = 0.102, Social: M = 3.32, SE = 0.166; all ps > 0.177).

#### **Reflection**

Similarly to Study 2, there was no main effect for the Type of Identities on reflection, F(2,181) = 0.21, p = 0.81, η 2 <sup>p</sup> = 0.00. Participants were equally likely to engage in this strategy, regardless of the types of identities in conflict (Role: M = 3.31, SE = 0.114, Relational: M = 3.38, SE = 0.111, Social: M = 3.44, SE = 0.182; all ps > 0.545).

# Discussion

Study 3 provides evidence for the stability of our measures of conflict experience and management as well as important insights into the experience and management of conflict, and their relationships with identity characteristics. With a markedly less diverse and slightly older sample relative to Studies 1 and 2, we found similar reliabilities for the measures of conflict experience and these measures were all positively related to each other. This suggests that magnitude, stress, and growth are useful measures for capturing the experience of conflict between multiple identities. Critically, we also found the same factors and similar reliabilities for the conflict management strategies. This suggests that retreat, reconciliation, realignment, and reflection are strategies that people use to manage conflicts between multiple identities. However, the present sample differed in the distribution identity conflicts, with almost 50% fewer bi-cultural conflicts being described relative to Study 2. Unsurprisingly, with these differences in the distribution of conflicts, we found both consistent and inconsistent results for the experience and management of conflicts between different types of multiple identities<sup>4</sup> .

First, in support of Hypothesis 1, the experience of conflict was positively related to its management. Conflict magnitude was positively related to retreat and reconciliation, and stress was positively related to retreat and realignment. However,

<sup>4</sup>Ethnic minorities might be more likely to experience and manage bi-cultural conflicts, which could explain why this conflict less prominent in the predominantly White sample of Study 3. In fact, several of the White respondents assigned to this condition explained the lack of conflict in similar ways before selecting another conflict (e.g., "I am white and British as are most British people. There is no conflict," "My ethnic and national identities have never conflicted. America is synonymous with my race"). Accordingly, we re-analyzed the data for Studies 2 and 3 controlling for whether participants were White or non-White. The pattern of findings was almost exactly the same with one exception: In Study 2, the main effect for conflict management strategies became marginal.

growth was unrelated to any management strategies. While these relationships still support the idea that individuals gravitate toward a strategy that enables them to choose one identity or avoid their conflicting identities, especially when they are associated with stress, it also suggests that as perceptions of the discrepancy between multiple identities increase (i.e., increases in magnitude), there are increased attempts to integrate them. Despite some different relationships relative to Study 2 (i.e., conflict magnitude was unrelated to realignment or reflection, growth was unrelated to any of the management strategies), the patterns from the present study still support the overall assertion that the experience of conflict is positively related to the management of conflict.

Second, we unexpectedly found that the experience of conflict differed as a function of the types of identities in conflict. Here, conflicts between role identities were associated with more stress relative to conflicts between relational or social identities. There were no differences in magnitude or growth by the types of identities in conflict. As this finding differed from Study 2 where no relationship was found, we considered whether characteristics of the conflict or samples might shed light on this discrepancy. Unfortunately, we could not find any reasons for why this might be the case: In Studies 2 and 3 the number of conflicts that were on-going vs. resolved was similar, and when supplementary analyses controlled for additional participant characteristics (i.e., age, race), the results remained the same.

Third, and in support of Hypothesis 2, the management of conflict differed as a function of the types of identities in conflict. Here it was retreat that was used more to manage relational conflicts relative to role or social identity conflicts. Reconciliation, realignment, and reflection were not differentially used depending on the types of identities in conflict. Although different to Study 2, these pattern of these findings add to, rather than challenge, our understanding of the management strategies for different types of identities in conflict. It suggests that retreat might be used more when contending with relational conflicts whereas realignment might be used more with role conflicts. This makes sense as it is likely more difficult to extricate one's self from relationships than from roles which can explain the reliance on different management strategies for different types of identity conflicts.

Fourth, we investigated whether identity characteristics in general might be associated with the experience and management of conflict. In partial support of Hypothesis 3a, student identity flexibility, a role, was positively related to the experience of conflict and social identity flexibility was positively related to growth. Although this was opposite to the direction expected, as the flexibility of employee identity, relational identities, and social identities were generally unrelated to the majority of the conflict experience measures it lends support to the idea that perceived flexibility is a key aspect of the experience of conflict. In partial support of Hypothesis 3b, student identity flexibility was positively related to realignment and retreat. Relational identity flexibility was also positively related to retreat. These relationships are generally consistent with the conflict management findings of Studies 2 and 3 and lend credence to our earlier assertion that flexibility is related tied to the strategic use of conflict management strategies. We also found that more important identities and more positive identities, regardless of type, were generally associated with reconciliation, suggesting that the perceived value of identities may influence whether individuals use this strategy. Furthermore, exploratory analyses<sup>5</sup> also suggested that role identities were perceived as more flexible relative to relational and social identities, and that relational identities were perceived as more flexible relative to social identities. However, there were no differences in the perceived importance or valence of role, relational, and social identities. While this finding provides initial evidence that identity characteristics are related to the experience and management of conflict, as these were general measures the findings should be interpreted with caution until they are assessed in terms of the conflicting identities themselves.

# GENERAL DISCUSSION

Most people have multiple identities. Yet, how these identities are experienced and managed when they conflict is not well understood. In three studies we uncovered new and interesting insights into the nature of identity conflicts, and the experience and management of identity conflicts. To our knowledge these findings are the first to develop a robust measure of identity conflict management strategies as well as the first to find evidence that the types of identities matter for the management though not necessarily the experience of conflict.

# The Nature of Identity Conflicts

In considering the nature of conflict, we found that conflicts between roles, relationships, and life-stages were reported most frequently rather than social identities which tend to figure prominently in the literature on multiple identities and conflict. This highlights the diversity of identity conflicts and suggests that although social identities are important, researchers should also start thinking more broadly about the range of identity conflicts that can be experienced and their implications for individuals' lives. Although role identities

<sup>5</sup>We conducted an exploratory analysis to determine whether the characteristics differed as a function of the types of identities in general. We conducted three single-factor within subjects ANCOVAs with gender as the covariate for each characteristic. Conflict occurrence and conflict choice were not used as covariates because the ratings of these characteristics were unrelated to the experience or management of conflict. All pairwise comparisons were conducted with Least Significant Difference (LSD): Flexibility: A 4 Type of Identity (Student, Employee, Relational, and Social) ANCOVA found that student identity was perceived as more flexible (M = 3.61, SE = 0.095) relative to relational identities (M = 3.27, SE = 0.085, p = 0.005) or social identities (M = 2.95, SE = 0.095; p < 0.001). Employee identity (M = 3.61, SE = 0.095) and relational identities were perceived as more flexible than social identities (both ps < 0.005). Student and employee identity did not differ from each other in perceived flexibility (p = 0.122). Importance: A 3 Type of Identities (Role, Relational, and Social) ANCOVA indicated that importance ratings did not differ as a function of the types of identities, F(2,346) = 0.39, p = 0.68, η 2 <sup>p</sup> = 0.00. Valence: A 3 Type of Identities (Role, Relational, and Social) ANCOVA indicated that valence not differ as a function of the types of identities, F(2,346) = 1.74, p = 0.18, η 2 <sup>p</sup> = 0.01. These findings suggest that flexibility might differentiate roles, relationships and social categories from each other.

have started to receive more attention (e.g., Hodges and Park, 2013; Rabinovich and Morton, 2016), relational identities have been relatively absent from the literature and require further consideration.

# The Experience of Identity Conflict

In examining the experience of conflict, individuals generally reported similar experiences of conflict between role, relational, and social identities. In particular, the magnitude and growth associated with these conflicts was equivalent across studies. The finding for magnitude in particular is consistent with the idea that experiencing identity conflicts leave individuals similarly torn: the perceived discrepancy associated with these conflicts does not differ whether contending with role, relational, or social identities. This finding is important because it provides further evidence for the use of the BII as a measure of the experience of conflict (Cheng and Lee, 2013) and demonstrates that when any types of multiple identities are implicated in conflict, these conflicts are likely to be felt in similar ways. Growth was also generally experienced in similar ways across different types of conflicts. This finding suggests that when individuals are faced with difficulties such as identity conflict, irrespective of the specific nature of these conflicts, they may be similarly likely to draw on these experiences to learn and change as individuals (e.g., Tedeschi and Calhoun, 2004).

# The Management of Identity Conflict

In considering the management of conflict, individuals reported engaging in a range of strategies, including reconciliation, retreat, and realignment, and we found evidence for the use of these three strategies and a fourth strategy, reflection, to manage these conflicts. All individuals endorsed the use of reconciliation, which suggests that this may reflect the default strategy when facing identity conflicts. This is in line with Nowak et al. (2000) work on the self-system which suggests that the global integration of information is a default process for the self as well as Berry's work on acculturation, where integration is associated with positive adaptation (Berry et al., 2006; Sam and Berry, 2010). When identities conflict individuals may therefore be motivated to 'bring their identities back into line' by engaging in reconciliation. Critically, conflicts between role and relational identities were associated with more use of realignment and retreat, respectively, relative to other types of identities, suggesting that while individuals try to hold onto both identities, roles and relationships may also be differentially shorn as a part of the management process.

# Identity Characteristics and the Experience and Management of Conflict

We also considered whether identity characteristics might help us to understand the experience and management of conflicts. In thinking about why conflicts between multiple roles and multiple relationships are associated with varied management strategies relative to conflicts we found that participants perceived role, relational, and social identities as differing in flexibility but not in importance or valence. Furthermore, the greater perceived flexibility of one role in general, student identity, was associated with the experience and strategic management of conflict. This suggests that when identities are perceived as flexible individuals might 'feel more' and believe that they can and/or need to 'do more' when they experience identity conflict. Although these relationships are encouraging we recognize that identity characteristics were assessed in general rather than in terms of the specific conflicting identities (e.g., Settles, 2004). However, these initial findings suggest that considering the perceived flexibility of the multiple identities in particular is likely to provide a clearer picture of the differences between roles, relationships, and social categories that might shape the experience and management of conflict.

Taken together, we contend that our measures of the experience and management of identity conflict are robust and provide meaningful insights into the ways that people experience conflict between multiple identities as well as what they might do to manage these conflicts. Yet, these findings also highlight that the experience and management conflicts between different types of identities may not be mirrored across samples. In hindsight, perhaps we should not have expected this to be the case given the subjective nature of identity conflict. Indeed, even though we assigned people to discuss the same conflicts, the nature of these conflicts could differ. Although assessing conflicts between different types of multiple identities helped us in identifying and building robust measures we recognize that moving these ideas forward will likely require a focus on specific types of multiple identities in isolation.

# Limitations and Future Research

Although this research provides initial evidence that experiencing identity conflict is associated with its management and that the types of identities matter for conflict management, these studies may be limited in their generalizability. In particular, the range of conflicts initially reported (i.e., Study 1) may reflect the context in which the data were collected. It is possible that data collected in settings other than university campuses, and with participants other than university students, might elicit very different identity conflicts, and potentially reveal different conflict management strategies. Moreover, university students may have more limited and/or more flexible roles than do older adults. As such, it would be important to explore the nature of role conflicts specifically, and identity conflicts in general, among adults in community samples who may have more fixed role identities, and who may have more experience in managing potential conflicts between their identities. Sample diversity will allow us to consider broader definitions of multiple role, relational, and social identities as well as enabling further tests of the experience of conflict, and the development if necessary, of additional strategies for identity conflict management.

The present findings were also all based on self-reports taken at one moment in time. Given the dynamic process of identity conflict and management, longitudinal studies that track individuals over time to see how the process of conflict unfolds, and whether there is a sequence in which management strategies are used, are recommended. Findings from Study 1 suggest that

individuals may also be 'doing nothing' or engaging in other neutral strategies that suggest that there might have to be some 'stew' time before people act to resolve these identity conflicts. In looking at this process longitudinally, it might be possible to determine whether there are different motivational factors associated with engaging in some strategies rather than others, whether there is a particular time in which this process occurs, and whether there are individual differences in the length of time it takes to resolve conflicts. This type of analysis can offer a clearer picture of the conflict process and can shed light on factors that may be associated with adaptive vs. maladaptive coping with identity conflict.

Finally, with regards to all of these studies, there is a focus on explicit conflict. However, conflict does not always occur explicitly. In fact, the activation of identities is likely an implicit process, with individuals not consciously attending to the cues that elicit these identities (e.g., Sinclair et al., 2006), particularly when contending with less flexible identities. This may explain why conflicts between role identities and relational identities, rather than social identities emerged as the most frequently reported conflicts – it may be easier for individuals to be consciously aware of conflicts arising from roles as they are associated with explicit actions, or between relationships because they can pinpoint disagreements associated with significant others. However, where social identities are concerned, conflicts may be less easy to identify or may be more ambiguous due to the attributions that one can make for conflict leading these types of conflicts to be under-reported. Asking people about their identity conflicts may therefore change the actual experience of these conflicts. Using implicit approaches to investigate identity conflicts (e.g., Hodges and Park, 2013) may provide important insights into the experience and management of conflicts between different types of multiple identities.

# CONCLUDING REMARK

Conflict may be inevitable as individuals navigate the multiple identities that help them to make sense of who they are and their social world. Although the present research offers important insights into the experience and management of conflict between

# REFERENCES


multiple identities, these relationships require further scrutiny to understand why and when the types of identities might matter.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of 'Human Research Ethics Committee, York University (Canada)' and 'Queen Mary University of London Research Ethics Committee' (United Kingdom) with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the 'Human Research Ethics Committee at York University (Canada) (Studies 1–3) and Queen Mary University of London Resarch Ethics Committee (Study 3).'

# AUTHOR CONTRIBUTIONS

JJ developed the idea, designed the studies, and analyzed and interpreted the data. MH assisted in the development of the idea, study design, and interpretation of the data.

## ACKNOWLEDGMENTS

We would like to thank Simorrah Colaco and Mariam Mahmood for their help with coding the open-ended responses in Study 1, the Social Identity and Intergroup Relations Lab at York University (Banu Cingöz-Ulu, Reeshma Haji, Benjamin Giguère, Yvonne Lai, Richard N. Lalonde, Ilil Naveh-Benjamin) for their feedback on the identity conflict management items used in Study 2 and Thomas A. Morton and Ilka H. Gleibs for comments on earlier drafts of this manuscript.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyg. 2017.01732/full#supplementary-material



conflict on well-being. Self Identity 15, 224–244. doi: 10.1080/15298868.2015. 1117524


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Jones and Hynie. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Does Identity Incompatibility Lead to Disidentification? Internal Motivation to Be a Group Member Acts As Buffer for Sojourners from Independent Cultures, Whereas External Motivation Acts As Buffer for Sojourners from Interdependent Cultures

#### Edited by:

Fabio Lorenzi-Cioldi, University of Geneva, Switzerland

#### Reviewed by:

Vassilis Barkoukis, Aristotle University of Thessaloniki, Greece Thomas Morton, University of Exeter, UK

\*Correspondence: Christina Matschke c.matschke@iwm-tuebingen.de

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

> Received: 30 June 2016 Accepted: 22 February 2017 Published: 07 March 2017

#### Citation:

Matschke C and Fehr J (2017) Does Identity Incompatibility Lead to Disidentification? Internal Motivation to Be a Group Member Acts As Buffer for Sojourners from Independent Cultures, Whereas External Motivation Acts As Buffer for Sojourners from Interdependent Cultures. Front. Psychol. 8:335. doi: 10.3389/fpsyg.2017.00335 Christina Matschke<sup>1</sup> \* and Jennifer Fehr<sup>2</sup>

<sup>1</sup> Leibniz-Institut für Wissensmedien, Tübingen, Germany, <sup>2</sup> Hector Research Institute of Educational Science and Psychology, University of Tübingen, Tübingen, Germany

Most individuals possess more than one relevant social identity, but these social identities can be more or less incompatible. Research has demonstrated that incompatibility between an established social identity and a potential new social identity impedes the integration into the new group. We argue that incompatibility is a strong risk factor for disidentification, i.e., a negative self-defining relation to a relevant group. The current research investigates the impact of incompatibilities on disidentification in the acculturation context. We propose that incompatibility between one's cultural identities increases the disidentification with the receiving society. It has, however, been shown that the motivation to be a group member serves as a buffer against negative integration experiences. Moreover, research from the intercultural domain has shown that intrinsic and extrinsic motivation has specific effects for members of cultures that differ in self-construal. In a European sample of High school exchange students (Study 1, N = 378), it was found that incompatibility was positively related to disidentification, but only for less (but not more) intrinsically motivated newcomers. In an Asian sample of international university students (Study 2, N = 74), it was found that incompatibility was also positively related to disidentification, but only for less (but not more) extrinsically motivated newcomers. Thus, the findings demonstrate that the effect of incompatibility between social identities on disidentification can be buffered by motivation. The results suggest that, depending on cultural self-construal, individuals have different resources to buffer the negative effect of incompatibility on the social identity.

Keywords: incompatibility, disidentification, self-construal, motivation, integration, acculturation

# INTRODUCTION

fpsyg-08-00335 March 4, 2017 Time: 16:56 # 2

Imagine Jane and Yukiko, two international sojourners who come to another country to study there. Jane is interested in the host culture and has always dreamt about living in that country. Yukiko's parents have gone through great difficulties in order to offer Yukiko the best education abroad. Thus, both Jane and Yukiko are strongly motivated to become an integrated member of the receiving society. However, they soon encounter expectations of the receiving society that differ from what they are used to, based on their prior cultural identity. In other words, they experience incompatibility between their internalized primary cultural identity and the potential new social identity of the receiving society. One way to solve this conflict is to discard the new social identity. If social groups are omnipresent and one is confronted with the expectation to integrate, however, discarding an identity may be impossible. If the social group remains important for the self-concept but has a negative meaning, an individual might develop a disidentification with the group instead. Jane and Yukiko, for instance, may start to describe themselves as contrary to the people from the receiving society, to have negative feelings about these people and to act against the interests of the people. We argue that incompatibility between social identities increases the risk for a disidentification with a new group. It has, however, been shown that the motivation to be a group member can buffer incompatibility effects. The purpose of the current research is to investigate the impact of incompatibility on disidentification with a new social group, and the potential of motivation to serve as a buffer.

# Incompatibility Impedes Integration

Most individuals possess more than one relevant social identity. Social identities are, however, not always compatible. Newcomers of social groups seem to be especially sensitive to the relationship between standing social identities and the potential new social identity. It has been demonstrated that incompatibility between pre-existing social identities and the new group identity hampers the integration into new groups. Incompatibility has been conceptualized as cultural distance between primary and secondary culture (Babiker et al., 1980; Ward and Kennedy, 1992), inconsistent attitudes toward acculturation between the sojourner and the receiving society (Bourhis et al., 1997; Roccas et al., 2000; Zagefka and Brown, 2002), identity conflict (Leong and Ward, 2000; Ward et al., 2011) or psychologically felt incompatibility between identities (Benet-Martínez et al., 2002; Jetten et al., 2008; Cheng and Lee, 2009; Iyer et al., 2009; Martinovic and Verkuyten, 2012; Matschke and Fehr, 2015). The integration outcomes have been captured with measures of psychological and physical well-being (e.g., Babiker et al., 1980; Roccas et al., 2000; Chirkov et al., 2008), socio-cultural adaptation (Ward and Searle, 1990; Ward and Kennedy, 1992), intergroup relations (Zagefka and Brown, 2002), and social identification (Iyer et al., 2009; Matschke and Fehr, 2015). Despite these different conceptualizations and measurements, the core finding is that incompatibility makes the integration into a new group more difficult.

Research has, however, scarcely investigated the impact of incompatibility on negative integration outcomes. This is especially striking, as the societal discourse lies on negative outcomes, such as the separation of cultural subgroups, or acts of aggression and terrorism conducted by societal newcomers. Such passive and active harm toward one's ingroup is often rooted in disidentification with that group (Elsbach and Bhattacharya, 2001; Becker and Tausch, 2014). Therefore, it is important to understand what factors affect newcomers' disidentification with the receiving society, and what factors act as a buffer against disidentification. The current research is one of the few that investigates the impact of incompatibility on disidentification and the first to consider motivation as a buffer.

# Disidentification – The Outcome of Negative Integration Processes

For newcomers, the new group is a potential source of a new social identity. In the acculturation context, the receiving society is a relevant group and has a great potential to be included into the newcomers' self-concepts, just like an ingroup (e.g., Berry, 1997; Sassenberg and Matschke, 2010). Whereas social identification is the positive form of such an inclusion, disidentification is the negative form. Disidentification is an active distancing of a group that is relevant for the self-concept. Disidentified group members feel contrary to the group and do not want to belong to that group (Elsbach and Bhattacharya, 2001; Becker and Tausch, 2014). Just like social identification (Tajfel and Turner, 1979), disidentification with a group affects an individual's self-description, affect and behavior. Disidentified sojourners, for instance, would describe themselves as contrary to members of the receiving society, have a negative affective association to the group and act against the interests of the receiving society (Matschke and Sassenberg, 2010b). Social psychological research has created explicit multidimensional measures of disidentification that either focus on capturing the active distancing of an individual from a group (Matschke and Sassenberg, 2010b), or the feeling of opposition to the group behaviorally, cognitively and affectively (Becker and Tausch, 2014). Disidentification, captured with these explicit measures, has both specific antecedents and consequences. It is especially induced by negative experiences with a group, such as rejection (Matschke and Sassenberg, 2010b), a bad reputation (Elsbach and Bhattacharya, 2001), or illegitimate group assignment (De Vreeze and Matschke, 2016, unpublished). What makes the study of disidentification especially relevant is the finding that it predicts negative behavior toward the ingroup as well as other outgroups: disidentification is related to active and passive harm toward the ingroup (Becker and Tausch, 2014), to turnover (e.g., Elsbach and Bhattacharya, 2001; Matschke et al., 2016, unpublished), to antinormative behavior (Matschke and Sassenberg, 2010a; Matschke et al., 2016, unpublished), to biased information behavior (De Vreeze and Matschke, 2016, unpublished; Matschke et al., 2016, unpublished), and to the derogation of other low status outgroups (Ikegami and Ishida, 2007; Matschke et al., 2016, unpublished). It is therefore important to reach a better understanding of the risk factors and buffers for disidentification

in newcomers in order to increase our knowledge on negative integration outcomes.

Recent research has demonstrated that incompatibility between prior group memberships and new group memberships induces disidentification with the new group (De Vreeze et al., 2016, unpublished; Matschke et al., 2016, unpublished). In acculturation research, disidentification has mostly been captured indirectly with decreased levels of identification (e.g., Jasinskaja-Lahti et al., 2009), but recently direct measures of disidentification have successfully been used to investigate negative integration processes (Verkuyten and Yildiz, 2007; Matschke and Sassenberg, 2010b). The impact of incompatibility on disidentification has, however, not been investigated in the acculturation context. Based on the prior findings, we propose that incompatibility increases the risk for disidentification.

# Motivation to Be a Group Member

In the process of integrating oneself into a new group, feelings of incompatibility occur easily. Especially when living in a new cultural society, the experience of incompatibilities is rather the rule than the exception. But not all newcomers disidentify with their new groups. What resources have newcomers in coping with incompatibility that make them refrain from disidentification?

There is evidence that the motivation to be a group member plays a crucial role for the integration success. According to Self-determination theory (SDT; Deci and Ryan, 2000), there are different motivations to pursue one's goals. Goals that fulfill the fundamental need for autonomy and selfdetermination are referred to as intrinsically motivated goals. Goals that are imposed upon individuals for reasons that lie outside their basic needs, like wishes of other persons or forces by circumstances, are referred to as extrinsically motivated goals. Based on this reasoning, research by Plant and Devine (1998) differentiates between intrinsic and extrinsic reasons for behavior. Whereas intrinsic motivation stems from internalized, personal beliefs, extrinsic motivation reflects the desire to avoid negative evaluations by others. Across different contexts, it has been shown that intrinsically motivated individuals are more successful in their goal pursuit than extrinsically motivated individuals (Plant and Devine, 1998; Deci and Ryan, 2000).

Recently, motivation has also become a matter of interest in the context of acculturation. Applied to the acculturation context, Chirkov et al. (2007, 2008) demonstrated that a selfdetermined motivation of students to study abroad increased their wish to acculturate, their socio-cultural adaptation, their subjective and physical well-being and the self-determination of academic activities. Thus, intrinsic motivation seems to increase the chances of successful acculturation processes. But why is intrinsic motivation so effective in goal pursuit?

According to the Self-completion theory (SCT; Wicklund and Gollwitzer, 1982) intrinsic motivation drives success because it leads to persistence in goal-pursuit. In other words: intrinsically motivated individuals are more successful because they do not give up when they meet obstacles. When intrinsically motivated goals are unfulfilled, a disagreeable sense of incompleteness is induced. In order to overcome this sense of incompleteness, more intrinsically motivated individuals increase their efforts to reach their goals. In contrast, goals with less intrinsic motivation do not induce this sense of incompleteness and will therefore be dropped when there are obstacles (Brunstein and Gollwitzer, 1996; Fehr and Sassenberg, 2009).

The superiority of intrinsic over extrinsic motivation is not, however, as global as assumed at first: There are theoretical accounts and empirical evidence that this finding is specific to certain cultures and may not be valid for individuals of all cultures (e.g., Markus and Kitayama, 1991; Iyengar and Lepper, 1999). We return to this question of the possible cultural specificity of intrinsic versus extrinsic motivation effects in the second study. Nevertheless, in Western cultures, it has been demonstrated that intrinsic motivation helps individuals to pursue their goals when they meet difficulties.

Sojourners that live in a foreign country with the goal to acculturate to the receiving society are likely to meet obstacles in form of identity incompatibility. Based on SDT and SCT, we suggest that intrinsic motivation buffers the effect of incompatibility on disidentification. In support of this idea, it has been shown that only less (but not more) intrinsically motivated newcomers disidentify with the new group when they feel rejected (Matschke and Sassenberg, 2010b). Moreover, Matschke and Fehr (2015) found that incompatibility impairs a developing social identification (i.e., a positive outcome criterion) of newcomers only if these are less (but not more) intrinsically motivated. We therefore hypothesize that incompatibility will be positively related to disidentification with the receiving society only for less (but not more) intrinsically motivated sojourners.

# STUDY 1

# Method

# Design, Participants, and Procedure

A correlational study with the continuous independent variables incompatibility and intrinsic as well as extrinsic motivation and the criterion disidentification was conducted. A sample<sup>1</sup> of 378 European exchange students (277 females, 97 males, 4 did not indicate gender, age M = 17, range 15–19 years) from 26 European countries, who had spent an exchange year in another European country, where they lived in local families and visited local High Schools filled in a questionnaire at their final seminar before returning home. Juristically, during the exchange year, i.e., also during the time when the data was collected, the organization stands in the position of the guardian of the adolescents. The organization encouraged the study and approved the questionnaire in detail. The questionnaire was voluntary, students were informed about the content of the study and consented to participate. There was, however, no written informed consent of the participant, because at the time of data collection, the approval of the guardian or organization was the ethical recommendation. Questionnaires were either in English (N = 184) or German (N = 193) and were introduced

<sup>1</sup>Post hoc linear multiple regression, fixed model, R 2 increase, power analysis conducted with G∗Power 3.1.5 (Faul et al., 2007), parameters set to effect size R 2 change = 0.029, alpha = 0.05, sample size N = 378, one tested predictor, five total predictors, indicated a good power of 0.95.

as a study about experiences during the exchange. Participants filled in measures on their intrinsic and extrinsic motivation to spend an exchange year in their host country, incompatibility, disidentification, and demographics. Debriefing was conducted via e-mail.

#### Measures

Motivation was measured with four items adapted from the Academic Self-Regulation-Questionnaire (Ryan and Connell, 1989). Two items measured intrinsic motivation (r = 0.33, p < 0.001, "I went as an exchange student to my host country because I felt like it" and ". . .because I thought that the experiences that I make in my host country would be fun," M = 5.83, SD = 1.09), two items measured extrinsic motivation (r = 0.41, p < 0.001, ". . .because I thought that it would be appreciated by people that are important to me." and ". . .because I wanted others to see that I can do it," M = 3.36, SD = 1.65). Intrinsic and extrinsic motivation measures were unrelated, r = −0.06, p = 0.271.

Incompatibility was operationalized as divergence of acculturation strategies (Bourhis et al., 1997; see also Roccas et al., 2000, for a similar conceptualization). Primary culture incompatibility was captured by two measures: (1) participants' attitudes toward the maintenance of the primary culture, captured with five items (α = 0.70, e.g., "It is important that exchange students maintain their home country's way of living") adapted by Zagefka and Brown (2002) and Geschke et al. (2010), and (2) the perceived attitudes of the host country, captured with the item "It is generally not popular in my host country if exchange students from my home country maintain their way of living." Likewise, secondary culture incompatibility was captured by two measures: (1) the exchange students' own attitudes on participation and contact with the secondary culture, captured with eight items (α = 0.78, e.g., "It is important that exchange students participate in their host country's culture") developed by ourselves and adapted from Zagefka and Brown (2002) and Geschke et al. (2010), and (2) the perceived attitudes of the host country, captured with the item "It is not liked in my host country when exchange students from my home country get into contact with local people." In order to represent that only the combination of strong attitudes and strong meta-attitudes induce a psychologically meaningful incompatibility, the product<sup>2</sup> of the attitudes and the meta-attitudes was calculated as an individual measure of incompatibility for each acculturation conflict. As a common measure of incompatibility, the two conflict measures (r = 0.34, p < 0.001) were aggregated (M = 14.66, SD = 7.60). The incompatibility measure was unrelated to intrinsic motivation (r = −0.03, p = 0.59), and positively related to extrinsic motivation (r = 0.21, p < 0.001).

#### TABLE 1 | Coefficients of the hierarchical regression in Study 1 (N = 378).


<sup>+</sup>p < 0.10, <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Ten items measured disidentification (α = 0.88). The items were adapted from Matschke and Sassenberg (2010a). This scale has a strong focus on cognitive (e.g., "I make myself aware that there are other groups and countries besides my host country that are important to me), affective (e.g., "I feel bad when I meet people from my host country") and behavioral (e.g., "I doubt that I will keep in contact with my host country for long") distancing of an individual from a specific group.

All other items were assessed on a 7-point scale (1 = I don't agree at all, 7 = I fully agree).

# Results

It was expected that incompatibility is positively related to disidentification, but only for those who are less (but not more) in intrinsic motivation. In order to test for potential effects of demographic variables, the independent variables and the criterion were correlated with gender and age. The correlation with gender indicated than females experiences less incompatibility (r = 0.16, p = 0.002) and disidentification (r = 0.19, p < 0.001), but were higher in intrinsic motivation (r = −11, p = 0.025), all other correlations remained nonsignificant, all rs < 0.09, all ps > 0.12. We therefore control for gender and age in the following analyses.

In order to test whether the buffering effect is specific to intrinsic motivation, a hierarchical regression analysis with the control variables gender and age and the continuous predictors Incompatibility, Intrinsic motivation, and Extrinsic motivation was conducted on disidentification. In the first step, gender and age were entered in the analysis. In the second step, the main effects were entered in the analysis, then the predicted Incompatibility × Internal motivation interaction (step 3), followed by the Incompatibility × External motivation interaction (step 4). Note that we did not expect or test the Incompatibility × Intrinsic motivation × Extrinsic motivation interaction. The regression analysis (see **Table 1**) revealed in the first step, that gender was indeed related to disidentification throughout the models, whereas the relation between age and disidentification remained marginal and became non-significant when other variables were entered into the model. In the second step, the analysis revealed that stronger Incompatibility was related to stronger disidentification, B = 0.025, SE = 0.004, p < 0.001, whereas Intrinsic motivation, B = −0.03, SE = 0.04 p = 0.421, and Extrinsic motivation B = 0.03, SE = 0.02, p = 0.204, were unrelated to disidentification. When entered into

<sup>2</sup>This measure is similar to the discrepancy measure introduced by Roccas et al. (2000). We chose to use the product of participants' own attitudes and the meta-attitudes because of the negative framing of the meta-attitudes. Moreover, the product reflects that only larger differences between sojourners' and receiving societies' attitudes results in psychologically felt incompatibilities, whereas lower passion toward attitudes, or lower meta-attitudes, should not result in a psychologically relevant feeling of conflict.

the model separately, the expected Incompatibility × Intrinsic motivation was significant, B = −0.11, SE = 0.04; p = 0.003, and remained significant with B = −0.12, SE = 0.04, p = 0.002, when the not significant Incompatibility × Extrinsic motivation, B = −0.05, SE = 0.04, p = 0.217, was entered into the model.

Simple slope analyses resolved the Incompatibility × Intrinsic motivation interaction (Aiken and West, 1991, with standardized independent variables). The analyses revealed that when less intrinsically motivated (1 SD below the mean), Incompatibility was positively related to disidentification, B = 0.32, SE = 0.05, p < 0.001. When more intrinsically motivated (1 SD above the mean), Incompatibility was less strongly related to disidentification, B = 0.09, SE = 0.06; p = 0.079 (see **Figure 1**).

# Discussion

Study 1 investigated a sample of European sojourners. It was expected and found that incompatibilities between acculturation attitudes are positively related to disidentification with the receiving society, but only if sojourners are less intrinsically motivated. When more intrinsically motivated, this relation did not disappear completely, but sojourners were less strongly affected by incompatibility. Extrinsic motivation did not affect disidentification, nor did it buffer the incompatibility effects. Study 1 thus supports the prediction that intrinsic motivation has the power to buffer the integration process from difficulties: it weakens the negative effects of incompatibility on disidentification.

#### Intercultural Differences in Motivation Effects

Study 1 demonstrated that intrinsic (but not extrinsic) motivation specifically buffered the effects of incompatibility on disidentification. There is, however, reason to suppose that even extrinsic motivation may serve as a buffer against incompatibility, depending on the self-construal of individuals in certain cultures. Across a wide range of theories, it is suggested that individuals differ across cultures in whether they construe their self as individualistic or independent vs. collectivistic or interdependent (e.g., Hofstede, 1980; Triandis, 1989; Markus and Kitayama, 1991; Singelis and Brown, 1995). Individuals with independent self-construal's experience and behavior are guided by their own inner thoughts, emotions, and actions. Individuals with interdependent self-construal's experience and behavior are guided by the thoughts, emotions, and actions of relevant others.

There is reason to suppose that individuals with interdependent self-construal are strongly driven by extrinsic motivation, because they are more strongly motivated by the goals of relevant others (Markus and Wurf, 1987) 3 . They are more motivated by activities that significant others chose for them, and more willing to be guided by significant others' opinions than individuals with independent self-construal (Markus and Kitayama, 1991; Cialdini et al., 1999; Hernandez and Iyengar, 2001; Iyengar and Brockner, 2001; Kitayama and Uchida, 2005). Motivation and persistence are stronger for individuals with interdependent self-construal, if the task is extrinsically motivated, compared to individuals with independent selfconstrual, whose motivation and persistence are strongest when intrinsically motivated (Iyengar and Lepper, 1999; Heine et al., 2001). Individuals with interdependent self-construal also prefer choices for others over choices for themselves, whereas the pattern is the opposite for individuals with independent self-construal (Pöhlmann et al., 2007). Study 1 demonstrated that intrinsic, but not extrinsic motivation buffered the relation between incompatibility and disidentification. Study 1, however, only tested this effect in Europeans, who are considered to be independent in self-construal (Hofstede, 1980; Markus and Kitayama, 1991). Based on the findings on motivation and self-construal, we suggest that for individuals from cultures with interdependent self-construal, extrinsic motivation is a buffer against the relation between incompatibility and disidentification. More precisely, we predict in a sample from a culture of interdependent self-construal that for less (but not more) extrinsically motivated newcomers, incompatibility is positively related to disidentification.

# STUDY 2

Study 2 is a correlational study with a small sample of individuals from different Asian countries. Asian cultures are considered to be cultures with interdependent self-construal (Hofstede, 1980; Markus and Kitayama, 1991). The study aims at collecting first evidence for the prediction that for individuals with interdependent self-construal, extrinsic motivation serves as a buffer against the effect that incompatibility has on disidentification.

# Method

#### Design, Participants, and Procedure

Study 2 was a correlational study that investigated the continuous independent variables incompatibility, intrinsic, and extrinsic motivation and the criterion disidentification in a sample

<sup>3</sup>While Markus and Kitayama (1991; see also Iyengar and Lepper, 1999) argue that individuals with interdependent self-construal have motivation driven by others that would become intrinsic, in classic terms of SDT this kind of motivation would be called extrinsic motivation, i.e., the motivation to do things enforced by circumstances or relevant other persons. For the sake of differentiation while reading this paper, we will continue to refer to the construct as extrinsic motivation.

of individuals from cultures that are interdependent in selfconstrual. Seventy-four<sup>4</sup> Asian students from 10 different countries (with the majority of 54 from Malaysia, 68 females, 6 males, age M = 22.58, range 19–48 years), who had moved to New Zealand about Median = 5 months (range: 0.5–156 months) ago in order to study at a local university, participated in the study. The questionnaire included an information sheet that informed the participants that their completion of the questionnaire will be understood as consent for the use and publication of the data. Study 2 was approved by the ethical committee of the School of Psychology, Victoria University Wellington (New Zealand). Participants filled in English language measures on intrinsic and extrinsic motivation to move to New Zealand, disidentification, and demographics, as well as other measures irrelevant to the research question. The study was part of a larger data collection advertised in mailing lists and introduced at meetings for international students as a study about experiences in New Zealand that could be filled in online (N = 4) or paper-and-pencil (N = 70). Participants were debriefed via e-mail.

#### Measures

As the motivation measure of Study 1 was rather short, in Study 2 motivation was measured with an adapted version of the Self-Regulation-Questionnaire – Study abroad (short version; Chirkov et al., 2008). The questionnaire captures intrinsic motivation to move to New Zealand with four items including intrinsic and identified regulation (α = 0.81; e.g., "I moved to New Zealand because I thought it was an exciting thing to do," M = 3.43, SD = 1.06). Extrinsic motivation was measured with six items that captured both extrinsic and introjected motivation (α = 0.80, e.g., "I moved to New Zealand because others (relatives and friends) forced me to do this," M = 2.04, SD = 0.95). Note that all extrinsic motivation items describe reasons to study abroad that are connected to one's relationships to others. The items were assessed on a 5-point scale (1 = not at all because of this reason, 5 = completely because of this reason). Intrinsic and extrinsic motivation measures were correlated, r = 0.46, p < 0.001.

Incompatibility (α = 0.76; M = 2.02, SD = 0.63) was measured with seven items adapted from the cultural distance scale of Babiker et al. (1980) and Piontkowski et al. (2000). Perceived differences between culture, religion, climate, food, educational level, material comfort, family life, and people in general were rated on a 5-point scale (1 = I don't agree at all, 5 = I fully agree). Incompatibility was unrelated to intrinsic and extrinsic motivation, both r < |0.15|, p > 0.20.

Disidentification (α = 0.68, M = 2.29, SD = 0.73) was measured as in Study 1, but one item ("I tell myself that I have other groups where I can play a part in") was deleted from the scale to improve intrinsic consistency (α = 0.71). Thus, the final scale consisted of nine items.

#### TABLE 2 | Coefficients of the hierarchical regression in Study 2 (N = 74).


<sup>+</sup>p < 0.10, <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

## Results

It was expected that for individuals with interdependent self-construal, incompatibility is positively related to disidentification, but only for less (but not more) extrinsically motivated sojourners. In order to test for potential effects of demographic variables, the independent variables and the criterion were again correlated with gender and age. Age was related to incompatibility, r = −0.27, p = 0.019, all other correlations remained non-significant, all rs < 0.17, all ps > 0.14. As in Study 1, we control for gender and age in the following analyses. The main prediction was tested in a hierarchical regression analysis with the control variables gender and age, and the continuous predictors Incompatibility, Extrinsic motivation, and Intrinsic motivation on disidentification. Demographics were entered into the analysis first, then the main effects (step 2), the Incompatibility × External motivation interaction (step 3), followed by the Incompatibility × Internal motivation interaction (step 4). For the demographic variables, the analysis (see **Table 2**) revealed no significant effects of gender and age, all βs < 0.18, all ps > 0.14. In step 2, neither the main effects of Incompatibility, B = 0.04, SE = 14; p = 0.789, nor Intrinsic motivation B = −0.04, SE = 0.09, p = 0.684, were significant. Extrinsic motivation was positively related to disidentification, B = 0.24, SE = 0.10, p = 0.023. When entered into the model separately, the expected Incompatibility × Extrinsic motivation was significant, B = −0.28, SE = 0.10, p = 0.006. When the not significant Incompatibility × Internal motivation, B = −0.20, SE = 0.15, p = 0.191, was entered into the model, the Incompatibility × External motivation interaction remained marginally significant, B = −0.21, SE = 0.11, p = 0.070.

When the Incompatibility × Extrinsic motivation interaction was resolved into simple slopes (Aiken and West, 1991, with extrinsic motivation standardized, 1 SD above/below the mean), the predicted pattern was found: incompatibility was marginally related to disidentification only for less extrinsically motivated sojourners, B = 0.40, SE = 0.21, p = 0.066, whereas more extrinsically motivated sojourners' disidentification was unaffected by incompatibility, B = −0.02, SE = 0.14, p = 0.887 (see **Figure 2**).

## Discussion

Study 2 investigated the effect of incompatibility and motivation on disidentification in a sample from a culture with interdependent self-construal. It was expected that

<sup>4</sup>Post hoc linear multiple regression, fixed model, R 2 increase, power analysis, parameters set to effect size R 2 change = 0.095, alpha = 0.05, sample size N = 74, one tested predictor, seven total predictors, indicated a medium power of 0.84. A sample size of N = 92 would have had a good power of 0.90 (A priori linear multiple regression, fixed model, R 2 change with parameters set as described as above).

incompatibility is only positively related to disidentification for those lower (but not higher) in extrinsic motivation. The data provided first evidence that extrinsic motivation has the predicted buffering effect on the relation between incompatibility and disidentification. This interpretation can, however, only be made with caution, because the effects of Study 2 may be underpowered, and Study 2 can only be taken as first evidence which should stimulate future research.

Intrinsic motivation did not significantly affect the relation between incompatibility and disidentification in Study 2, but there was a descriptive relation between the two constructs that might reach conventional levels of significance in studies with greater power. Moreover, both types of motivation were, other than in Study 1, correlated. The fact that extrinsic, but not intrinsic motivation served as a buffer in Study 2 strengthens the assumption that extrinsic motivation is a stronger resource in a sample from interdependent cultures compared to independent cultures, but it cannot be ruled out that any kind of motivation, or both intrinsic and extrinsic motivation serve as buffers for incompatibility experiences in samples from cultures with interdependent self-construal.

# GENERAL DISCUSSION

Two correlational studies investigated the effect of incompatibility and motivation on disidentification with the new group in the acculturation context. It was expected that motivation buffers the effect of incompatibility on disidentification with the new group. The data supported the predictions. In a European sample (Study 1), incompatibility was positively related to disidentification, but only if newcomers were less (not more) intrinsically motivated. In an Asian sample incompatibility was also positively related to disidentification, but only when newcomers were less (not more) extrinsically motivated.

# The Impact of Incompatibility

When individuals possess several group memberships, it is easily possible that these are incompatible and difficult to reconcile in the self-concept (Roccas and Brewer, 2002; Amiot et al., 2007). Individuals who move to another country to live there are likely to experience incompatibility somehow or other. Research has demonstrated that incompatibility hampers the integration process (e.g., Jetten et al., 2008; Iyer et al., 2009; Matschke et al., 2016, unpublished). The current research used two different kinds of incompatibilities that sojourners may experience: differences in acculturation attitudes and cultural distance. Whereas differences in acculturation attitudes are a rather subjective measure, cultural distance is more objective. Both measures do not directly capture whether individuals actually think that the differences are problematic (compare, for instance, Matschke and Fehr, 2015). Despite these differences, the patterns are similar, and both types of incompatibilities could be buffered by motivation. The ability to keep on track despite cultural or attitudinal differences between oneself and the new group may prove essential for the healthy integration process. Even though incompatibilities have received considerable attention by acculturation researchers, motivation has so far never be considered as a buffer for incompatibility experiences, let alone in a way that differentiates between extrinsic and intrinsic motivation (for an exception, see Matschke and Fehr, 2015). The current research contributes to the growing body of evidence that demonstrates that motivation can make a difference in the integration process, especially when the process turns out to be thorny.

# Disidentification in Integration Processes

The present research is one of the few that take a deeper look into negative integration outcomes. When newcomers are strongly immersed in their group, as is the case on sojourns and during migration, it is unlikely that the self-concepts of newcomers remain untouched by incompatibility. Most research that investigated the impact of incompatibility of multiple social identities has captured positive outcomes, such as the lack of social identification or socio-cultural adaptation (e.g., Ward and Kennedy, 1992; Iyer et al., 2009; Matschke and Fehr, 2015). So far, negative outcomes of incompatibility captured by research were passive constructs like the indifference toward the receiving society (e.g., anomie or individualism in Bourhis et al.'s model) and physical symptoms (e.g., headaches or other health issues; Chirkov et al., 2008; for an exception, see Verkuyten and Yildiz, 2007; Matschke and Sassenberg, 2010a,b). Thus, individuals with incompatible multiple identities are often treated as passive victims of the negative integration processes. By investigating disidentification, which is an important predictor of negative behavior toward an ingroup (Becker and Tausch, 2014), group members are considered an active part of a negative integration process. Acts of aggression conducted by newcomers underline the importance of the investigation of active, negative integration outcomes in newcomers. Even if acts of aggression are, of course, the tip of the iceberg, disidentification in group newcomers is hardly constructive for the long-term integration success. It is therefore important to investigate the early development of disidentification in newcomers.

# Motivation as a Resource

Motivation seems to be a powerful resource when people enter new groups: it can harden people's persistence toward difficulties.

Even though different motives to move abroad have been suggested some 20 years ago (e.g., Bierbrauer and Pedersen, 1996; Berry, 1997; Ward et al., 2001), the differentiation into different kinds of motivation is still under-investigated. In line with SDT and SCT, Study 1 demonstrated in a European sample that especially intrinsic motivation hardens people's persistence toward difficulties. However, Study 2 suggests that in Asian samples, extrinsic motivation may play a similar buffering role as internal motivation does in the European sample. This finding supports the assumption that individuals with different self-construal are motivated by different things. Individuals with interdependent self-construal rely more strongly on the judgment of significant others and take their opinions more strongly as a guideline for their own behavior than individuals with independent self-construal (see also Hernandez and Iyengar, 2001). The present research thus underlines that the principles of SDT and SCT are applicable to more than one cultural background, if adjusted to the context: it is not only the intrinsic motivation that leads to success (Deci and Ryan, 2000) or persistence (Wicklund and Gollwitzer, 1982), but individuals with interdependent self-construal can similarly be driven by extrinsic motivation. Theories on motivation should take these differences more strongly into account and acknowledge that there may be different paths to a healthy goal-pursuit.

# Limitations and Future Avenues

Studies 1 and 2 contain two correlational datasets that differ in a variety of dimensions (e.g., age, intercultural context, host culture, temporary vs. permanent residence) in addition to selfconstrual. Thus, the interpretation of the results as a crosscultural comparison can only be made with caution. In fact, self-construal has not been measured in the current set of studies, but was derived from the cultural background of the sample. This approach, however, over-simplifies individuals, as they differ in their self-construal also within cultures (Markus and Kitayama, 1991). Future research should replicate the present findings in one sample with heterogeneous self-construal and use it as a factor in the design, e.g., by measuring self-construal in an intercultural sample of sojourners or by manipulating self-construal (e.g., Kühnen et al., 2001) in order to rule out alternative explanations for the differences in the samples.

The differences in the relations between intrinsic and extrinsic motivation between the two samples is another indication that the samples differ in their motivational structure. Considering that studying abroad is something that usually only high social class background people can afford, and that high social class background is related to independent (rather than interdependent) self-construal (e.g., Stephens et al., 2012), the specific pattern of intrinsic and extrinsic motivation in Study 2 might also be due to a high social class sample from Asia. Thus, the sample might not be purely interdependent in their self-construal. Measuring self-construal and social class background in future studies could deepen our understanding of different sources of self-construal and their effects on motivation.

The finding that intrinsic motivation is a strong resource for individuals with independent self-construal seems to be robust, whereas the evidence whether individuals with interdependent self-construal are driven by extrinsic or intrinsic motivation is somewhat mixed. On the one hand, Chirkov et al. (2007, 2008) found, for instance, that the effects of self-determined motivation on socio-cultural adaptation are not significant in the Chinese sample compared to the international sample. On the other hand, the relation between the self-determined motivation and well-being was found, but weaker for Asians than for Europeans. Research on the persuasiveness of messages has demonstrated that collectivistic advertisements are more appealing in collectivistic societies than in individualistic societies, but that under certain circumstances, members from collectivistic cultures find both individualistic and collectivistic messages appealing (Han and Shavitt, 1994). Taken together, there is evidence that intrinsic motivation might be more universal in its effect than assumed, and that only extrinsic motivation is especially effective in interdependent cultures. Thus, individuals with interdependent self-construal might benefit from both intrinsic and extrinsic motivation to be a group member. In other words, any kind of motivation would be useful when these newcomers experience incompatibility. In the context of international students, it is likely that the students were, one way or the other, motivated to live in the receiving society. There are, however, certainly contexts (e.g., individuals who take refuge from war, natural disasters, political persecution or poverty) where the motivation to come live in another country is less driven by the receiving society itself, but by a motive to leave the country of origin. The present research only indirectly captures the effects of amotivation and does not explicitly investigate the effects of low levels of both intrinsic and extrinsic motivation. It demonstrates, however, that it is worthwhile to investigate the effects of different kinds of motivations, and future research should broaden our knowledge about the relation between kinds of motivation (including amotivation), self-construal, incompatibility, and integration outcomes.

# Practical Implications

The current findings demonstrate that incompatibility is a risk factor for disidentification. Disidentification has disagreeable and potentially dangerous consequences for both the new group and the individual trajectories within the group. These findings have practical implications for the avoidance of undesirable negative integration processes. One way to minimize the risk of disidentification is to decrease incompatibility. This could be done by negotiating differences in acculturation attitudes, by encouragement of perspective taking, or by a focus on similarities rather than differences (compare Rosenthal and Crisp, 2006). Experiences of incompatibility are, however, likely to occur in the intercultural context and sometimes impossible to rule out (e.g., cultural distance). In these cases, the motivation of newcomers should be strengthened. The current research suggests that it is worthwhile to focus on both intrinsic and extrinsic motivation, because both kinds of motivation may help newcomers not to

be discouraged when they experience incompatibilities. Thus, interviewers in selection processes should ask both "Why do you want to be a part of the group?" as well as "What does your social context think about you're joining the group?". When individuals with interdependent self-construal express that they have been "sent" by, for instance, parents, Westerners should stop seeing these people as pitiful, hard-driven individuals that are not allowed to follow their nature, but consider that for individuals with interdependent self-construal, the goals and opinions of relevant others are just as motivating and gratifying as inner needs and wishes are for individuals with independent selfconstrual. For the training of newcomers, it seems promising to strengthen both kinds of motivation, for instance by having them recall their inner reasons for the group membership or think about significant others that are happy about their being in the group.

# Conclusion

The current study is the first to demonstrate the joint impact of incompatibility between different social identities and motivation in different cultures on the newcomers' social identities. It thus contributes to research on multiple identities as well as motivation and cross-cultural research. The experience of incompatibilities between pre-existing and new social identities is a normal and probably unavoidable part of an integration process. The current research demonstrates that individuals have resources to stay on track and refrain from disidentification despite these difficulties. These resources are, however, not the same for everyone. It is thus a fruitful research avenue to investigate what resources are effective in whom to improve the integration processes when social identities turn out to be incompatible.

# REFERENCES


# AUTHOR CONTRIBUTIONS

CM lead the conception and design of the work, the data collection, conducted the analyses and lead the interpretation of the data, and drafted the manuscript. JF contributed substantially to the analyses and the interpretation of the data, and revised the manuscript critically. CM and JF approve the final version of the manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

# FUNDING

This work and the preparation of this article was supported by the German Federal Ministry of Education and Research and the European Social Fund (grand number: 01PF 08005 C), as well as the European Social Fund in Baden-Wuerttemberg. The funding sources were not involved in decisions on the preparation, procedure, analyses or report of the present research.

# ACKNOWLEDGMENTS

We are thankful to the German Youth for Understanding for their cooperation in data-collection of Study 1. Solveig Schudeiske conducted the data collection of Study 2 in cooperation with Colleen Ward and her lab, and we thank Solveig Schudeiske, Colleen Ward and her lab in Wellington for their commitment and fruitful discussions on motivation and acculturation. Moreover, we thank Fabio Lorenzi-Cioldi and the two reviewers for their constructive feedback on an earlier version of the paper.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Matschke and Fehr. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Combining Gender, Work, and Family Identities: The Cross-Over and Spill-Over of Gender Norms into Young Adults' Work and Family Aspirations

Loes Meeussen\*, Jenny Veldman and Colette Van Laar

Center for Social and Cultural Psychology, Department of Psychology, University of Leuven, Leuven, Belgium

The current study investigates how descriptive and prescriptive gender norms that communicate work and family identities to be (in)compatible with gender identities limit or enhance young men and women's family and career aspirations. Results show that young adults (N = 445) perceived gender norms to assign greater compatibility between female and family identities and male and work identities than vice versa, and that young men and women mirror their aspirations to this traditional division of tasks. Spill-over effects of norms across life domains and cross-over effects of norms across gender-groups indicated that young women, more than young men, aimed to 'have it all': mirroring their career ambitions to a male career model, while keeping their family aspirations high. Moreover, young women opposed traditional role divisions in the family domain by decreasing their family aspirations in face of norms of lower family involvement or higher career involvement of men. Conversely, in line with traditional gender roles, young men showed lower family aspirations in the face of strong male career norms; and showed increases in their career aspirations when perceiving women to take up more family roles. Young men's family aspirations were, however, more influenced by new norms prescribing men to invest more in their family, suggesting opportunities for change. Together, these findings show that through social norms, young adults' gender identity affects aspirations for how to manage the co-presence of their work and family identities. Altering these norms may provide leverage for change to allow both men and women to combine their multiple identities in an enriching way.

Keywords: identity compatibility, identity management, gender, work, family, social norms

# INTRODUCTION

Many people on a daily basis manage the combination of two identities central in their lives: their work identity and their family identity. The co-presence of these identities may be experienced as conflictual or as enriching, with important consequences for well-being, health, and performance (Allen et al., 2000; Van Steenbergen and Ellemers, 2009). We argue that people's gender identity impacts the way men and women cope with the co-presence of work and family identities differently. More specifically, decisions about engaging in or refraining from work and family

#### Edited by:

Natasza Kosakowska-Berezecka, University of Gdañsk, Poland

#### Reviewed by:

Eric Mayor, University of Neuchâtel, Switzerland Michele Hoffnung, Quinnipiac University, USA

> \*Correspondence: Loes Meeussen loes.meeussen@kuleuven.be

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

Received: 18 August 2016 Accepted: 31 October 2016 Published: 17 November 2016

#### Citation:

Meeussen L, Veldman J and Van Laar C (2016) Combining Gender, Work, and Family Identities: The Cross-Over and Spill-Over of Gender Norms into Young Adults' Work and Family Aspirations. Front. Psychol. 7:1781. doi: 10.3389/fpsyg.2016.01781

roles are influenced by prevailing gender norms in one's social context – norms that indicate what roles are and should be for men and women, and as such, communicate to what extent male and female identities are (in)compatible with work and family identities (Eagly et al., 2000; Eagly and Karau, 2002). Advancing research that shows norms are highly influential within a gender group and life domain (Major, 1994; Wood and Eagly, 2009; Brown and Diekman, 2010), we argue that due to the intertwined nature of gender, work, and family identities, norms may 'spill-over' from one life domain to the other (e.g., family norms affecting career aspirations) and 'cross-over' from one gendergroup to the other (e.g., norms for men affecting women's aspirations).

Laying bare such inter-identity dynamics of norms allows us to identify potential barriers to and leverages for change toward greater compatibility between male and family identities, and between female and work identities, with various benefits. Increasing male opportunities in the family domain is beneficial for men's psychological and physical health for their children's upbringing, and for reducing women's second shift (for a review, see Croft et al., 2015). Increasing female opportunities in the work domain is beneficial for women's feelings of life-enrichment and well-being (Van Steenbergen et al., 2007; Jacob, 2008), for performance of organizations (Carter et al., 2003; Herring, 2009), and to decrease children's gender stereotypes (Davis and Greenstein, 2009).

# The Importance of Gender Norms: Communicating (in) Compatibility with Family and Work Identities

Gender norms indicate which roles are for men and which roles are for women (Eccles, 1994; Eagly et al., 2000; Eagly and Karau, 2002). With regard to work and family identities, normative beliefs persist that male identities are more compatible with work identities, and thus that men should be -or arebetter at being the provider; and that female identities are more compatible with family identities, and thus that women should be -or are- better at being the caregiver (Riggs, 1997; Etaugh and Folger, 1998; Eagly et al., 2000; Hodges and Park, 2013). Research has shown that such gender norms affect aspirations, decisions, and behavior as people internalize social norms (Major, 1994; Wood and Eagly, 2009) and experience social and personal benefits when they adhere to gender norms and punishments when they do not (Brown and Diekman, 2010; Diekman et al., 2011). Replicating such research, we expect that people's family and career aspirations will be influenced by norms for their own gender with regard to that same life domain. Extending this research, we argue that work, family, and gender identities are so intertwined that gender norms are not only influential within gender-groups and life domains, but also from one life domain to the other (referred to as spill-over effects, cf. Bakker and Demerouti, 2013) and from one gender-group to the other (referred to as cross-over effects).

# Gender Norms Spill-Over from One Life Domain to the Other

Many people manage both work and family identities on a daily basis, sometimes experiencing a conflicting and sometimes an enriching combination (Carlson et al., 2000; Hakanen et al., 2011). Given this close link between work and family identities, we argue that norms about the compatibility between one's gender and one life domain will also affect one's aspirations in the other life domain, indicating how people combine work and family identities. We predict that particularly for women, norms regarding one domain will reinforce aspirations in the other domain, as research indicates that women more than men experience the combination of a career and family identity as enriching and rewarding (Van Steenbergen et al., 2007). Moreover, societal changes lead women especially to increasingly consider "having it all," that is, pursuing a successful career and family life (Hoffnung, 2004). Indeed, in the last decades, women's career involvement has increased greatly (England, 2010, 2011), while at the same time women continue to have the main responsibility for family tasks (Hochschild and Machung, 2012; European Commission, 2014).

# Gender Norms Cross-Over from One Gender Group to the Other

In addition to spillover from one domain to another, we predict there also to be cross-overs from one gender group to another. Partners need to find a division of work and family tasks between them, since the less one partner does, the more the other will have to fill in (Belsky and Kelly, 1994; Biehle and Mickelson, 2012). As such, within heterosexual relationships male and female identities are very much intertwined (Elder, 1998). Given this close link, we argue that norms about the compatibility between one gendergroup and work or family will also affect the other gender group's aspirations.

First, gender norms for the other gender may cross-over to influence people's aspirations within one life domain, indicating that people may mirror themselves to roles for the other gender. For example, women may increase their career aspirations the more they perceive men to invest in their career. As gender roles are increasingly egalitarian (Botkin et al., 2000), the other gender becomes a more relevant comparison group for one's own aspirations. We predict that such relations will especially occur for women, since their roles have changed more toward those of men than vice versa (Eagly and Diekman, 2003; Croft et al., 2015). Moreover, matching their aspirations to male norms means an opportunity for self-improvement for women (Corcoran et al., 2011) since roles typically occupied by men tend to be higher in status and authority than roles typically occupied by women (Eagly et al., 2000; England, 2010, 2011).

Second, gender norms for the other gender in one domain may cross-over to influence one's aspirations in the other domain, indicating how work and family roles are negotiated between partners. For example, men may increase their career aspirations the more they perceive women to invest in their family (i.e., influence in traditional direction) or women may decrease their family aspirations the more they perceive men to invest in their careers (i.e., influence in non-traditional direction). We predict

that women more than men will oppose traditional gender divisions, because such divisions are most disadvantageous for them in a society in which women's career involvement has greatly increased while men's involvement in family tasks has remained behind, thus saddling women with a second shift (Bianchi et al., 2000; Laflamme et al., 2002; Pleck and Masciadrelli, 2004; Bartley et al., 2005; Hochschild and Machung, 2012). Conversely, men are likely to oppose traditional divisions less, or even follow them, since they experience less disadvantages of such divisions and may thus be more likely to follow systemjustifying beliefs that what "is" is what "ought" to be (Heider, 1958; Sidanius and Pratto, 1999; O'Brien and Major, 2009).

To examine these relationships, we had young adults anticipate their life as working parents and respond to how they imagined their opportunities and responsibilities in the family and work domains. Young adulthood (18–30 years old; Rindfuss, 1991) is a crucial life stage in which decisions are made regarding ways of managing the combination of work and family identities (Brown and Diekman, 2010; Weisgram et al., 2010), and a stage in which instigating possible change toward gender equality in the future is most likely. We examined both descriptive and prescriptive norms. Descriptive norms indicate what men and women currently do (Eagly, 1987; Heilman, 2001, 2012; Prentice and Carranza, 2002). They reflect what young adults see in their social context: to what extent do men and women currently take up family and career roles? We included descriptive norms as people's behavior is shaped by what they encounter in their social surrounding (Oishi and Graham, 2010), and young adults' choices regarding families and careers are influenced by what they have seen around them (Croft et al., 2014; Gerson, 2010). Prescriptive norms meanwhile indicate what men and women should do (Eagly, 1987; Heilman, 2001, 2012; Prentice and Carranza, 2002). They reflect what is deemed desirable: to what extent men and women should invest in work and family roles. Both descriptive and prescriptive norms are important, since what people should do according to their social environment may differ from what people actually do. Also, what people see around them and what they believe to be appropriate may both push social change, and this change could push toward more traditional or more egalitarian gender roles.

# MATERIALS AND METHODS

# Participants

Participants were young adults at the University of Leuven in Belgium who participated in an online study about 'their present and future life.' Belgium can be seen as an average Western country in terms of gender equality: Its gender equality index is comparable to other Western European countries such as France or Germany, and in-between Eastern European countries scoring in the lower range and Scandinavian countries scoring in the higher range of gender equality (European Institute for Gender Equality, 2015). Compared to the United States, Belgium scores somewhat higher on gender equality with a 10th place worldwide based on the global gender gap index (Unites States 20th place; World Economic Forum, 2014) and a 9th place worldwide based on the gender inequality index (United Stated 47th place; United Nations Development Programme, 2014).

Participants took part in this study in exchange for course credits or a chance to win a number of €9–35 prizes in a lottery. Participants were recruited through student mailing lists, social media, flyers in university buildings, and promotion talks during lectures at various faculties. Of the 522 students who completed the survey instrument, a number were excluded as they did not fit the population of interest: seven were excluded from the sample because they were over 30 years old and hence no longer fit the group of young adults (Rindfuss, 1991); 37 were excluded as they indicated not being heterosexual (which meant gender norms would not be expected to cross-over for them in the same way), and 33 were excluded because they indicated that they did not want to have children (which made the family aspiration measures about childcare not applicable to them). Of the resulting sample of 445 participants, 37.8% were male and 62.2% female with a mean age of 19.3 (SD = 1.92). The sample reflected the typical university campus: Participants came from a wide variety of majors, such as law, economics and business, engineering, medicine, humanities, and social sciences. Most participants self-identified as Belgian (93.4%), and 19.1% (also) identified with another group (mostly other European countries).

# Measures

All items were answered on a 7-point Likert scale from (1) strongly disagree to (7) strongly agree. Measures were scored such that higher scores indicate stronger scores on the concept.

## Own Family and Career Aspirations

To assess own future family and career aspirations, participants were asked to imagine their life in 15 years. Given that participant's mean age was 19 years, this meant that they indicated how they saw their family and career investment around the age of 34, an age at which many people in the Western world are combining a career with a household and taking care of one or more young children (Mathews and Hamilton, 2009). Participants' own family aspirations were measured with four items (two childcare items, two household items): '15 years from now. . . I will make time to take care of all the needs of my children'; 'I will not really concern myself with my children's upbringing' (reversed); 'I will spend considerable time making sure that the household is in order'; and 'I will spend little time on household tasks' (reversed). Own career aspirations were measured with two items '15 years from now, I will spend time on building a successful career' and '15 years from now, I will spend little time on my career' (reversed).

## Descriptive Norms Regarding Family and Career Investment

Descriptive norms about men's family investment, men's career investment, women's family investment, and women's career investment were measured with the same items as for own aspirations but rephrased to inquire about what the average man or woman in Belgium does (cf. Gill, 2004). For instance, an item measuring descriptive norms regarding men's family investment was 'The average man in Belgium spends little time on household

chores' (reversed) and an item measuring descriptive norms regarding women's career investment was 'The average woman in Belgium spends time on building a successful career.' It was emphasized that we were interested in what participants believed the average man or woman in Belgium did, not what they thought about this personally.

#### Prescriptive Norms Regarding Family and Career Investment

Similarly, the same items were adapted to measure prescriptive family and career norms for men and for women by asking what they should do (cf. Heilman, 2001, 2012). Every item was measured twice, asking both what men prescribe (e.g., 'The average man in Belgium believes that men should make time to take care of all the needs of their children') and what women prescribe (e.g., 'The average woman in Belgium believes that men should make time to take care of all the needs of their children'). The norms for women (prescribed by women and men) were then combined into one scale, as were the norms for men (prescribed by women and men). In this way, we ensured that the items measured consensual norms rather than participants' own opinion as to what men and women should do.

**Table 1** provides reliability scores, means, standard deviations, and correlations between all measures. Items were worded rather strongly to ensure it would not be too easy to fully agree or disagree, thus avoiding ceiling or floor effects. The means on the scales (ranging between 3.88 and 5.67 on 7-point scales) indicated this was successful. As described above, for each measurement two items were used (one of which was always a reversed item): two items that measured household investment, two items that measured childcare investment, and two items that measured career investment. As exploratory factor analyses for all of the measures (i.e., own aspirations, descriptive and prescriptive norms for men and women) revealed a two-factor solution with career investment on the one hand and childcare and household investment on the other (57.15–63.89% explained variance), childcare and household investment were combined into one family investment scale.

# RESULTS AND DISCUSSION

# Traditional and Changing Gender Norms

Before looking into the relation between gender norms and young adults' own aspirations, we examined how young males and females perceived descriptive and prescriptive gender norms regarding family and career identities. Supporting the notion that consensual norms rather than personal opinions or preferences were being assessed, within-group reliability indices were all above the 0.70 criterion (James et al., 1984; LeBreton and Senter, 2008) indicated that young adults (MRwg = 0.82, SD = 0.03), as well as young males (MRwg = 0.79, SD = 0.02) and young females separately (MRwg = 0.83, SD = 0.03), showed considerable consensus on what the descriptive and prescriptive norms are for what men and women do and should do in terms of family and work. **Table 2** presents differences and similarities between young males' and females' in their perceptions of the


TABLE 1 | Reliability

 scores, means, standard

deviations,

 and correlations

 between all measures

 for male and female participants.

TABLE 2 | Difference tests investigating how descriptive and prescriptive norms about career and domestic investment differ from one another for men and women (bottom half of table), and in the perception of male and female participants (final column top half of table).


∗∗∗p < 0.001, ∗∗p < 0.01, <sup>∗</sup>p < 0.05, †p < 0.10.

norms (independent samples t-tests: top half of table), as well as differences between descriptive and prescriptive norms (paired samples t-tests: bottom half of table).

Specifically, the results showed that generally, young males and females had a relatively traditional consensus about patterns in descriptive and prescriptive norms. Both young males and females believed that norms show greater compatibility between male and work identities and between female and family identities: As indicated in the lower half of **Table 2**, they believed norms to indicate that men currently do (difference #1) and should (difference #5) invest more in their career than in their family, and that women currently do (difference #2) and should (difference #6) invest more in their family than in their career. Similarly, the task-division between partners was perceived to be rather traditional, as young adults believed norms to indicate that men currently do (difference #4) and should (difference #8) invest more in their career than women, and that women currently do (difference #3) and should (difference #7) invest more in their family than men. Moreover, differences between young males and females' norm perceptions (top half **Table 2**, last column) indicated that young females estimated men's current career investment significantly higher than young males do, and they perceived prescriptive norms to indicate more strongly that men should invest in a career and that women should invest in a family. Thus, young females perceived gender norms to be even more traditional than young males, which could be explained by the fact that women experience more disadvantages from traditional gender divisions than men (Eagly et al., 2000; England, 2010, 2011; Gerson, 2010).

Importantly, there was also some indication of change toward greater gender equality when comparing descriptive with prescriptive norms: Young males and females believed that norms indicate that men should invest more in family tasks than they currently do (bottom half of **Table 2**, difference #9), and young men believed that norms indicate women should invest less in family tasks (difference #11) and more in their career (difference #12) than they currently do. Thus, both young males and females appear ready for some movement toward less traditional role divisions; which could be fostered so that this is translated into actual behavior when these young adults become parents.

# Gender Norms and Young Adults' Own Family and Career Aspirations

We then examined the relationship between gender norms and young adults' aspirations. For this, we conducted hierarchical regressions for young males and young females separately, predicting their family or work aspirations (while controlling for work and family aspirations, respectively) from (1) norms for their own gender within the same life domain; (2) norms for their own gender in the other life domain (i.e., spill-over); and (3) norms for the other gender (i.e., cross-over). **Tables 3** and **4** present the results of the final regression models for career and

TABLE 3 | Hierarchical regression models predicting young male and female participants' career aspirations from different gender norms (standardized coefficients).


∗∗∗p < 0.001, ∗∗p < 0.01, <sup>∗</sup>p < 0.05, †p < 0.10.

TABLE 4 | Hierarchical regression models predicting young male and female participants' family aspirations from different gender norms (standardized coefficients).


∗∗∗p < 0.001, ∗∗p < 0.01, <sup>∗</sup>p < 0.05, †p < 0.10.

family aspirations, respectively, as well as R square difference tests for each step. While gender norms were related to one another (as would be expected), VIF indices between 1.20 and 2.45 and Tolerance indices between 0.41 and 0.83 indicated no multicollinearity concerns (cf., Dormann et al., 2013).

### Norms for Own Gender within Life Domains: Young Adults Mirror Their Aspirations to (Traditional and Changing) Gender Norms

Looking at the role of norms for one's own gender within life domains, the results showed that young adults' career aspirations were related to career norms for their own gender: As **Table 3** shows, the more young males believed that men should invest in their career, the more they aspired to invest in their career themselves (β = 0.41 p < 0.001). The more young females perceived women to currently invest in their career, the more they aspired to invest in their career themselves (β = 0.23, p = 0.001). Similarly, as shown in **Table 4**, young adults' family aspirations were related to family norms for their gender: The more young females perceived women to currently invest and the more they perceived norms to indicate they should invest in their family, the more they aspired to invest in their family themselves (β = 0.37, p < 0.001 and β = 0.19, p = 0.019, respectively). Also, the more young males believed that men should invest in their family, the

more they aspired to invest in their family themselves (β = 0.27, p = 0.004).

Thus, young adults mirrored their own aspirations to rather traditional gender norms, indicating that gender equality in both the family and career domain may not occur rapidly unless change is fed. The findings point to opportunities for change especially in the family domain: young men's own family aspirations were related to prescriptive rather than descriptive norms. Given that both young males and females perceived norms to ascribe that men should invest more in their family than what they currently invest, gender equality in the family domain is likely to be improved when such changing gender norms are made salient. Also, the advantages of men taking part in family tasks for men's psychological and physical health, for women's well-being, for relationship satisfaction, and for children's adjustment (Biehle and Mickelson, 2012; Croft et al., 2014, 2015) could be highlighted and communicated more. More generally, since young adults mirror their own aspirations to the norms which they perceive for their own gender-group, investing in altering their perceptions of gender norms to be less traditional is likely to feed social change. For instance, such change may be initiated by more non-traditional gender role models in education and the mass media (Hogben and Waterman, 1997; Greenwood and Lippman, 2009) and through children's home environment (Goldberg et al., 2012; Croft et al., 2014).

### Spill-Over Effects of Norms between Life Domains: Young Women But Not Young Men Want it All?

Next, we examined spill-over effects of own-gender norms on young adults' aspirations from one life domain to the other. Results showed that family aspirations were positively related to own-gender career norms for women, while they were negatively related for men: That is, the more young women perceived women to currently invest in a career, the more they aspired to also invest in family tasks (β = 0.16, p = 0.015). Conversely, the more young men perceived men to currently invest in a career, the less they aspired to invest in family tasks (β = −0.19, p = 0.040). This gender difference was also visible in the relations between family aspirations and career aspiration, which were significant and positive for women and non-significant and negative for men.

These findings suggest that young women, more than or even in contrast to young men, 'want it all' (cf. Hoffnung, 2004): they want to pursue high aspirations in the family as well as in the career domain. This corresponds to the societal trend that while their participation in the labor market and 'agentic' type tasks has increased substantially, women continue to have the main responsibility for family and communal tasks (Diekman et al., 2011; Hochschild and Machung, 2012; European Commission, 2014). On the one hand, wanting it all may be beneficial for women, given that they experience the combination of work and family roles as even more enriching than do men (Van Steenbergen et al., 2007). On the other hand, these high aspirations may also make women more vulnerable to workfamily conflict once they enter parenthood (Nomaguchi and Milkie, 2003; Hodges and Park, 2013).

While young women's family aspirations were positively related to female career norms, their career aspirations were not related to female family norms. This difference may be a result of high expectations for mothers (Thurer, 1995; Hays, 2003; Johnston and Swanson, 2006): a strong focus on women's careers may bring about feelings of guilt and hence push aspirations for their mother identity, while a stronger focus on family does not bring about the same guilt for their work identity and thus does not push career identity (Rotkirch and Janhunen, 2010; Liss et al., 2013).

Young males' family aspirations were negatively related to male career norms, but their career aspirations were not related to male family norms. This suggests that when experiencing conflict between family and career roles, young males would follow more traditional gender role patterns, decreasing family aspirations for career aspirations rather than the other way around. This is consistent with research showing that men are penalized for decreasing career time for family reasons (Rudman and Mescher, 2013).

#### Cross-Over Effects of Other Gender Norms within Life Domains: Young Women's Aspirations Follow Male Norms

We next examined cross-over effects of norms by investigating relationships between young males' and females' aspirations and norms for the other gender. With regard to cross-over effects within one life domain, as expected young females' but not young males' aspirations were related to gender norms for the other gender. Specifically, the results show that young females' family aspirations were higher the more they believed men currently invested in family tasks (β = 0.17, p = 0.006). This may suggest that young females aim for a 'tit for tat' strategy toward a more equal partner division of family tasks in a world where men spend only half the time on family tasks women do (Bianchi et al., 2000; Bartley et al., 2005). Moreover, young females related their career aspirations to a more male career model: The more they believed norms to ascribe men should invest in a career, the higher their own career aspirations (β = 0.28, p = 0.001)<sup>1</sup> . This also fits with the notion that male career roles tend to be higher in status than female roles, and therefore, matching their aspirations to male career norms also means an increasing opportunity for self-improvement for women (Eagly et al., 2000; England, 2010, 2011).

#### Cross-Over Effects of Other Gender Norms between Life Domains: Women More Than Men Oppose Traditional Task-Divisions

With regard to cross-over effects between life domains, opposite patterns emerged for young males and females: The more young females believed norms to ascribe men should invest in a career, the less they themselves aspired to invest in family tasks (β = −0.17, p = 0.034) – and the more they also aspired a career, as described above. This suggests women may increasingly

<sup>1</sup>The marginally significant relation between descriptive career norms for men and young female's career aspirations showed a similar trend: the more women perceived men to currently invest in a career, the higher their own career aspirations (β = 0.11, p = 0.091).

actively oppose traditional role divisions when males have a strong emphasis on careers. Conversely, the more young males believed women invest in family tasks, the more they aspired to invest in their career (β = 0.20, p = 0.046), reinforcing a traditional role division with men as providers and women as caregivers, in line with system-justifying beliefs that what "is" is what "ought" to be (Heider, 1958; Sidanius and Pratto, 1999; O'Brien and Major, 2009). Thus, relating these patterns to the descriptive results, young females perceived norms to be more traditional and they opposed traditional role divisions more than young males<sup>2</sup> . These gender differences correspond to the fact that women experience more disadvantages from traditional gender divisions than men (Eagly et al., 2000; England, 2010, 2011; Gerson, 2010); and they provide opportunities for change: Fostering young women's resistance to traditional gender divisions by decreasing family aspirations may decrease a second shift and maternal gatekeeping behaviors for women (Allen and Hawkins, 1999; Hochschild and Machung, 2012); and our results indicate that this, in turn, could also change men's focus on career over family roles.

More generally, the results showed that young females' aspirations more than young males' were related to prevailing gender norms (i.e., more significant relationships and higher R squares among females- thus more variance in their aspirations explained by gender norms). This is in accordance with previous research showing that especially young females assume that the choices they make need to fit their social context (Oyserman and James, 2011). This difference between young males and females was also due to the fact that women's aspirations were influenced more by male gender norms than vice versa, which, as noted, may reflect an opportunity for self-improvement for women (Eagly et al., 2000; England, 2010, 2011).

# Limitations and Directions for Future Research

The current study investigated how descriptive and prescriptive gender norms that communicate work and family identities to be (in)compatible with gender identities limit or enhance young men and women's family and career aspirations. Advancing research that shows norms are highly influential within a gender group and life domain (Major, 1994; Wood and Eagly, 2009; Brown and Diekman, 2010), we argued that due to the intertwined nature of gender, work, and family identities, norms may 'spill-over' from one life domain to the other (e.g., family norms affecting career aspirations) and 'cross-over' from one gender-group to the other (e.g., norms for men affecting women's aspirations). Together, the findings show that through social norms, young adults' gender identity affects aspirations for how to manage the co-presence of their work and family identities. Altering these norms may provide leverage for change to allow both men and women to combine their multiple identities in an enriching way.

One potential limitation of the current study is that we cannot automatically generalize data collected in a particular norm setting to other nations or norm settings. However, as indicated earlier, Belgium is comparable to other Western countries in terms of gender inequality (United Nations Development Programme, 2014; World Economic Forum, 2014; European Institute for Gender Equality, 2015). Also, while the levels of young adults' aspirations and gender norms may differ between countries (Crompton and Lyonette, 2006; Fahlén, 2014), the relations between norms and aspirations – the main focus of this paper – are likely to be generalized more broadly.

In this paper, we studied young adults' aspirations for their future work and family life. While we believe young adults to be an important sample because decisions regarding work and family life are shaped before people actually become working parents, we do not wish to imply that these aspirations will necessarily directly translate into the future lives of these individuals. Young adults' hopes, choices, and decisions are not fixed but continue to change, shaped by personal attitudes and factors in their social environment (cf. Hoffnung and Williams, 2013), an important one of which, as we show, is gender norms. Also, many of these young adults may not yet have a very clear notion of how they will balance their family and career responsibilities and how they will negotiate tasks with their partner. Therefore, we did not ask participants directly about the challenges in combining or dividing tasks, but more implicitly about their aspirations and the norms in their environment to then infer the combinations of life domains and partners from the spill-over and cross-over analyses. The current sample were young adults attending university, a more highly educated segment of the general population. As is typically the case for higher educated groups, this sample is likely to represent where gender role change or maintenance is headed. Our results indicated that these young adults still showed attachment to traditional gender roles, as well as clear elements of change.

A limitation of the current study is its correlational design, which does not allow for firm causal conclusions from the results. While it is possible that young adults' family and career aspirations are projected onto rather than influenced by their perception of gender norms, this alternative explanation is less likely for a number of reasons: First, the finding that young males and females showed strong consensus in their perceptions of gender norms but show very different relations between these norms and their own aspirations speaks against the idea that young adults would base their perception of gender norms on their own aspirations. Similarly, the spill-over and cross-over effects of norms are hard to explain through reverse causation. For instance, if young females project their career aspirations onto gender norms, then it is unlikely that they would also project their aspirations onto career norms for men instead of only on career norms for women. Moreover, the fact that gender norms affect people's own aspirations and behavior has

<sup>2</sup>Three effects did suggest a trend away from traditionalism for young males, though not significant at the 0.05 level: Young males' family aspirations were higher the more they perceived women to currently invest in a family (β = 0.18, p = 0.071; suggesting a tendency 'to do their part') and the more they perceived women to currently invest in a career (β = 0.16, p = 0.057; suggesting they would want to enable a future partner's career). Moreover, the more young males perceived norms to indicate women should invest in a family, the lower their own career aspirations (β = −0.19, p = 0.085; suggesting a trend to oppose traditional divisions, similar to our findings for young females).

been well documented in previous research (Major, 1994; Brown and Diekman, 2010; Diekman et al., 2011). Still, future studies could further investigate the causal link between gender norms and young adults' aspirations using longitudinal or experimental designs. For instance, it would be fruitful to investigate whether the relationships that suggest leverage for change toward gender equality can be manipulated: Do men increase their family aspirations when they are informed that changing norms ascribe men to do so? Do they become more interested in family roles when the status of such roles is highlighted? Do women decrease their family aspirations when norms ascribing a strong career focus for men are highlighted? Moreover, one could examine actual partner negotiations with regard to family and work responsibilities: How do partners come to a particular division of tasks and to what extent are gender norms implicitly or explicitly at play during such negotiations? We are examining these questions in ongoing research.

# CONCLUSION

In sum, this paper examines how young men and women expect to fill in their future multiple identities, and how this relates to the norms they perceive in the environment around them. The results showed that young men and women's management of the co-occurrence of their work and family identities is differentially influenced by norms for their and the other gender identity. While norms were perceived to be rather traditional, results also revealed several areas in which there is leverage for change toward gender equality in both the family and work domains, such that men and women can regulate their gender, professional, and family identities as an enriching rather than conflicting combination.

# REFERENCES


# ETHICS STATEMENT

The study has been approved by Social and Societal Ethics Committee, University of Leuven, Belgium. Before starting the questionnaire, participants agreed to an informed consent that provided the general research aims. They were informed that participation was voluntary and could be stopped at any moment during the study; and that their responses are anonymous and treated confidentially. Moreover, they were provided with room for questions and comments as well as all contact information of the researchers and the ethical committee – and they received a full debriefing at the end of the study.

# AUTHOR CONTRIBUTIONS

LM, JV, and CvL contributed to the development of hypotheses, data collection, interpretation of results, and this manuscript. LM conducted the statistical analyses.

# FUNDING

This research was supported by an Odysseus grant to CvL from the Research Foundation of Flanders (FWO) grant number G.O.E66.14N.

# ACKNOWLEDGMENT

We thank Toni Schmader, Alyssa Croft, Katharina Block, Sarah Martiny, and Carolin Schuster for their valuable feedback in the development of this project.


careers: evidence for a goal congruity perspective. J. Pers. Soc. Psychol. 101, 902–918. doi: 10.1037/a0025199



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Meeussen, Veldman and Van Laar. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Role of Indian Caste Identity and Caste Inconsistent Norms on Status Representation

#### Sindhuja Sankaran<sup>1</sup> \*, Maciek Sekerdej<sup>2</sup> and Ulrich von Hecker<sup>3</sup> \*

<sup>1</sup> Department of Psychology, University of Warsaw, Warsaw, Poland, <sup>2</sup> Institute of Psychology, Jagiellonian University, Kraków, Poland, <sup>3</sup> School of Psychology, Cardiff University, Cardiff, UK

The Indian caste system is a complex social structure wherein social roles like one's profession became 'hereditary,' resulting in restricted social mobility and fixed status hierarchies. Furthermore, we argue that the inherent property of caste heightens group identification with one's caste. Highly identified group members would protect the identity of the group in situations when group norms are violated. In this paper, we were interested in examining the consequence of caste norm violation and how an individual's status is mentally represented. High caste norms are associated with moral values while the lower caste norms are associated with immorality. We predicted a 'black sheep effect,' that is, when high caste individuals' group identity (caste norm violation condition) is threatened their salient high caste identity would increase, thereby resulting in devaluing the status of their fellow in-group member if the latter is perceived as perpetrator. We presented participants with a social conflict situation of a victim and a perpetrator that is 'Caste norm consistent' (Lower caste individual as a perpetrator and higher caste individual as a victim) and vice versa 'Caste norm inconsistent' condition (higher caste individual as perpetrator and lower caste individual as a victim). Then, participants had to choose from nine pictorial depictions representing the protagonists in the story on a vertical line, with varying degrees of status distance. Results showed evidence for the black sheep effect and, furthermore, revealed that no other identity (religious, national, and regional) resulted in devaluing the status of fellow in-group member. These results help us understand the 'black sheep' effect in the context of moral norms and status representation and are discussed in the framework of the Indian society.

Keywords: black sheep effect, caste identity, norm-violation, social identity threat, status

# INTRODUCTION

People in general belong to many social categories that could either be achieved, such as one's profession, or inherited, such as one's gender. The consequences of social categorizations are often not only seen in the dynamics of social interactions, but also in the way social status is represented. For the present research, the Indian/Hindu caste system is of interest, which is an integral feature of the Indian societal structure. The caste system provides a hierarchy of social roles that hold inherent characteristics and, more importantly, remain stable throughout life (Dirks, 1989). An implicit status is attached to one's caste which historically changed from the social roles to hereditary

#### Edited by:

Natasza Kosakowska-Berezecka, University of Gdañsk, Poland

#### Reviewed by:

Marika Rullo, University of Lausanne, Switzerland Peter Bevington Smith, University of Sussex, UK

#### \*Correspondence:

Sindhuja Sankaran sindhuja.p.s@gmail.com Ulrich von Hecker vonheckeru@cardiff.ac.uk

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

> Received: 11 November 2016 Accepted: 15 March 2017 Published: 31 March 2017

#### Citation:

Sankaran S, Sekerdej M and von Hecker U (2017) The Role of Indian Caste Identity and Caste Inconsistent Norms on Status Representation. Front. Psychol. 8:487. doi: 10.3389/fpsyg.2017.00487

roles. This, created status hierarchies on hereditary basis with limited social mobility. For instance, individuals born into the highest caste, that is, the Brahmin caste have usually been priests and scholars. Individuals born into the Kshatriya caste have been warriors and kings. Individuals born into the Vaishya caste have been merchants. Finally, individuals born into the Shudra caste have been laborers. Besides, there was an additional 'out-casted' group called the Dalits or the 'untouchables' who occupied the lowest step of the social ladder (see Ambedkar, 1925/1989; Pick and Dayaram, 2006). In modern India, the Indian government introduced a categorization scheme in which the untouchable castes were categorized as scheduled castes (SC), the backward tribes were categorized as scheduled tribes (ST) and the disadvantaged castes as other backward castes (OBC). The Forward caste (FC) community generally constitute the high caste group. The SC, ST, and OBC comprising the historically disadvantaged groups, were provided job opportunities by the government through affirmative action (Sheth, 1987; Kumar, 2001; Gupta, 2005; Dreze and Khera, 2009). The FC has historically been and, continues to be, in a strong socioeconomic position with the highest status in society<sup>1</sup> . Thus, one of the main objectives of the present research was to examine how status is cognitively represented in the Indian society as a consequence of the way caste is perceived<sup>2</sup> . Even now, people in India continue to define their self-identity by means of the caste they belong to and the social group that they find themselves in. Caste membership is thus ingrained in the society and there is considerable reason to claim that caste as a type of social identity would probably be one of the most salient identities in the Indian context. This aspect is addressed by Social Identity Theory (Tajfel and Turner, 1986), to which we now turn.

# Social Identity as a Basis for Caste Identity

Social identity claims that people derive an important part of their identity from an affirmation of membership with the group they belong to. Tajfel and Turner (1986)suggested that any group (e.g., social class, family, football team etc.) can act as a source of pride and self-esteem, therefore, we tend to enhance our self-esteem by promoting and endorsing the status of the group we belong to, the so-called "in-group" (as opposed to "out-groups" being those groups that we do not belong to, see also Hogg and Turner, 1987). The Indian societal structure provides a fertile ground to examine the interactive roles of multiple identities like religious, national, regional (north vs. south), class and caste wherein one could discard or fuse these identities for the benefit of societal functioning (Miller et al., 1990; Miller and Bersoff, 1992). But many researchers have stressed the importance and the influence of caste as an integral social identity among many South Asians compared to other social identities like gender and ethnicity (for example, Gayer, 2000; Mand, 2006). It has in fact been argued that caste identity may override other social identities, because of its primary importance for many South Asians (Judge and Bal, 2008). We argue that in the context of status representation, caste identity (as opposed to religion, national and regional identities) would be the most prominent identity in explaining the differences in status perception, due to the inherent associations of caste and status. Thus, according to social identity theory, individuals would strive to maintain a positive image of their caste identity. We further argue below that caste identity will especially be more salient for high caste individuals.

A strong caste identity could provide feelings of belongingness or self-esteem, thereby relying on some caste norms. Particularly, it is known that high caste individuals see caste identity as a more stable construct wherein this identity is inherited at birth. They tend to essentialise their identity and this is predominantly attributed to the feelings of connectedness with previous generations of one's caste group. High caste individuals also develop feelings of temporal continuity, positive distinctiveness, and heightened self-esteem from essentialisation of their caste identity (Jaspal, 2011). In fact, in a study conducted by Cotterill et al. (2014), it was argued that the caste system tends to be legitimized through the ideology of Karmic beliefs (beliefs that general good or bad deeds in one's life are rewarded or reprimanded by being born into a high or low caste in the next life) especially by those high on social dominance orientation (SDO), that is, those who demonstrate a general preference for hierarchical social relations (Sidanius and Pratto, 1999; Pratto et al., 2000). Furthermore, when members of higher castes essentialise their caste identity (Mahalingam, 2003) they permit themselves to stigmatize members of the lower castes. The low caste members or the Dalits on the other hand, do not believe that their caste identity is inherited and therefore do not essentialise it. They may thus enhance their self-efficacy, through the possibility of social mobility, based on the idea that caste identity can be seen as less permanent (Mahalingam, 2003). We thus argue that caste identity is more salient amongst high caste individuals due to the belief they have about being privileged to have inherited this positive image of high caste at birth. Low caste individuals would not have a salient caste identity because they believe that this identity is not essentialised and belonging to this group has negative consequences.

# Social Identity Threat and Caste Norms

Social identity effects are based on the protection of self-concepts (Tajfel and Turner, 1986) and thus any threat to this self-concept would be associated with strong identity effects. Research has shown that highly identified group members would find ways to protect their in-group identity (see Spears et al., 1997). However, Branscombe et al. (1999) claim that threat to one's social identity in fact depends on the degree of group identification. For

<sup>1</sup>Caste is not always related to socioeconomic class even though the two constructs are connected. For instance, a high caste and a high class individual could be a university professor while a high caste and a low class individual could be a Brahmin temple priest. Similarly, a politician could be a low caste and a high class individual and a typical example of a low caste, low class individual would be household help or a flower seller.

<sup>2</sup>The terms backward and forward are used based on the official classification issued by the government of India and recognized in the Constitution of India. Essentially the backward caste groups comprise of historically disadvantaged people in India while the forward caste comprise those of the more privileged high caste individuals. These terms by no means denote in the linguistic sense of moving forward or backward.

instance, they suggest that those who are highly identified with their in-group are more likely to show defensive responses than those who are not so highly identified. We can assume that high caste individuals who legitimize their inherent high caste would also show strong high caste identity.

So, what specifically could elicit an identity threat related to caste? We claim that norms and expectations that are associated with caste membership, when questioned, could fundamentally be a source of threat. In fact, it is most commonly seen that a person engaging in any sort of norm violation (especially of the higher caste) is ostracized and devalued (Mahalingam, 2007). One of the most deeply rooted caste norms relates to marriage. For instance, when people violate the norm of marrying within one's own caste by engaging in inter-caste marriage, the higher caste individual is believed to bring shame to the family and this norm transgression is considered to be immoral. Branscombe et al. (1999) argued that when an identity related to the morality value is threatened, high identifiers will show more defensive reactions. We therefore argue that the threat to one's own caste, if related to moral values or norms would motivate strong caste identifiers to alleviate this threat and protect their identity.

For many years, the high caste members in general had greater status in the society, and viewed themselves as living to higher moral standards and values, as compared to low caste individuals (Mahalingam, 2003). It is generally believed that high caste individuals hold qualities related to wisdom, intelligence, honesty, austerity, and morality while low caste individuals possess qualities of dullness, stupidity, immorality, impurity, and other negative qualities (Deshpande, 2010). These ancient established norms carried over into modern day Indian society and thus certain norms were explicitly attached to a caste type. Thus we can argue that morality is perhaps a significant value attached to one's caste and violating such a norm could be a source of threat particularly amongst high identifiers. As introduced earlier, inter-caste marriages can be seen as a typical norm-violation in India and are often viewed as 'polluting' the sanctity of the caste system (Dube, 2001), thereby touching upon the value of morality. Marriages between high and low caste persons are especially harshly punished and sometimes lead to public lynching of couples or their relatives, murder (of the bride, groom or their relatives), rape, public beatings and other sanctions (NYU 2007, p. 11 as cited in Hoff et al., 2009). In fact, in Northern India, inter-caste marriages frequently result in family members choosing to kill the couple (Flintoff, 2010; Goli et al., 2013). Thus, when a high caste member commits norm violation s/he is devalued in society. This effect can especially be understood by the 'Black sheep effect' (BSE) wherein people in general derogate deviant in-group members (Marques and Paez, 1994).

# Norm Violation Effects and Identity

When a norm is violated, members often perceive this deviant behavior as potentially threatening to the group identity, and therefore deal with the deviance in order to reduce the threat (Jetten and Hornsey, 2014). However, research has shown that the tendency for a group to defend the threat depends on the extent to which an individual is identified with the group (Marques and Paez, 1994). Those who are not as much identified with the group, are typically less motivated to protect one's social identity (Spears et al., 1997; Rimal and Real, 2005). It can thus be understood that high identifiers would show greater motivation to engage in in-group protection to defend the threat (Doosje et al., 1995). We argue that high caste individuals would be high identifiers with their caste, and low caste individuals would be low identifiers with their caste. However, we claim that in-group identity protection will be seen in the form of black sheep effect and not as in-group favoritism. In certain situations, in-group members are known to exclude undesirable members from the in-group in order to maintain a positive and distinctive social identity (Marques and Paez, 1994). For instance, research by Otten (2009) claims that an aggressive social interaction between a victim and a perpetrator would lead to generally biased responses that could either lead to in-group favoritism or black sheep effect; the latter effect being most likely to occur in situations. More specifically it is said that in-group favoritism is particularly observed when the deviant behavior of the perpetrator was ambiguous or unintentional (Duncan, 1976; Sagar and Schofield, 1980). However, when there is explicit evidence suggesting that in-group perpetrators deliberately "committed the crime" (see also Abrams et al., 2000; Marques et al., 2001) one would observe the black sheep effect. Wang et al. (2016) also found neural evidence showing that intentional aggressive interactions result in patterns of the black sheep effect. Thus, there is some evidence indicating that aggressive, intentional, and unambiguous interactions would lead to more in-group derogations.

Furthermore, this pattern of in-group derogation tends to be more distinct among individuals who are highly identified with their group (Biernat et al., 1999) than those who are not. High-caste individuals would indeed be high identifiers, owing to the notion of 'being born into' one's high caste and thus would be especially motivated to protect one's ingroup by excluding the undesirable member (see Branscombe et al., 1999). Furthermore, Castano et al. (2002) explain that well- established group members (high caste members) are especially aware of the pertaining rules and norms and therefore, any kind of deviance from such norms would pose a threat to one's group identity, which will be responded to by devaluating the perpetrating in-group member and seeing him/her as low in typicality. Research by Stamkou et al. (2016) further also add that especially those who highly identify with one's in-group perceive the in-group deviant as less typical of the in-group. Thus, in general, norm-deviating in-group members are seen as more negative than non-deviating members (Marques et al., 1998, 2001; Pinto et al., 2010). Also, according to the threat classification by Branscombe et al. (1999), it is seen that in the face of moral value threat, high identifiers (high caste individuals) are most likely to respond to such a threat by engaging in defensive reactions (black sheep effect) and wanting to be different from the deviant member of the group. We would thus argue that high caste individuals who are also high identifiers with their caste would devalue another in-group member committing norm transgression (that

is aggressive and intentional) and would find the transgression morally unacceptable in order to protect their threatened social identity.

Elaborating on the black sheep effect, according to subjective group dynamics theory (SGD; Marques et al., 1998; Abrams et al., 2000) group members are motivated to maintain a positive social identity. This motivation then results in positive evaluations of in-group conformers and negative evaluations of in-group deviants (Marques et al., 1998, 2001). In a similar vein, Stamkou et al. (2016) found that higher ranked group members showed more preference for norm followers than norm violators. They suggest that this could be because higher ranked members were more threatened by the norm violator's challenge to the status quo. Thus we can argue that high caste individuals would be more motivated to protect one's in-group identity by making negative evaluations of the deviant member. Likewise, according to relational models theory (RMT, Fiske, 1991, 1992), a derogation of in-group member in order to protect a group identity and integrity is explained by a transgression of moral norms regulated by specific in-group relations. In our context it particularly refers to moral motives for unity and hierarchy. Unity is aimed at caring for and supporting the integrity of in-group by avoiding or eliminating threats of contamination. When a group member commits a moral violation, the whole group feels contaminated and shamed until it purifies itself. Hierarchy in turn is aimed at maintaining linear orderings of social status where subordinates are motivated to respect and obey, and superiors to guide, protect, but also take moral responsibility for the actions of their subordinates (for review of all moral motives see Rai and Fiske, 2011). Thus, high caste individuals who break the strongly ingrained high caste norm of morality, purity, self-control, and pastoral care must expect group aversion or even a punishment. We were therefore interested in identity threat in the form of caste norm violation, and the ensuing cognitive representations of caste and status, which could be identity-maintaining. We assume in this context only the caste-based identity will be activated whilst other identities, such as religion, national and regional affiliation, will not play a role.

# Caste and Social Consequences

One of the most common social problems of the caste system was the discrimination of low caste members as explained earlier. In 1950, independent India's constitution banned caste-based discrimination and in order to compensate for historical injustices the authorities introduced quotas in government jobs and educational institutions to improve the quality of life of low castes ("What is India's caste system?", 2016, February 26., para. 13). A reservation system was introduced wherein a certain number of seats were reserved for members of the lower castes at places of higher education and government jobs. However, this legislation was soon met with a lot of resistance from the high caste community who felt that the system was not meritocratic, and provided an unjust advantage to the low caste members (e.g., Siddique, 2011). We believe that the reservation system is one of the most important social consequences of the caste system in modern times, and attitudes toward the system would have to be a reflection of one's caste identity.

# The Present Research

It was argued that social identity would play an integral role in the way people represent status. Threat to one's social (caste identity) in the form of norm violations would thus result in engaging in in-group protection. It is important to note that we refer to norm violations as behavior of fellow in-group members and not to one's own norm violations. Therefore, the aim of this study was to examine the role of caste-norm inconsistent (violating) vs. caste norm-consistent (non-violating) situations in making mental representations of status. Furthermore, we also wanted to examine how high caste and low caste individuals make moral judgments in situations of norm violation vs. non-violation. We predict that caste identity threat would be most salient for high caste individuals because these individuals in general have higher levels of caste identity. In general, we predicted that high caste individuals would show a black sheep effect when caste norm violations are introduced. We thus expected that in the face of identity threat, in a norm-inconsistent condition, there will be salient caste identity resulting in a mental representation that reflects the devaluation of an in-group perpetrator (black sheep effect), and lowered moral acceptability of an in-group perpetrator's action, particularly amongst high caste individuals. Moreover, we expect that this black sheep effect is mediated by high caste identity (and not mediated by regional, religious or national identity).

In this study, we used vignettes describing a social conflict situation which comprised norms of violence and morality, as reflecting common norms in the Indian context. The social conflict situation always entailed an intentional aggressive interaction between a victim and a perpetrator. Participants read two stories that depicted two protagonists as either a victim or perpetrator. The protagonists were either of high or low caste which was depicted using implicit stimuli of names and faces (pre-tested). They then had to choose and pick, as a response, one out of nine pictorial depictions that represented the protagonists in the story on a vertical line with varying degrees of status distance. At the end of each story participants also had to rate the moral acceptability of the behavior that took place between the protagonist and the victim. This procedure was chosen based on the general idea that status judgments are supported by an automatic simulation of vertical location (von Hecker et al., 2013; see also Schubert, 2005 for the same argument on social power). We assume that abstract concepts such as status can be mentally represented in an embodied way (Niedenthal et al., 2005). Giessner and Schubert (2007) showed that information about a leader's power influenced participants' vertical positioning of the leader. The paradigm used here involves such status representations via the vertical dimension. In this study, we interpret black sheep effect in the form of status evaluations wherein the in-group member is not necessarily excluded but devalued in status instead. That is, when the perpetrator is portrayed as someone from their in-group (high caste), they would then derogate that individual by representing him as having lower status. Low caste individuals

on the other hand would not differ in the way they represent status regardless of the condition, that is, they would generally show the trend of depicting a perpetrator as having greater status than a victim regardless of whether the perpetrator is of high or low caste<sup>3</sup> . Finally, we also predict that high caste identity will also mediate the relationship between caste affiliation and social outcomes of caste like endorsement of affirmative action. Thus, high caste individuals' opposing attitude toward educational affirmative action will be mediated by high caste identity.

# MATERIALS AND METHODS

## Participants

One hundred and two South Indians from various parts of Chennai, India, were recruited in person by choosing a convenience sampling technique. The participants were approached from various parts of the city and explained the study with examples and in the local language – Tamil. As a cover story, they were informed that the survey was part of a large research project that would help in understanding the Indian society better. The test materials were administered using paper and pencil. The researcher stayed in the area until the participants had completed the survey and they answered questions if necessary. Participants volunteered to participate for free in this research (age M = 33.25, SD = 12.46; 56 female; 45 high caste). Descriptive details of participant demographics are outlined in **Table 1**.

<sup>3</sup>We would like to argue that participants in fact have a pre-disposed idea of status based on information about caste. Thus, while the stories in the vignettes do imply a power relation, this pre-disposed notion of status has short term influence in making status representations. We argue that the transgression as an outcome of power relation is still evaluated based on which person is perpetrator –the high or low caste individual.


#### Design and Procedure

All participants completed the demographic questions and were then instructed that they would be reading a few stories and were required to answer a few questions based on them. Participants were then randomly assigned to one of the two conditions. (i) caste-norm consistent, wherein the perpetrator was of a low caste and the victim was of high caste, and (ii) caste-norm inconsistent, wherein the perpetrator was of high caste and the victim was of low caste. Participants then read the stories pertaining to each condition and answered questions on status representation and moral acceptability for each story. This was then followed by answering questions on attitudes toward affirmative action, about their national, regional, caste and religious identities and, finally, the manipulation check. At the end of the questionnaire all participants were debriefed.

# Materials and Measures Caste Stimuli Materials

Names and faces were pre-tested, from which four names and faces emerged; two belonging to low caste and two belonging to high caste from South India. The names contained the first name and a surname, indicative of a certain caste. The length of the name was not controlled since certain surnames are more indicative of the caste than others. The low caste names included Selvaraj Mani and Kannan Pandian. The high caste names included Murali Raman and Ravi Krishnamurthy. The characteristic features used to distinguish faces belonging to different caste included (i) Skin color, wherein a high caste face typically had lighter skin color and a low caste face had a darker skin color (ii) Facial hair, wherein a higher caste face would have less to no facial hair while a lower caste face would have more facial hair, and (iii) General grooming, wherein a high caste face looked more groomed than a low caste face. Participants were never explicitly informed that the people in stories were either of high or low caste.

#### Social Conflict Stories

Participants read two stories that described a social threat situation wherein two protagonists in a story were either a victim or a perpetrator. The particular conflict situations, involving issues of violence and morality, were chosen because of their commonness in the Indian society and because such issues are heavily rooted in distinguishing high and low caste characteristics. In the caste norm-consistent condition all participants read two stories in a South Indian context, wherein the victim was portrayed as a high caste member and the perpetrator was portrayed as a low caste member, based on the caste associations which names and faces would elicit. For example, "At the junction of Mount Road there was a huge traffic jam where two cars collided. While nobody got hurt, the drivers jumped out of their cars and started hurling verbal abuses at each other. That quarrel soon turned into a proper fight when Selvaraj Mani jumped out from his car and attacked Murali Raman. Murali Raman was pulled out of his car and was punched hard by Selvaraj Mani. Murali Raman finally fell to the ground and shouted apologies, but Selvaraj Mani did not stop kicking him until the police came and pulled the two drivers apart." In

the caste norm-inconsistent condition, participants also read two South Indian stories wherein the victim was portrayed as a low caste member and the perpetrator was portrayed as a high caste member. For example, "At the junction of Mount Road there was a huge traffic jam where two cars collided. While nobody got hurt, the drivers jumped out of their cars and started hurling verbal abuses at each other. That quarrel soon turned into a proper fight when Murali Raman jumped out from his car and attacked Selvaraj Mani. Selvaraj Mani was pulled out of his car and was punched hard by Murali Raman. Selvaraj Mani finally fell to the ground and shouted apologies, but Murali Raman did not stop kicking him until the police came and pulled the two drivers apart." At the end of each story, two photos were presented of the faces and their respective names below the photos. Thus, the low caste name was presented below the low caste photo and the high caste name was presented below the high caste photo. Two different conflict situations were chosen. One was about a traffic jam situation and the second was about a situation in the office where two colleagues get into a dispute (see Appendix 1 for all stories). The stories were counterbalanced within the South Indian context and also between conditions. All stories were in Tamil<sup>4</sup> .

#### Demographic Questions

Participants completed a survey that included several demographic questions indicating one's religion, perceived socio-economic status, education and caste. The caste categories included most backward caste (MBC), backward class (BC), ST, SC, and FC. Participants were asked to indicate their caste by choosing one of the categories. For the purpose of the research, MBC, BC, ST, and SC were grouped together as low caste and FC remained as high caste. **Table 1** shows all frequencies and percentages for other demographic variables.

#### Status Representation

After reading the stories participants had to choose one out of nine pictorial representations which, according to them, best depicted the social status between the two protagonists in the story. Each picture comprised a vertical line with two circles representing each of the protagonists with varying degrees of status. A wider distance between the two circles on the vertical dimension indicated a greater status difference (see **Figure 1** for details). The first and the ninth picture represented greatest status distance but in different directions, for instance, the victim on the lower part of the vertical line and the perpetrator on the upper part of the vertical line, or the positions of victim and perpetrator reversed. Picture 5 had two circles merged into one on a vertical line, indicating equal status. These meanings were explained to participants. Distance was scored based on the picture whereby the picture representing the greatest distance got a score of 4 and the picture representing least distance got a score of 1. The picture representing equal status got a score of 0. Positive distance values were scored for congruent representation of status and stories, that is, victim at lower position and perpetrator at higher position. Negative distance values were assigned for incongruent representation of status and stories, that is, victim at higher position and perpetrator at lower position. Each participant received a score for each of the stories. All chosen distances pertaining to the South Indian stories were combined together, to yield a distance score.

#### Moral Acceptability

After choosing status representations for all stories, participants were asked to indicate on a scale (1 = extremely unacceptable to 5 = extremely acceptable) to what extent they perceived the behavior between the protagonist and the victim in the scenarios as morally acceptable.

#### Attitudes toward Reservation System

Participants rated a question on whether they supported or opposed the affirmative action policy (Indian quota system) on a scale (1 = strongly oppose to 5 = strongly support).

#### Multiple Identities

Participants were asked to indicate to what extent they identified themselves with being Indian, south Indian, high caste, low caste, and being religious on a scale (1 = strongly do not identify to 5 = strongly identify). All items were answered in Tamil.

#### Manipulation Check

At the end of the study participants were presented with all eight names of the protagonists and were asked to indicate which caste they belonged to by choosing high caste, low caste, or an unsure option.

# RESULTS

Descriptive statistics of the manipulation check results are presented in **Table 2** which shows the response category frequencies for each of the protagonists. In general, the majority of participants categorized both South Indian high caste protagonists as belonging to the high caste and South Indian low caste protagonists as belonging to the low caste. Thus, we can assume that the names and faces of the protagonists were sufficient to identify the caste. Descriptive statistics and bivariate correlations of identity variables are presented in **Table 3**. For the purpose of the current study the status distances from the South Indian story 1 (M = 0.92, SD = 2.23) and South Indian story 2 (M = 0.61, SD = 2.49) were averaged and used since the sample was a South Indian sample. All participants read the stories and made individual status depiction for each story. To test our predictions, we checked whether caste-norm consistency (coded 0 as cast norm-inconsistent, 1 as caste norm-consistent) was associated with high-caste identity (mediator). In turn, high-caste identity was expected to predict status representation (DV) and moral acceptability (DV), but only amongst high caste members as compared to

<sup>4</sup>Apart from the two South Indian stories, all participants also had to respond to two other stories that were pertaining to the North Indian context. Since the participants were all from Chennai, Tamil Nadu in South India we did not include the distance and moral acceptability scores of the North Indian stories. Furthermore, the manipulation check reveled that when asked to classify the protagonists as belonging to a particular caste, most North Indian protagonists were classified in the 'unsure' category. Thus we only used the data pertaining to the South Indian context.

low caste members (moderator), as depicted in the theoretical model in **Figures 2**, **3**, respectively. That is, we predicted a caste norm inconsistent condition would lead to incongruent status representations (victim seen as of higher status and perpetrator seen as of lower status), as mediated by high-caste identity, but that this effect might be seen only amongst high caste individuals. Similarly, we predicted that a caste-norm inconsistent condition would lead to less moral acceptability of the high-caste perpetrator, as mediated by greater high caste identity, and again only amongst high-caste individuals. We used model 5 in Process macro for SPSS (Hayes, 2013) to test these moderated mediation effects. In each case 20,000 bootstrap samples were used, 95% bias-corrected confidence intervals and unstandardized coefficients are reported. Religion, education, gender, perceived socioeconomic class, national, regional, and religious identities were all controlled for. Furthermore, since we wanted to examine the role of one identity in particular, it was essential to control for these other identities.

In line with our predictions, we found that caste norm condition was related to status representation through the indirect effect of high-caste identity (IE = 0.24, 95% CI [0.06; 0.54]). The results showed that caste norm condition negatively predicted high-caste identity (b = −0.68, t = −2.81, p = 0.006, 95% CI [−1.17, −0.20]) and high-caste identity negatively predicted status distance (b = −0.35, t = −2.98, p = 0.004, 95% CI [−0.58, −0.12]). Thus, we see that caste norm inconsistency leads to higher caste identity which in turn results in incongruent status distances, however, only amongst high caste individuals.

TABLE 2 | Frequencies and percentage for caste categorized for each protagonist.


The conditional direct effect of caste norm condition on status representation was significant only among high-caste individuals (b = 4.84, t = 11.47, p < 0.0001, 95% CI [4.00, 5.68]) but not among low-caste individuals, (b = 0.39, t = 1.37, p = 0.175, 95% CI [−0.17, 0.95]). The interaction between condition and caste on status representation is graphically presented in **Figure 4**. The interaction between condition and caste is significant (b = 4.46, t = 8.52, p < 0.0001, 95% CI [3.42, 5.49]) wherein there is a significant difference between caste norm consistent and inconsistent condition amongst high-caste individuals (b = 0.37, t = 14.97, p < 0.0001, 95% CI [4.79, 6.26]) but no significant difference amongst low-caste individuals (b = 0.32, t = 1.10, p = 0.273, 95% CI [−0.26, 0.91]). In addition, results showed high caste individuals showed more status incongruent representations compared to low caste individuals in the norm inconsistent condition (b = −5.83, t = −17.87, p < 0.0001, 95% CI [−6.47, −5.17]). The analyses were repeated for each of the other identities (religious, national, regional, and low caste) that were used as mediators while controlling for the other. The indirect effects for lower caste identity (IE = −0.01, 95% CI [−0.09; 0.07]), South Indian identity (IE = −0.00, 95% CI [−0.07; 0.04]), national identity (IE = 0.00, 95% CI [−0.03; 0.08]) and religious identity (IE = −0.02, 95% CI [−0.16; 0.06]) were not significant.

We also found a replication of the moderated mediation model wherein caste norm condition was related to moral acceptability through the indirect effect of high-caste identity (IE = 0.11, 95% CI [0.01; 0.30]) while controlling for other identities. The results revealed that caste norm condition negatively predicted high-caste identity (b = −0.57, t = −2.53, p = 0.013, 95% CI [−1.02, −0.12]) and high caste identity negatively predicted moral acceptability (b = −0.19, t = −2.12, p = 0.039, 95% CI [−0.37, −0.01]). The conditional direct effect of caste norm condition on moral acceptability was again significant only among high-caste individuals (b = 1.84, t = 5.63, p < 0.0001, 95% CI [1.19, 2.48]) but not among low-caste individuals, (b = 0.13, t = 0.58, p = 0.561, 95% CI [−0.30, 0.56]). The interaction between condition and caste on status representation is graphically presented in **Figure 5**. The interaction between condition and caste is significant (b = 1.71, t = 4.17, p < 0.0001, 95% CI [0.90, 2.52]). Thus, we see that caste norm inconsistency leads to more high-caste identity which in turn results in lesser moral acceptability, however, only amongst high-caste individuals. The analyses were repeated for each of the other identities (religious, national, regional, and low caste) that were used as mediators while controlling for the other. The indirect effects for low-caste identity (IE = −0.04, 95% CI [−0.03;


0.11]), South Indian identity (IE = −0.01, 95% CI [−0.10; 0.04]), national identity (IE = 0.00, 95% CI [−0.03; 0.05]) and religious identity (IE = −0.00, 95% CI [−0.09; 0.01]) were again not significant.

To test the effects of actual caste on the social consequences of the caste system, such as attitudes toward affirmative action as mediated through upper caste identity, a mediation analysis was conducted (Process model 4, Hayes, 2013) as theoretically depicted in **Figure 6**. Again, in line with our predictions, the results showed that actual caste positively predicted high caste identity (b = 1.10, t = 4.78, p < 0.0001, 95% CI [0.65, 1.56]) and high caste identity negatively predicted the affirmative action attitudes (b = −0.39, t = −8.76, p < 0.0001, 95% CI [−0.53, −0.25]). We found an indirect effect of caste on affirmative action attitude partially mediated by high caste identity (IE = −0.43, 95% CI [0.07; 0.46]). The direct effect of actual caste on affirmative action attitudes was also significant (b = −1.56, SE = 0.18, p < 0.0001, 95% CI [−1.91, −1.21]). Thus, high caste leads to opposition of affirmative action partially due to heightened upper caste identity. The analyses were repeated for each of the other identities while controlling for the other. The indirect effects for lower caste identity (IE = 0.03, 95% CI [−0.20; 0.23]), South Indian identity (IE = −0.03, 95% CI [−0.06; 0.16]), national identity (IE = 0.02, 95% CI [−0.08; 0.15]) and religious identity (IE = −0.00, 95% CI [−0.08; 0.09]) were also not significant.

# DISCUSSION

In this research, we apply a well-established theoretical framework incorporating social identity theory (Tajfel and Turner, 1986), the black sheep effect (Marques et al., 1988) and moral motives derived from relational models theory (Fiske, 1991; Rai and Fiske, 2011) to understand a small part of a complex social structure, namely the Indian caste system. It is important to state that we are not claiming that this study answered all the nuances seen in exploring the caste system but has paved the way toward a better understanding of this concept. Overall, we found support for our argument that group membership to a

particular caste has consequences in the way one perceives status amongst fellow members in the society. More specifically, we validated that when high-caste individuals learn that their fellow in-group member engaged in norm-violating behavior then they devalue the status of that in-group member. Thus, for instance, when a high-caste member marries a low caste member, this high caste member will be devalued by his fellow in-groupers owing to norm-violation. Moreover, we find that this trend is particularly observed as a consequence of heightened high-caste identity. Our results show that in a situation of norm violation, high caste identity predicts the way people make status representations. We argue that this norm violation induces identity threat, especially among high-caste individuals and as a result of the threat they heighten their high-caste identity, thereby engaging in status devaluation. The stories in the study also depict an intentional aggressive, deviant behavior and in line with previous research (Abrams et al., 2000; Marques et al., 2001) this behavior led to black sheep effects in the form of status devaluation by high-caste members. Classic studies by Branscombe et al. (1993) and Biernat et al. (1999) show that highly identified individuals show more in-group derogation especially when one's identity is threatened. Moreover, we also claim that since high-caste members tend to see caste identity as inherited at birth and also tend to essentialise their caste (Dube, 2001; Mahalingam, 2007) a high-caste identity will be more salient amongst the high-caste individuals than low caste-identity amongst the low-caste individuals. Thus, high caste members would especially defend the identity threat to maintain positive social identity due to heightened identification with the group (Marques and Paez, 1994) and would perceive the in-group deviant as less typical of the in-group (Castano et al., 2002).

Thus, in the present study we were able to show how status is represented without explicit information about the protagonist's caste. The information people received was merely the person's name and a photograph and yet it is evident that existing knowledge about caste was activated. As seen in the manipulation check, in general, most high-caste protagonist names were identified as belonging to the high caste and low-caste protagonist names as belonging to the low caste. Moreover, the study

incorporated a novel method to measure perception of status by means of a projective task using the vertical dimension, as this dimension has been shown to be used in reasoning to simulate power (Schubert, 2005) and social status (von Hecker et al., 2013), in order to generate judgments on these characteristics. Therefore, when participants were depicting relations in such a way that a perpetrator was represented on the bottom of the vertical line and the victim on top, we believe the abstract concept of status and in-group derogation was clearly embodied.

It is important to understand that the reason a norm-violating social situation was chosen to induce identity threat was because in the Indian context caste norm violation is perhaps the most salient identity threat situation. Deshpande (2010) explains that caste has acquired hereditary characteristics in the Indian society which comes with certain norms that evolved from ancient scriptures. Thus, when these norms are violated in the society, it is received with punitive costs. The threat used in this context violated the basic morality values of the caste and according to Branscombe et al. (2000), threat to moral values leads to defensive reactions and perceived in-group heterogeneity, especially amongst high identifiers. Thus, when the high caste moral code of being virtuous and non-aggressive was violated, highly identified high caste members strengthened their identity and showed in-group derogation by devaluing the status of the in-group deviant. This reasoning was further confirmed when high-caste individuals in the norm-inconsistent (violating) condition showed lower moral acceptability of the behavior when the perpetrator was a high-caste member.

Secondly, the results also show that only high-caste individuals seem to be affected by the norm violation identity threat in making status representations but not the low-caste individuals. In fact, the low-caste individuals seem to show no differences in the way they represent status and moreover they seem to do so in a status-congruent direction, that is, the perpetrator is always depicted as having high status and a victim is always depicted as having low status. This could possibly be the case because historically low-caste individuals are the stigmatized low status groups (Mahalingam, 2003). Therefore, they probably do not experience the same amount of identity threat in a normviolation condition as the high-caste individuals. The low caste individuals, due to the belief that their caste was not acquired through birth (Dube, 2001) and the lack of an established status, probably also showed lesser identification with their lowcaste. Thus, when a low- caste individual reads a story about a norm-violating low-caste member who is in fact a victim, there is no threat that is activated probably because they do not show high identification. Furthermore, being portrayed as a victim does not threaten the image of the in-group of low caste individuals, thereby lowering the need to protect the in-group image.

Another important outcome of this study was that only highcaste identity seemed to play an integral role in predicting status effects and moral acceptability. As already explained by Judge and Bal (2008), caste identity dominates over all other identities in the Indian context, probably because of the inherent characteristics the concept of caste entails. As seen in the correlations (**Table 2**), only high caste and low caste identities seem to negatively correlate with each other; understandably because those with greater high caste identities would have lesser low caste identities. Interestingly, low-caste identity did not act as a significant mediator in predicting status relationships or understanding moral acceptability. Understandably because previous studies have shown that caste identity is essentialised and legitimized (Dube, 2001; Mahalingam, 2007) particularly by high caste individuals; thus possessing high-caste identity could result in understanding and representing status differences more than having low-caste identity. Our results also show that none of the other identities – national, regional or religious played an important role in predicting status distances or moral acceptability. While India is a complex social structure with an interactive role of multiple identities, we have reasonable evidence to believe that perhaps in the context of status representation, hierarchies and caste-related social outcomes, other identities are not as salient, so perhaps people discard these identities when being involved in status-relevant situations. Thus, it seems like in the context of how individuals see status in the society, only high-caste identity seems to be activated while alternative identities might be discarded. This is a very important finding because whilst studies have shown that caste identity is the most salient identity (Gayer, 2000), perhaps heightening other identities like regional or national identities could result in reducing caste-based status hierarchies or status representations. Interventions such as this could be seen in the work by Brewer and Gaertner (2008) where they speak of cross-categorization as a technique to overcome prejudice. A typical example could be seen in the sporting context when athletes all over the country from different socioeconomic backgrounds and castes come together as a team to play for the country. The heightened in-group national identity perhaps diminishes existing caste based identities that could potentially create situations of prejudice amongst various players. While there are many religious and regional conflicts we see that caste based discrimination is the most salient form when it involves differentiating groups of varied status.

Finally, the results also confirm the social consequences of a caste-based system in terms of affirmative action. As pointed out by Siddique (2011), affirmative action policies were introduced to improve the quality of life of low-caste individuals. However, it was soon met with opposition from the high-caste community. Sometimes caste is not always related to class, so it is possible that someone of low caste could be of high class and therefore have sufficient means and opportunities at places of education and government jobs. Thus, the present results confirmed that high-caste individuals would generally oppose affirmative action, whereby this link was partially mediated through high-caste identity. More importantly, again, none of the alternative identities played a role in determining attitudes toward affirmative action. This is unsurprising because affirmative action was introduced as an outcome of revisiting the consequence of the caste system for low-caste individuals. Therefore, other identities do not seem to play a vital role in predicting these effects.

The present study, however, has some limitations especially with respect to sample size and the sampling technique (convenience sampling). While only single-item measures were used for measuring identity, as a follow-up study caste-based identity should also be measured on its own using a questionnaire and relate this to the present findings. The present findings should also be extended to other norm-violating instances apart from the victim-perpetrator context. In India, several other norm violations exist, for instance, consuming beef in a purely 'Hindu' state, or women dressed in 'western' clothes, or engaging in intercaste marriage, the latter being the most predominant caste-norm transgression. Furthermore, it would be essential to re-run the study, this time manipulating the class of the protagonists using different professions (without names), in order to rule out the influence of class in these processes. It would also be essential to replicate these findings with a North Indian sample. Finally, it is important to note that when a high caste individual is engaged in a norm-violation, there is a greater negative connotation as a high caste individual's consistent norm is that of purity and morality (Deshpande, 2010). The low caste individuals' consistent norm, however, is that of impurity and immorality, therefore, a negative expectation of their behavior is already established as a norm and therefore in a norm-violation condition, wherein they are portrayed as a victim, they are probably not as threatened as highcaste individuals. A positive connotation attached to a normviolation perhaps does not induce as much threat and therefore low-caste individuals remain unaffected by the two conditions. This limitation is something that needs to be addressed in future studies.

# CONCLUSION

The present study tried to utilize basic social identity theories in explaining one outcome of a very complex social system, such as the Indian caste system. It also touched upon the most dominant identity that is useful in predicting status differences in the Indian society. Considering the complexity of the Indian caste system, it was crucial to look into a small consequence of the caste system rather than the whole structure on its own. This study also paves the way into the understanding of the black sheep effect as seen in status representation as supposed to general in-group derogation

# REFERENCES


attitudes. It is fundamental to expand this study by examining other aspects of social identity threat in the Indian context, such as religion, to gain some more understanding of how multiple identities interact with each other in the functioning of the Indian society.

# ETHICS STATEMENT

The study was approved by the Scientific Research and Ethics Committee at the Institute of Psychology, Jagiellonian University, Krakow, Poland (the date of the record July 2, 2014). All participants signed the informed consent form and were told that they were free to withdraw from the study at any time.

# AUTHOR CONTRIBUTIONS

SS and UvH initially conceptualized the idea. SS and MS developed the idea and primarily designed the study. SS was mainly involved in data collection and data analyses. SS and MS were involved in data interpretation and UvH contributed to the discussion. SS was involved in writing the paper draft of the paper. Both MS and UvH were involved in the revised versions of paper.

# ACKNOWLEDGMENTS

We would like to thank Usha Sankaran from Chennai, India who volunteered to collect data for this study and helped coding the questionnaire responses on excel. This research is carried out as part of the FUGA grant (2016/20/S/HS6/00317) awarded to SS from the National Centre for Science, Poland (2016–2019). The publication costs are supported by Cardiff University.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpsyg. 2017.00487/full#supplementary-material



"What is India's caste system?," (2016). Available at: http://www.bbc.com/news/ world-asia-india-35650616

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Sankaran, Sekerdej and von Hecker. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# "I Want, Therefore I Am" – Anticipated Upward Mobility Reduces Ingroup Concern

Marion Chipeaux<sup>1</sup> \*, Clara Kulich<sup>1</sup> \*, Vincenzo Iacoviello<sup>2</sup> and Fabio Lorenzi-Cioldi<sup>1</sup>

<sup>1</sup> Social Psychology, Section of Psychology, Faculty of Psychology and Educational Sciences, University of Geneva, Geneva, Switzerland, <sup>2</sup> Social Psychology, Psychology, Faculty of Behavioural and Social Sciences, University of Groningen, Groningen, Netherlands

#### Edited by:

Marcel Zentner, University of Innsbruck, Austria

#### Reviewed by:

Andrew Edgar, Cardiff University, United Kingdom Mark Hallahan, College of the Holy Cross, United States

#### \*Correspondence:

Marion Chipeaux marion.chipeaux@unige.ch Clara Kulich clara.kulich@unige.ch

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

Received: 11 November 2016 Accepted: 10 August 2017 Published: 28 August 2017

#### Citation:

Chipeaux M, Kulich C, Iacoviello V and Lorenzi-Cioldi F (2017) "I Want, Therefore I Am" – Anticipated Upward Mobility Reduces Ingroup Concern. Front. Psychol. 8:1451. doi: 10.3389/fpsyg.2017.01451 Empirical findings suggest that members of socially disadvantaged groups who join a better-valued group through individual achievement tend to express low concern for their disadvantaged ingroup (e.g., denial of collective discrimination, low intent to initiate collective action). In the present research, we investigated whether this tendency occurs solely for individuals who have already engaged in social mobility, or also for individuals who psychologically prepare themselves, that is 'anticipate', social mobility. Moreover, we examined the role of group identification in this process. In two studies, we looked at the case of 'frontier workers', that is people who cross a national border every day to work in another country where the salaries are higher thereby achieving a better socioeconomic status than in their home-country. Study 1 (N = 176) examined attitudes of French nationals (both the socially mobile and the non-mobile) and of Swiss nationals toward the non-mobile group. As expected, results showed that the mobile French had more negative attitudes than their non-mobile counterparts, but less negative attitudes than the Swiss. In Study 2 (N = 216), we examined ingroup concern at different stages of the social mobility process by comparing the attitudes of French people who worked in Switzerland (mobile individuals), with those who envisioned (anticipators), or not (non-anticipators), to work in Switzerland. The findings revealed that anticipators' motivation to get personally involved in collective action for their French ingroup was lower than the non-anticipators', but higher than the mobile individuals'. Moreover, we found that the decrease in ingroup concern across the different stages of social mobility was accounted for by a lower identification with the inherited ingroup. These findings corroborate the deleterious impact of social mobility on attitudes toward a lowstatus ingroup, and show that the decrease in ingroup concern already occurs among individuals who anticipate moving up the hierarchy. The discussion focuses on the role of the discounting of inherited identities in both the anticipation and the achievement of a higher-status identity.

Keywords: multiple social identities, social mobility, status inconsistency, ingroup concern, identification

# INTRODUCTION

fpsyg-08-01451 August 24, 2017 Time: 14:48 # 2

Individuals are members of inherited social groups defined, for instance, by their gender, ethnicity, or age. They belong simultaneously to more malleable categories qualified by their educational and professional achievements. The present work is interested in how individuals cope with such multiple group memberships, in particular when these memberships are associated with different value and prestige (i.e., social status). Nowadays, most societies are still organized around a hierarchical principle of distribution of resources and power, creating and reinforcing economic, cultural, and political inequalities. Some groups are associated with a high social status whereas others with a low one (Tajfel and Turner, 1986; Sidanius and Pratto, 1999). Nevertheless, while in traditional societies the various group memberships tended to be aligned in status (Lenski, 1954), the stronger social fluidity of contemporary societies leads individuals to belong to multiple groups of conflicting status. For example, individuals from disadvantaged inherited backgrounds (e.g., women, ethnic minorities) may achieve higher status through professional attainments. The present research seeks to better understand the socio-psychological processes at play when individuals are confronted with such status inconsistency due to upward mobility. We first consider how they cope with the contradicting demands arising from such multiple group memberships, by investigating their concern for the inherited low-status group members who did not achieve social mobility (Study 1) and more generally toward the inherited low-status ingroup (Study 2). We then investigate whether the anticipation of social mobility already leads to a decreased ingroup concern.

Although being hierarchically organized, modern societies are characterized by an ideal of meritocracy that leads people to believe that personal investment and efforts are main causes of success (McCoy and Major, 2007). People are encouraged to focus on their personal trajectory and to engage in individual strategies in order to improve their social standing and to achieve self-worth (Tajfel and Turner, 1986; Wright, 2001). The social mobility strategy, as defined by social identity theory (SIT: Tajfel and Turner, 1986), describes individuals who suffer from the low status associated with their group membership and decide to quit their group to join a better valued one. However, although clear-cut scenarios can be designed in the laboratory in order to make salient one specific membership, the study of identities in real life is more complex. Most often, individuals are confronted with contexts in which several of their group memberships are salient. Moreover, we argue that the possibility to leave a social group for another drastically varies depending on the nature of the group memberships. Whilst some group memberships are achieved by individuals throughout their lives (e.g., professional occupation, political affiliation), other groups are imposed from birth and are thus inherited in quality (e.g., gender, ethnicity). For individuals who are members of low-status achieved groups, the social mobility process can effectively occur as they move from one group to another. An illustration is an individual's attempt to quit their employee status by moving up the social ladder and becoming a manager. However, when the low-status membership is inherited, this status is quite impermeable, meaning that the individual has little power to modify it. For instance, a woman cannot easily change her sex, but she can focus on her professional standing and become a manager (Ellemers, 2001; Derks et al., 2016, for a review).

Past research has shown that there is a tendency toward status crystallization, meaning that the probability to achieve a high-status membership is greater for members of highstatus inherited groups than for members of low-status inherited groups (Lenski, 1954; Bourdieu, 1984). Nonetheless, societies have become increasingly fluid over the past decades, notably because of social and political movements (e.g., feminism, human rights movements) which have contributed to break societal barriers. A product of these more fluid societies is the increasing number of individuals experiencing statusinconsistent identity configurations. According to Lenski (1966), individuals who are simultaneously members of low- and highstatus groups experience a psychological tension derived from their motivation to improve their social identity while still belonging to a low-status group. In line with this idea, Wright and Taylor (1999) showed that low-status group members who succeeded as a token felt more negative emotions than individuals who succeeded in a non-discriminatory context. This means that, when being simultaneously members of low- and highstatus groups, individuals face contradicting social expectations. Indeed, such expectations (e.g., stereotypes) differ to a great extent according to group status (Fiske et al., 2002; Kervyn et al., 2009). Moreover, while high-status groups promote norms and values related to independence, individualism, and selffulfillment, low-status groups convey norms and values that promote interdependence and solidarity among its members (Lorenzi-Cioldi, 1988, 2009). Thus, by conforming to the norms of one of their memberships, individuals in statusinconsistent identity configurations deviate from the norms of their other membership, and expose themselves to various forms of social punishments. As an illustration, female managers may be punished for enacting agentic behaviors, because these behaviors contradict the female stereotype despite the fact that agency is expected for the professional role (Rudman and Glick, 2001).

Consistent with the motive to achieve self-worth as posited in SIT, individuals who possess multiple social identities have "a natural tendency to think of themselves in terms of that status or rank which is highest, and to expect others to do the same" (Lenski, 1966, 87). Providing evidence of Lenski's reasoning, a series of studies conducted by Derks and colleagues showed that, among women who achieved a high-responsibility professional role, those who reported low levels of gender identification and who reported having experienced gender discrimination tended to describe themselves as more similar to the highstatus group, compared to women high in gender identification and/or women having experienced low gender discrimination. In this way, they portrayed themselves using more masculine traits (the characteristics of the high-status group), while using the same amount of feminine traits as other women (Derks et al., 2011a,b). In addition, research showed that female faculty

rated their male Ph.D. students more favorably than their female Ph.D. students, whilst no difference was observed among the male faculties' evaluations (Ellemers et al., 2004). Of interest, the same pattern of results was observed among Hindustanis in the Netherlands who self-described as more Dutch when lowly identified with their ingroup and when having experienced discrimination (Derks et al., 2015). Lack of ingroup support has also been reported from the perspective of achieved lowstatus individuals. Research showed that female employees and Non-White employees felt less support from ingroup supervisors (i.e., female and Non-White supervisors) than from outgroup supervisors (i.e., male or White supervisors) in organizations with an adverse diversity climate (Paustian-Underdahl et al., 2017). Such parallel findings across social categories suggest that low ingroup concern among socially mobile women is not specific to gender, and that it can be broadly attributed to the status dynamics between different group memberships. Providing evidence to this reasoning, Kulich et al. (2015) compared the ingroup concern of low-status inherited group members who had successfully engaged in social mobility to their congeners who had not. In their research, the authors observed that mobile members of different social categories of inherited low-status groups (e.g., Afro-Americans, immigrants, and women) expressed greater hostility and lesser support for the inherited low-status group compared to non-mobile members. Taken together, these findings suggest that individuals who experience a status-inconsistent identity configuration describe themselves as more similar to the achieved highstatus group and are less supportive of the low-status inherited group.

An important issue is the kind of motivation that leads to such ingroup unsympathetic attitudes. Predictions derived from SIT (see Ellemers, 2001) would explain the lack of ingroup concern by a decrease in ingroup identification. However, Kulich et al.'s (2015) findings revealed that identification with the inherited group did not play a role, as mobile and non-mobile participants were similarly identified with their inherited group. This suggests that, despite their self-distancing from the inherited group on the attitudinal dimension and their stronger counter-stereotypical self-descriptions, the mobile remained identified with this group, and also similarly selfdescribed on the dimension that is stereotypical of it. In addition, the authors found that the lack of support for the inherited ingroup was accounted for by an increased identification with the achieved high-status group. Mobile individuals identified more strongly with their achieved group than the non-mobile. Such pattern suggests that individuals with multiple identities do not necessarily disengage from their low-status group. They may have to cope with the simultaneous presence of several identities and thus the coping strategies on the attitudinal, self-evaluation, and self-categorization levels are not aligned. For example, the disparagement of the inherited low-status ingroup may be motivated by an effort to become accepted in the new high-status group (Wright and Taylor, 1999), while keeping the ties with the low-status inherited group. Indeed, the conflicting nature of their identity configuration becomes evident in the contrasting phenomena of identification and simultaneous negative attitudes toward the low-status ingroup. Research is thus needed to identify the specifics of this assimilation process which does not appear to influence to the same extent the different dimensions of analysis (e.g., attitudes, identification).

One way to look at this assimilation process is to compare the attitudes of mobile individuals to the attitudes of the high-status group members. Indeed, considering the motivation of low-status groups' members to enhance the positivity of their social identity, one may expect that mobile individuals adopt similar attitudes as those of high-status groups' members, that is, members of groups that mobile individuals joined through their social mobility. This could be considered as a strategy to increase their chances to be accepted in the new group by showing to its members that they do not consider themselves as members of the low-status (out)group anymore (Merton, 1968). Indeed, Van Laar et al. (2014) showed that members of the high-status group offered support to the mobile only if they perceived that the latter were not behaving in a manner that was prototypical of their low-status group. Nevertheless, we also know from the social identity perspective that social groups' members need to feel their membership not only as positive, but also as distinct (Tajfel and Turner, 1986). Thus, to achieve this feeling of distinctiveness, individuals tend to express ingroup bias. This bias describes the tendency of individuals to show more negative attitudes toward outgroup members, compared to ingroup members, especially when social categorization is salient (Mullen et al., 1992). This may help to ensure the distinctiveness of their group membership, but also indirectly to increase their self-esteem (see for example, Brown, 2000). In sum, it appears more reasonable to think that in order to maintain the distinctiveness of their membership, members of the high-status group will still express more negative attitudes toward the low-status group members, than the mobile individuals, thus preventing a complete assimilation of mobile individuals which would threaten their distinctiveness.

A second way of looking at this assimilation process is to wonder whether social mobility in itself is related to a lack of ingroup concern, or whether this relationship is already present among individuals who merely aspire to undertake social mobility. Indeed, Merton (1968) suggests that social mobility is often preceded, or facilitated, by the expression of positive attitudes toward the group to which the individual seeks to belong to. He theorized this process as an anticipatory socialization which contributes to increase the probability of successful individual mobility, as well as the integration in the new group once joined. In line with this idea, research by Ellemers et al. (1993) showed that, when people find themselves in a permeable intergroup context, they tend to act in order to defend their individual interests rather than the interests of their group (see also Wright et al., 1990). Ellemers et al. (1990) further showed that, in a permeable context where individual mobility is facilitated, competent individuals strongly identify with the high-status group. Thus, on the basis of this literature, we expect different levels of ingroup concern between the non-mobile who do not strive for social mobility and the non-mobile who do. In conformity with Merton's (1968) suggestion, anticipators of social mobility should reduce their ingroup concern in order to enhance their chances to achieve the mobility. Finally, even if anticipators should reveal a lower concern than the nonanticipators, we expected them to still be more concerned than mobile individuals. The latter, who successfully achieved social mobility, should be motivated to maintain their distinctiveness by distancing even more strongly from their low-status inherited ingroup.

# The Present Research

fpsyg-08-01451 August 24, 2017 Time: 14:48 # 4

In order to examine the identity management strategies in dual identity configurations, we conducted two correlational studies. Our target group consisted of French nationals living in areas around the Swiss border. As the costs of living are higher in this region than the average costs in France, many French from this region attempt to join the Swiss work-force which grants a number of financial and symbolic advantages: The unemployment rate in Switzerland is almost half (5.1%, FSO, 2014) of the rate in France (9.9%, INSEE, 2014), and the median salary in Switzerland (5,560 euros) is three times higher than the median salary in France (1,712 euros). We took advantage of this natural setting and compared these socially mobile French 'frontier workers', who achieved a considerably higher socioeconomic standing, with the non-mobile French who worked in France.

Our first aim was to test whether mobile individuals (i.e., French frontier workers) are less concerned with the achieved low-status group that they have left (i.e., French workers in France; see Study 1) and with their inherited ingroup as a whole (i.e., French people who live in border regions of Switzerland; see Study 2), compared to non-mobile individuals (i.e., French workers in France). Moreover, we looked at the relevant inherited high-status outgroup (i.e., Swiss workers in Switzerland), who are granted consistency between their inherited and achieved memberships (see Study 1). This design bore the opportunity to assess the extent to which mobile individuals assimilate to the high-status group. Finally, we examined whether the actual achievement of social mobility is a necessary condition to undermine ingroup concern, or whether the mere prospect of undertaking social mobility is sufficient to do so. This was done by measuring non-mobile participants' willingness to engage in social mobility (Study 2). From these general goals, we derived the following two hypotheses:

Hypothesis 1 (tested in Study 1) predicts a linear effect of social mobility on the concern for the low-status achieved group. More specifically, we expect non-mobile individuals to express more concern than the mobile individuals, who in turn should express more concern than the high-status inherited group members. We thus aimed to demonstrate that mobile individuals, who achieved a high-status position through individual mobility, express a lower concern for the fate of their inherited group members who did not succeed individually, and that they have the willingness to assimilate with the high-status group. Moreover, we sought to highlight that the high-status group's members can feel threatened by mobile individuals and should therefore express an even lower concern. This would safeguard their ingroup distinctiveness.

Hypothesis 2 (tested in Study 2) predicts a linear effect of the stages of social mobility (non-anticipators, anticipators, mobile individuals) on the concern for the inherited low-status group. More specifically, we expect the concern for the low-status inherited ingroup to be highest among non-mobile individuals who do not wish to undertake mobility (i.e., non-anticipators), moderate among non-mobile individuals who strive for mobility (i.e., anticipators), and lowest among individuals who have succeeded in their mobility (i.e., mobile) – the latter being motivated to claim their distinctiveness in the face of the anticipators. Indeed, even if anticipators have a strong desire to improve their social status, they are still part of the non-mobile group (i.e., by being simultaneously members of low-status achieved and inherited groups), and for this reason, they should still express a higher concern for their inherited membership compared to mobile individuals who can focus on their high-status achieved membership in order to reduce the identity threat associated with their low-status inherited membership.

In addition, we aimed to investigate the mechanisms underlying these attitudinal differences. Although the social identity perspective leads to the expectation that self-ingroup distancing derives from a lower identification with the inherited group, this has not been found in past research (Kulich et al., 2015). We believe that different levels of inherited ingroup identification could have been concealed in previous studies because such studies only looked at non-mobile individuals without making the distinction between anticipators and nonanticipators of social mobility. To move a step forward, our research considered social identification with the inherited group among these two non-mobile subgroups. From this, follows our next hypothesis:

Consistent with assumptions derived from SIT, Hypothesis 3 (tested in Study 2) predicts that the lower ingroup concern among anticipators of social mobility (as compared to non-anticipators) and even lower concern among mobile individuals should be explained by a lower identification with the low-status inherited group.

# Study 1

Study 1 tested H1, which predicts a linear effect of social mobility on concern for the low-status achieved group. More specifically, we examined this concern among French workers in France (i.e., the non-mobile individuals), French workers in Switzerland (i.e., the mobile individuals), and Swiss workers in Switzerland (i.e., the members of the high-status group).

In addition, we also explored the identification patterns associated with these different categories.

#### Method

#### **Participants**

A total of 176 participants (122 women and 54 men, Mage = 34.53, SDage = 9.23, ranging from 19 to 62 years old) were recruited through social networks and were asked to complete an online questionnaire. One-hundred and fifteen participants were French and 61 were Swiss.

#### **Materials and measures**

fpsyg-08-01451 August 24, 2017 Time: 14:48 # 5

Participants indicated their citizenship and were then presented with a short excerpt in which France and Switzerland were compared on several domains, such as employment rates and average wages. The aim of this introductive part was to emphasize the current socio-economic status gap of the two national groups. Participants were then asked to answer several items, which are listed below in the chronology of their occurrence.

Social mobility. The social mobility variable distinguished between three groups of participants. French working in France were the non-mobile group (n = 43). These participants are characterized by their relative low inherited status (i.e., French as compared to Swiss nationals) along with a low achieved status (i.e., they work in France). French nationals working in Switzerland were the mobile group (n = 72). These participants are characterized by a low inherited status combined with a high achieved status (i.e., French working in Switzerland). Finally, Swiss nationals working in Switzerland (n = 61) were the high-status group, characterized by high inherited and achieved statuses.

Identification with the inherited and the achieved groups. Identification, with the inherited and the achieved groups, was assessed with the 10 items of the self-investment dimension of the hierarchical model of ingroup identification (Leach et al., 2008). Sample items are "I feel a bond with [Ingroup]", "I'm glad to be [Ingroup]", or "I often think about the fact that I am [Ingroup]" (1 fully disagree to 7 fully agree). First, participants were asked to answer these items for their inherited group (i.e., "French people in general" or "Swiss people in general"). The reliability of this scale was satisfactory for both targets (respectively, α = 0.90; M = 4.77, SD = 1.36 for the French, n = 115, and α = 0.89; M = 5.42, SD = 1.03 for the Swiss, n = 61). Second, they were asked to answer these items for their achieved group (i.e., "workers in France" or "workers in Switzerland"). The reliability of this scale was also satisfactory for both targets (respectively, α = 0.84; M = 4.43, SD = 1.16 for the workers in France, n = 43, and α = 0.84; M = 5.29, SD = 0.96 for the workers in Switzerland, n = 133).

Concern for the low-status achieved group. We measured participants' motivation to engage in social action aimed at improving the situation of French nationals who lived in border regions of Switzerland and worked in France. We measured support for both personal involvement and group involvement in social action because this allowed to capture potential psychological distancing from the group as a result of a simultaneous expression of high support for group involvement and low motivation to get personally engaged. These measures were taken on a 7-point scale, from 1 not at all to 7 totally.

Support for group involvement was assessed with two items (e.g., "French people who work in France and live in border areas should fight collectively for financial compensation for the difference they face between the cost of living and the level of their wages", r = 0.64, p < 0.001; M = 4.67, SD = 1.79). Personal involvement was assessed with two items measuring participants' motivation to get personally involved in social action (e.g., "I would be willing to sign a petition to call for more economic support for French people who work in France and live in border regions", r = 0.55, p < 0.001; M = 3.78, SD = 1.87).

We also introduced a direct measure of concern for the lowstatus achieved group with the single item: "I feel concerned by the fate of French people living in border areas of Switzerland and working in France" (1 not at all to 7 totally; M = 4.19, SD = 1.95).

Socio-demographic information. Finally, participants indicated their gender, their professional status (with in general 67.6% employees, 8.5% entry-level managers, 9.1% middle managers, 8% senior managers and 6.9% missing data), and their age (M = 34.53, SD = 9.23; ranging from 19 to 62 years old). We also measured the subjective social status of their occupation with two items (i.e., "To what extent do you think that your professional occupation is valued – and – prestigious in society?") (1 not at all to 7 totally, r = 0.46, p < 0.001, M = 3.98, SD = 1.34). Moreover, we measured the perceived status of working in Switzerland and France with two items ("To what extent do you think that working in France/Switzerland is valorizing?", 7 point-scale from 1 not at all to 7 totally, with, respectively, M = 3.34, SD = 1.62 for the French, and M = 5.29, SD = 1.22 for the Swiss).

## Results

#### **Preliminary analyses**

We performed several analyses in order to detect potential differences on relevant socio-demographic indicators between the three groups of participants. First, we looked at gender, as the occupational gender divide may lead to men and women occupying professions that differ in type and status (e.g., Charles and Grusky, 2004). Chi-square analysis showed that men and women were similarly distributed across the three mobility groups, χ 2 (2, N = 176) = 3.34, p = 0.18. Moreover, the three groups did not differ in professional status, χ 2 (6, N = 164) = 8.80, p = 0.18. ANOVA of the continuous variable measuring the subjective social status of the occupation, with social mobility as a between-participants factor, revealed no effect, F(2,172) = 0.69, p = 0.50, η 2 <sup>p</sup> = 0.16. We also tested if the status that the three investigated groups attributed to working in France and in Switzerland differed. In a repeated-measures ANOVA with the two status items and the three participant groups as a between-participants factor, we observed that all groups of participants rated working in Switzerland as more valued than working in France, F(1,171) = 144.49, p < 0.001, η 2 <sup>p</sup> = 0.46. Furthermore, this effect was qualified by a significant interaction effect between the two factors, F(2,171) = 4.11, p = 0.018, η 2 <sup>p</sup> = 0.05. Although the French workers also believed that working in Switzerland was more valued than working in France (p < 0.001), pairwise comparisons revealed that the French working in France attributed a higher value to working in France than the other two groups (ps < 0.04). Finally, we tested for age differences in the three mobility groups and observed a marginal effect, F(2,173) = 2.53, p = 0.08, η 2 <sup>p</sup> = 0.03. Pairwise comparisons showed that French workers in France (M = 31.81, SD = 6.94) were younger

than French (M = 35.28, SD = 9.04, p = 0.051) and Swiss (M = 35.57, SD = 10.54, p = 0.04) workers in Switzerland. No difference was observed between the two groups working in Switzerland (p > 0.85). In light of this unexpected finding, participant age was entered as a covariate in all of the following analyses<sup>1</sup> .

#### **Hypotheses testing**

Concern for the low-status achieved group. In order to test H1 which predicts a linear effect of social mobility on concern for the low-status group, we computed two orthogonal contrasts with the social mobility variable. The first contrast (C1) opposed the French workers in France (i.e., non-mobile), coded −1, to the Swiss (i.e., high-status group members), coded 1, with the French workers in Switzerland (i.e., mobile) coded 0, lying in between these groups. The residual contrast (C2) opposed the French workers in Switzerland, coded −2, to the two others groups, the French workers in France and the Swiss, both coded 1. H1 predicted a significant effect of C1 but not C2, thus highlighting a linear effect of the social mobility variable (Judd et al., 2011).

We performed a repeated-measures ANCOVA with personal and group involvement measures of social action as a withinparticipant factor, the two orthogonal contrasts as betweenparticipants factors, and age as a covariate. The findings showed a significant main effect of involvement in social action, F(1,172) = 59.57, p < 0.001, η 2 <sup>p</sup> = 0.26, such that personal involvement (M = 3.78, SD = 1.87) was lower than group involvement (M = 4.67, SD = 1.79). The analysis further produced an interaction between involvement in social action and C1, F(1,172) = 15.02, p > 0.001, η 2 <sup>p</sup> = 0.08, showing that C1 had a significant effect on the motivation to get personally involved in social action, t = −3.61, p < 0.001, but did not impact significantly the group dimension, t = −0.67, p = 0.50. As the interaction between involvement in social action and C2 was not significant, F(1,172) = 0.34, p = 0.56, η 2 <sup>p</sup> = 0.002, the C1 effect can be interpreted as linear effect. As predicted in H1, Swiss participants (M = 3.20, SD = 1.81) were less motivated to get personally involved in social action for the French working in France compared to the French workers in France themselves (M = 4.53, SD = 1.83, t = −3.61, p < 0.001) with the French workers in Switzerland (M = 3.82, SD = 1.81) situated between these two groups. The interaction between involvement measures and age was not significant, F(1,172) = 0.89, p > 0.34, η 2 <sup>p</sup> < 0.01.

We then conducted an ANOVA on the single item of concern for the low-status achieved group, with the two orthogonal contrasts as between-participants factors and age as a covariate. The pattern of results was quite similar to the one observed for personal involvement in social action. The analysis showed a significant effect of C1, F(1,172) = 37.75, p < 0.001, η 2 <sup>p</sup> = 0.18, and a marginal effect of C2, F(1,172) = 3.14, p = 0.078, η 2 <sup>p</sup> = 0.018. Thus, as predicted in H1, we observed a linear tendency showing that Swiss participants (M = 3.03, SD = 1.95) expressed a lower concern for the low-status achieved group compared to the non-mobile French (M = 5.16, SD = 1.97). The French workers in Switzerland (M = 4.60, SD = 1.81) were situated between these two groups, and were closer to the French workers in France than to the Swiss.

Identification with the inherited and the achieved groups. In order to investigate the identity patterns of the non-mobile, the mobile and the high-status group members, we performed a repeatedmeasures ANCOVA with inherited versus achieved identification as a within-participant factor, the three participant groups as between-participants factor (i.e., corresponding to the social mobility variable), and age as a covariate. Results revealed an interaction between group and identification, F(1,172) = 18.61, p < 0.001, η 2 <sup>p</sup> = 0.18 (means are presented in **Figure 1**, left panel). Pairwise comparisons first indicated that mobile participants identified more strongly with their achieved group than the Swiss (p = 0.01) and the non-mobile French (p < 0.001), while the Swiss identified more strongly with their achieved group than the non-mobile participants (p = 0.002), F(2,172) = 15.47, p < 0.001, η 2 <sup>p</sup> = 0.15. Second, we also observed that the Swiss identified more strongly with their inherited group than non-mobile (p = 0.009), and mobile French (p = 0.003), while no differences were observed between these two last groups (p = 0.99), F(2,172) = 5.60, p < 0.01, η 2 <sup>p</sup> = 0.06. We then investigated this interaction by focusing on the difference between identification with the inherited group and identification with the achieved group. Findings revealed that mobile participants identified more strongly with the achieved than the inherited group, F(1,172) = 27.68, p < 0.001, η 2 <sup>p</sup> = 0.14, while the Swiss and the non-mobile French showed a reversed pattern and identified more strongly with the inherited than the achieved group, with respectively, F(1,172) = 6.60, p = 0.01, η 2 <sup>p</sup> = 0.04 (for the Swiss), and F(1,172) = 3.69, p = 0.056, η 2 <sup>p</sup> = 0.02 (for the non-mobile French).

#### Discussion

Consistent with Hypothesis 1, Swiss participants reported lower concern and lower motivation to get personally involved in social action for the French working in border regions of Switzerland, compared to the French working in these regions themselves. Moreover, the attitudes of the frontier workers (i.e., mobile individuals) were situated in between these two groups' attitudes. Thus, we observed that even if the mobile French appeared to take distance from the low-status achieved group, they did not fully assimilate to the Swiss high-status group members. Indeed, they still expressed a higher concern and a higher motivation to get involved in social action compared to what reported the Swiss participants. These discrepancy between French mobile and Swiss participants can be interpreted through the lens of SIT (Tajfel and Turner, 1986). More specifically, by considering that the Swiss constituted an outgroup on both identity dimensions (i.e., inherited and achieved), it is expected that they show less support for the French, as compared to the French (either mobile or non-mobile). From the perspective of mobile individuals, French workers in France also belong to an outgroup, but on

<sup>1</sup>Controlling for gender, professional status, and subjective social status in our analyses on concern for the low-status achieved group and identification did not affect the pattern of results, thus we do not report the details here.

the achieved dimension exclusively. Consistent with previous research (e.g., Kulich et al., 2015), these findings highlight the negative impact of social mobility on the attitudes toward the ingroup. They also provide evidence that the assimilative dynamic toward the high-status group can be a consequence of the social mobility process. Indeed, we observed that even if French mobile participants effectively expressed less concern than their non-mobile counterparts, they still appeared more concerned than the high-status group members (i.e., the Swiss). As discussed in the Introduction, we argue that this attitudinal difference between French mobile and Swiss participants may also have been due to a motivation of Swiss participants to maintain their distinctiveness, potentially threatened by the arrival of mobile individuals.

Of interest, we observed different degrees of involvement in social action for the low-status achieved group on the individual and the group level. Although we found the predicted negative impact of social mobility on the motivation to get personally involved in social action, we did not observe any differences on the motivation for group involvement. These findings suggest that individuals, regardless of their status, acknowledged the disadvantaged conditions endured by the French working in border regions of Switzerland, and that they were favorable toward group involvement in social action. If we consider the normative pro-egalitarian context of contemporary societies, we can apprehend such a support for group involvement as a socially valued opinion resulting in a shared conformism to social norms.

Concerning the identification dimension, we observed a significantly higher identification with the achieved group among mobile French and Swiss workers compared to non-mobile French workers. Consistent with Kulich et al. (2015), this result indicates that mobile participants clearly focus on their achieved higher-valued identity, and distance themselves from their lowstatus group, as it was also observed on the attitudinal dimension. It is also consistent with the literature showing that individuals identify more strongly with ingroups that are more socially valued (e.g., Ellemers et al., 1990; Roccas, 2003). Moreover, still in line with the findings of Kulich et al. (2015), we did not observe any difference in the identification with the inherited group between mobile and non-mobile participants. This suggests that the mobile keep their ties with their inherited group and manage their social mobility through an increase in the identification with the new high-status achieved group. Nevertheless, as we claimed in the Introduction, we believe that the absence of an effect of the identification with the inherited group may be due to the fact that the non-mobile group is quite heterogeneous in the group members' desire to engage in a mobility in the future. Study 2 will address this issue. Finally, comparison of the inherited identity patterns of the Swiss nationals and the mobile individuals revealed a higher identification with the inherited identity by the Swiss compared to the two French groups. This is not surprising as the Swiss have a more positive inherited identity in this intergroup context. Moreover, we observed that the Swiss revealed a preference for their inherited identity as compared to their achieved identity, which was the exact opposite of the pattern observed for the mobile. The difference between the two identification levels was considerably smaller for the Swiss than for the mobile French. This suggests that the mobile French were motivated to emphasize their higher valued identity and to distance it from the lower valued one. The Swiss, although to a smaller extent, focused more on the inherited than the achieved identity. This may be because it is their inherited identity that clearly differentiates them from the French mobile individuals, and so fulfills their need of a positive, but also, distinct social identity (Tajfel and Turner, 1986). Indeed, their achieved identity is more malleable and also shared by a portion of the French and it may thus be considered as less important.

# Study 2

In Study 2, we focused on French nationals in order to examine concern for the inherited low-status group at different stages of the social mobility process. We used the same setting as in Study 1, with three modifications. The main modification consisted in distinguishing, among the French working in France, between those who anticipated social mobility by expressing the desire to work in Switzerland in the future (i.e., mobility

anticipators) and those who did not (i.e., non-anticipators). Second, the target of the involvement dependent measure was the inherited low-status group (and not the achieved lowstatus group). Third, we measured prejudice toward the mobile group. The aim was to test if the mere anticipation of social mobility is sufficient to produce a tendency toward self-ingroup distancing, a phenomenon that should be most prominent among the mobile. As predicted in H2, we expected a linear effect of social mobility on concern for the low-status inherited group, showing the highest concern among non-anticipating individuals, a moderated concern among anticipators, and the lowest concern among mobile individuals. In addition, we investigated whether the differentiation of mobility anticipating, mobility non-anticipating, and mobile individuals, revealed different levels of inherited group identification. As predicted in H3, such differences should, in turn, account for the gap in these groups' ingroup attitudes.

#### Method

#### **Participants and procedure**

Participants were 216 French nationals (137 women and 79 men, Mage = 34.54, SDage = 10.16, ranging from 20 to 61 years of age) living in border regions of Switzerland. We used the same recruitment procedure as in Study 1. After reporting their nationality, we presented a short introductive text in order to prime the status gap between French nationals working in France (low-status achieved group) and French nationals working in Switzerland (high-status achieved group). Following this, participants indicated the country of their employment. Then, participants proceeded to the measures outlined below in chronological order.

#### **Measures**

Social mobility. We distinguished between three groups of different social mobility stages. Participants who worked in Switzerland were categorized as mobile (n = 95). Participants who worked in France (n = 121) were asked to report the extent to which they would like to work in Switzerland in the future (1 not at all to 7 totally, M = 4.69, SD = 2.32). A mediansplit on the responses to this question (median = 5) provided two subgroups of participants: the mobility anticipators (n = 58) who expressed strong desire to work in Switzerland (M = 6.76, SD = 0.43), and the non-anticipators (n = 63; M = 2.78, SD = 1.61) who reported a lower desire for mobility. Thus, the mobile and the anticipators can be considered as "psychologically mobile" because they either have, or are considering, to be mobile, whereas the non-anticipators have not.

Identification with the inherited and the achieved groups. As in Study 1, participants' identification with the inherited and the achieved groups was measured with Leach et al.'s (2008) identification scale. Participants were first asked to state their identification with the inherited group (i.e., French people in general; α = 0.93; M = 4.38, SD = 1.44), and with the achieved group (i.e., workers in France for anticipators and non-anticipators, α = 0.92, M = 4.10, SD = 1.43; or workers in Switzerland for mobile participants, α = 0.89; M = 5.40, SD = 1.11).

Concern for the inherited group. To assess ingroup concern, we measured participants' motivation to get involved in actions aimed to improve the situation of French people living in border regions of Switzerland. As in the first study, we used two items indicating support for group involvement in collective action (e.g., "The French should unite and show solidarity with each other to collectively fight against a decline in their standard of living", r = 0.64, p < 0.001; M = 4.67, SD = 1.79), and two items indicating the motivation to get personally involved in collective action (e.g., "I would be willing to get personally involved to improve the economic and social situation of the French in a precarious situation (e.g., pay more taxes)", r = 0.55, p < 0.001; M = 3.78, SD = 1.87). Both constructs were measured with 7-point scales from 1 not at all to 7 totally.

Prejudice toward frontier workers. In addition, we measured participants' prejudice toward frontier workers with four items. Sample items are: "Because of their special status, frontier workers should pay a solidarity tax to help French people living in border areas and working in France, who are suffering from rising prices (e.g., estate market)" and "Frontier workers only think of their own interest and often forget their origins" (α = 0.71; M = 3.68, SD = 1.58).

Socio-demographic information. Finally, participants reported their gender, their professional status (with in general 62.5% employee, 8.8% entry-level manager, 12% middle manager, 4.6% senior manager, and 12% of missing data), their age, and the subjective status of their occupation (same two items as in Study 1, r = 0.36, p < 0.001, M = 3.86, SD = 1.34).

#### Results

#### **Preliminary analyses**

As in Study 1, we conducted preliminary analyses in order to examine potential socio-demographic differences between the three social mobility groups. A chi-square test showed that gender was not similarly distributed across the three groups, χ 2 (2, N = 216) = 10.26, p = 0.006, Cramer's V = 0.22. Whilst men and women were equally represented among mobile participants, the sample showed an overrepresentation of women among the non-anticipators (73% of women vs. 27% of men) and the anticipators (72.4% of women vs. 27.6% of men). A further chi-square analysis showed no differences in terms of the professional status between the three groups, χ 2 (6, N = 190) = 2.80, p = 0.83. An ANOVA testing the effect of the three groups on subjective professional status of the participants' occupation revealed no significant effect, F(2,216) = 2.00, p = 0.14, η 2 <sup>p</sup> = 0.02. Finally, an ANOVA showed an effect of participants' social mobility stage on age, F(2,216) = 6.28, p = 0.002, η 2 <sup>p</sup> = 0.07, revealing that nonanticipators (M = 37.75, SD = 11.28) were significantly older than anticipators (M = 30.47, SD = 8.88, p < 0.001), and marginally older than mobile participants (M = 34.91, SD = 9.33, p = 0.08). The latter group was also older than the anticipators (p = 0.007). Based on these results, we included gender and age as covariates in all the following analyses<sup>2</sup> .

<sup>2</sup>Again, controlling for professional status and subjective social status in our analyses on ingroup concern, prejudice toward the frontier workers, and group

#### **Hypotheses testing**

fpsyg-08-01451 August 24, 2017 Time: 14:48 # 9

Concern for the inherited ingroup. In order to test H2, which predicts a linear effect of social mobility, we computed two orthogonal contrasts with the social mobility variable. The first contrast (C1 opposed the non-anticipators of social mobility (i.e., French workers in France who do not wish to work in Switzerland in the future), coded −1, to the mobile (i.e., French workers in Switzerland), coded 1. Anticipators (i.e., French workers in France who wish to work in Switzerland) were coded 0 and were thus situated between the two former groups. The residual contrast (C2) tested differences between the anticipators, coded −2, and the two other groups, the nonanticipators and the mobile, both coded 1. As expected in H2, we predicted a significant effect of C1 but not of C2, thus highlighting a linear effect of our social mobility variable (Judd et al., 2011).

We conducted a repeated-measures ANCOVA with the two orthogonal contrasts as between-participants factors, involvement in collective action (group versus personal involvement) as a within-participant factor, and age and gender (coded women −1 and men 1) as covariates. The findings first showed a main effect of involvement, F(1,211) = 165.28, p < 0.001, η 2 <sup>p</sup> = 0.44. Participants reported greater support for group involvement in collective action (M = 5.55, SD = 1.31), than for personal involvement (M = 3.96, SD = 1.67). The interaction between the involvement dimensions and C1 was also significant, F(1,211) = 4.61, p = 0.03, η 2 <sup>p</sup> = 0.02, but while C1 had a significant impact on the personal dimension of collective action, t = −2.62, p < 0.01, its impact on the group dimension was not significant, t = −0.55, p > 0.58. Moreover, we also observed a significant interaction between the involvement dimensions and C2, F(1,211) = 6.45, p = 0.01, η 2 <sup>p</sup> = 0.03, showing a significant impact of C2 on the group dimension of collective action, t = −2.26, p = 0.02, indicating that the anticipators expressed a higher support for the collective action (M = 5.84, SD = 1.20) from the group than the support of the non-anticipators and the mobile combined (respectively, M = 5.46, SD = 1.23 and M = 5.38, SD = 1.40). As C2 was not significant for the personal dimension, the effect of C1 on the personal dimension could be interpreted as linear effect: As expected in H2, the non-anticipators (M = 4.40, SD = 1.63) were more motivated to get personally involved in collective action than the mobile (M = 3.69, SD = 1.73), with the anticipators (M = 3.92, SD = 1.55) situated between these two groups. Finally, the analysis produced a significant interaction between gender and involvement, F(1,211) = 4.09, p = 0.04, η 2 <sup>p</sup> = 0.02. However, none of the pairwise comparisons reached significance for this interaction (ps > 0.27, η 2 <sup>p</sup> < 0.006). All other effects were non-significant (ps > 0.29, η 2 <sup>p</sup> < 0.005).

Prejudice toward the frontier worker status. We performed an ANCOVA on the prejudice expressed toward frontier workers with the two contrasts as between-participants factors, and gender and age as covariates. The findings revealed a main effect of C1, F(1,216) = 41.73, p < 0.001, η 2 <sup>p</sup> = 0.16, but not of C2, F(1,216) = 1.44, p = 0.23, η 2 <sup>p</sup> < 0.01. Non-anticipators (M = 4.40, SD = 1.54) reported more prejudice toward frontier workers than mobile participants (M = 2.95, SD = 1.28), with anticipators (M = 4.09, SD = 1.60) situated between these two groups. Consistent with H2, we therefore observed a linear effect of the social mobility variable on the prejudice toward frontier worker status.

Identification with the inherited and the achieved groups. In order to test H3, which predicts that inherited identification should explain anticipators' and mobile's lower ingroup concern (i.e., compared to non-anticipators), we performed PROCESS Model 4 mediation analysis, using 10,000 bootstrapped samples following Hayes' 2013 recommendations. The model included C1 (non-anticipators versus anticipators versus mobile) as a predictor, personal involvement in collective action as the dependent variable, identification with the inherited group as potential mediator, controlling for C2 (anticipators versus nonanticipators and mobile participants), gender, and age (see full results in **Table 1**). The analysis revealed a significant effect of mobility stage (i.e., C1) on identification with the inherited group (path a: B = −0.34, SE = 0.12, p = 0.007), which in turn was positively associated with personal involvement (path b: B = 0.30, SE = 0.08, p < 0.001). Moreover, identification with the inherited group proved to be a significant mediator, CI 95% [−0.23, −0.02], of the relation between social mobility (C1) and the motivation to get personally involved in collective action for the ingroup. The direct effect of C1 became marginally significant when controlling for the mediator (path c': B = −0.28, SE = 0.14, p = 0.051). A Sobel test confirmed that the difference between path c and path c' was significantly different from 0 for the indirect effect of the identification with the inherited group, z = −2.16, p = 0.03, corroborating the mediating role of the identification with the inherited group. In sum, this study provides evidence of an identity discount strategy in social mobility trajectories. Findings are graphically represented in **Figure 2**.

In addition, in order to test the replicability of the identification pattern observed among non-mobile and mobile participants previously observed in Study 1, we performed a repeated-measures ANCOVA with inherited versus achieved identification as a within-participant factor, and the three mobility groups as between-participants factor, controlling for age and gender. Means are displayed in the right panel of **Figure 1**. Results only revealed an interaction between group and identification, F(1,211) = 31.5, p < 0.001, η 2 <sup>p</sup> = 0.23. Pairwise comparisons first showed that for achieved identification [F(2,211) = 39.34, p < 0.001, η 2 <sup>p</sup> = 0.27] mobile participants identified more strongly with the achieved group than the nonanticipators (p = 0.007), and the anticipators (p < 0.001), and the anticipators identified less than the non-anticipators (p < 0.001). Second, pairwise comparisons showed for inherited identification [F(2,211) = 4.32, p = 0.01, η 2 <sup>p</sup> = 0.04] that while anticipators and mobile participants identified to a similar extent (p = 0.96), they both identified less with their inherited group than non-anticipators (p = 0.01 for the anticipators,

identification did not affect the pattern of results, thus we have not reported the details here.

Chipeaux et al. Anticipated Upward Mobility

TABLE 1 | Identification with the inherited group as mediator of the relationship between social mobility stages and personal involvement in collective action (Study 2).


C1, Non-anticipators (coded −1) vs. Mobile (coded 1; with Anticipators coded 0); C2, Anticipators (coded −2) vs. Non-anticipators and Mobile (both coded 1); Estimates are unstandardized; †p < 0.10, <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

and p = 007 for the mobile). Finally, we also examined the discrepancy between identification with the inherited group and identification with the achieved group. Findings showed that the mobile identified more with the achieved than with the inherited group, F(1,211) = 61.58, p < 0.001, η 2 <sup>p</sup> = 0.23, and that the reverse pattern occurred for the anticipators, F(1,211) = 11.6, p = 0.001, η 2 <sup>p</sup> = 0.05. No difference between identification to the achieved and the inherited groups was observed for the non-anticipators, F(1,211) = 0.28, p < 0.60, η 2 p = 0.001.

#### **Discussion**

Study 2 investigated the role of the mobility stage on self-distancing from the inherited ingroup, by looking at ingroup attitudes and identification. The novelty of Study 2 is that it took a more fine-tuned perspective on the nonmobile group by distinguishing between those who desired to engage in social mobility in the future and those who did not. Such analysis provided a preliminary insight on the crucial role of the mobility stage on ingroup attitudes and identification.

On the attitudinal dimension, we observed different effects of social mobility on the support for group involvement in collective action and on the motivation to get personally involved in it. First, results revealed the unexpected effect that anticipators of social mobility expressed a greater support for group involvement in collective action compared to the two other groups aggregated. Such finding highlights the particular dissatisfaction of anticipators regarding the fate of their inherited membership. In line with our expectations, results on personal involvement in collective action further revealed that anticipators preferred to focus on their personal trajectory in order to improve their chances to enhance their social identity rather than to join the group in its claim. This result is consistent with Taylor and McKirnan's (1984) model of social mobility stages arguing that low-status group members would only act collectively if they had failed to individually mobilize. Indeed, as predicted by H2, even if anticipators were more motivated to get personally involved compared to the mobile, they were less motivated compared to the non-mobile, preferring to focus on an individualistic strategy to improve the value of their social identity. Consistent with previous evidence in the literature showing a negative impact of the social mobility process on ingroup concern (Derks et al., 2011a,b, 2015; Kulich et al., 2015), mobile individuals showed the weakest levels of personal involvement. Finally, moving a step further, the present findings demonstrated that the mere anticipation of social mobility is sufficient for triggering a decrease in ingroup concern. We thus argue that experiencing the socialization process of social mobility per se is not a necessary condition for lower ingroup concern, but that imagining the possibility to be socially mobile, thus a purely psychological process, is sufficient to engage in attitudinal change.

In this study, we also assessed prejudice toward the mobile group. In parallel to what was observed on ingroup concern, the findings showed that prejudice toward frontier workers was also contingent on the social mobility stage. Again, we observed a linear effect of the social mobility, such that nonmobile participants expressed higher prejudice toward frontier workers than mobile individuals, with the anticipators of social mobility situated between these two groups. Although nonmobile participants had to evaluate their most direct and relevant outgroup, individuals who anticipated social mobility expressed less prejudice toward this group, compared to

the non-anticipators. Consistent with Merton's theorization concerning anticipatory socialization, these findings highlight the positive orientation individuals develop toward an outgroup they aspire to belong to (Merton, 1968). Moreover, the fact that non-anticipators have meanwhile demonstrated a higher level of prejudice toward frontier workers may be related to the previous literature investigating the reaction toward deviance, and particularly the "black sheep effect" (Marques et al., 1988; Pinto et al., 2010). According to this literature, deviance tends to be more severely punished when it comes from an ingroup member than when it comes from an outgroup member. Individuals indeed perceive the deviant's behavior as threatening to the identity of the ingroup. By the rejection of this behavior and its actor, they reaffirm the ingroup's standards and contribute to the longevity of the group. Thus, it is not surprising that non-anticipators had unfavorable attitudes toward frontier workers who are ultimately perceived as betrayers, preferring to improve their own status while the whole group continues to suffer from inferior conditions (Blair and Jost, 2003).

As for group identification, novel insights were obtained through the distinction between non-anticipators and anticipators in the non-mobile group. Indeed as expected in H3, the linear decrease of ingroup concern observed throughout social mobility stages was accounted for by a lower identification with the inherited group. Such a finding suggests an identity discount strategy, as derived from SIT assumptions. This strategy points to individuals who distance themselves from their inherited low-status ingroup on both the attitudinal and the identification dimensions (Ellemers, 2001). Extending past research (e.g., Derks et al., 2011b, 2015; Kulich et al., 2015), the findings from Study 2 illustrate the willingness of individuals to increase their chances to attain a better valued social identity through individual mobility, despite the fact that they are unable to actually part with their low-status membership. Moreover, by revealing similar levels of ingroup identification among anticipators and mobile individuals (even though these two groups still differed in their ingroup concern), our data support the idea that ingroup identification plays a crucial role in the process of social mobility (Tajfel and Turner, 1986; Ellemers, 2001). This is also in line with Taylor and McKirnan's (1984) claim that members of low-status groups who perceive themselves as competent, and so as non-prototypical of their low-status membership, will try everything possible to dissociate themselves from this group.

Consistently, Study 2 revealed the same identification pattern for the mobile French as in Study 1. The mobile identified more with their higher valued achieved group than with their lower valued inherited group. As for the non-mobile individuals, their identification pattern strongly depended on their desire to engage in social mobility. Indeed, those who wished to be socially mobile were less identified with both the inherited and the achieved groups compared to the nonanticipators. These results thus highlight the conflict anticipators may feel between their desire to improve their condition and their actual low-status memberships (Lenski, 1966). Moreover, the non-anticipators were more identified with the inherited group compared to the mobile and the anticipators. Thus, the absence of a difference in inherited identification between non-mobile and mobile participants in Study 1, as well as in Kulich et al. (2015), may have been due to the fact that all non-mobile individuals were treated as one group, thereby mixing two groups (anticipators and non-anticipators) of very different identification patterns. In support of this reasoning, the marginal difference observed in Study 1 between inherited and achieved identification became significant only for the anticipators in Study 2. In sum, it seems that no distance in the identification occurs for non-mobile non-anticipators but a clear motivation to distance between the two arises for anticipators.

The correlational nature of the present research limits the interpretation of causal relationships between social mobility, attitudes and identification. This study only shows results from anticipators and mobile individuals but not the actual process of people who move from the anticipator to the mobile stage. Thus, at least two different mechanisms could be responsible for the patterns observed for anticipators. First, they may start to disidentify from the unwanted achieved group in order to replace it by a higher identification with the new highstatus group as soon as they actually successfully engage in social mobility. Interestingly, we observed in line with this idea that anticipators showed in Study 2 the same identification with the inherited group as the mobile participants. As previously discussed, this may describe a psychological strategy through which anticipators are adapting to a potential social mobility rather than behaving like other non-mobile individuals. According to Sidanius and Pratto (1999), such individual orientation rests on meritocratic beliefs, which conceptualize as legitimating myths that contribute to protect the social hierarchy by valuing individualistic behavior and strengthening the unequal treatment of members of the two groups. Moreover, this is also in line with the theorization of Merton (1968) and the results we observed on the attitudes toward frontier workers in Study 2. Second, an alternative explanation that cannot be completely discarded is that anticipators and mobile individuals differ from non-mobile due to previous experiences or socialization processes. Controlling for all the possible differences in socialization, attitudes, and employment histories that may exist between the three investigated French groups is an important but difficult task. Future research should thus aim at experimentally manipulating social mobility in order to investigate its impact on attitudes and identification. Longitudinal studies could also be informative as they would allow to assess actual changes in attitudes and (dis)identification patterns.

In addition, further research is needed to determine the extent to which the present findings in ingroup concern can be generalized to broader intergroup attitudes, such as ingroup bias and prejudice expression. Indeed, the measures we used in this research were targeting the personal interests of individuals (e.g., the reduction of the costs of life in the French border regions of Switzerland), thus maximizing the differences between the three groups of participants.

Another limitation of this study is its incapacity to provide insights concerning the consequences of such identity discount strategy on the quality of the relationship mobile individuals maintain with their inherited group members. Indeed, it is reasonable to think that by being simultaneously lowly identified with the inherited ingroup and expressing lower concern for it, mobility anticipators take the risk of being judged as disloyal (Blair and Jost, 2003). This may in turn motivate ingroup members to devalue them as a way to punish the deviance and at the same time to reaffirm the norm. Paralleling this idea, research demonstrates that when women attain high-responsibility positions in the workplace, meaning that they successfully achieved social mobility in a male-dominated domain, they are perceived as less communal than women in general (see for example Rudman and Glick, 2001). Research conducted by Heilman and Okimoto (2007) illustrated that unexpected (i.e., gender incongruent) competence demonstration of agentic women led to punishment in hiring procedures. However, by adding information reaffirming the communality of the female candidates (i.e., by emphasizing their mother status), the selective bias against agentic women decreased as they appeared more stereotypical. Consistently, Phelan et al. (2008) showed that whereas perceived competence was the most important factor predicting the selection of candidates in hiring procedures, the criteria of selection shifted when they came to concern agentic female candidates: Rather than being evaluated based on their competence, they were evaluated based on their social skills, thus being punished if they did not live up to expectations that women should be socially skilled.

# GENERAL DISCUSSION

This research was aimed at providing further insights about the deleterious impact of upward mobility on attitudes toward the inherited low-status ingroup. In Study 1, we observed that socially mobile individuals expressed negative attitudes toward non-mobile group members, simultaneously with a higher identification with their achieved group. In parallel, inherited identification was not different between these groups, a pattern that illustrates their desire to assimilate to the high-status group. On the one hand, mobile individuals have the willingness to improve their social condition, but on the other hand they are bound by the inevitable membership in a low-status group. Of interest, Van Laar et al. (2014) showed that members of low status groups and those of high social status groups do not offer support for mobile individuals under the same conditions. On the one hand, highstatus group members appear to be more sensitive to the behavioral dimension than the affective one, preferring mobile individuals who did not behave as a prototype of their lowstatus ingroup regardless of their identification with this group. On the other hand, low-status group members prefer mobile individuals who keep a strong identification with their ingroup, regardless of their level of behavioral prototypicality, or their competence (Campos et al., 2016). In light of these recent works, the negative ingroup attitudes expressed in the present research by the French mobile together with the expression of affective proximity with this group can be apprehended as a strategy to increase their integration in the high-status group, without breaking the affective ties they have with their inherited ingroup.

Extending these findings to actual social mobility, Study 2 further examined the impact of the willingness to engage in social mobility among individuals who have been so far nonmobile. Of interest, results showed that the mere anticipation of social mobility was sufficient to produce a lower concern for the ingroup. In addition, this tendency appeared to be due to an identity discount strategy (Tajfel and Turner, 1986; Ellemers, 2001). Indeed, we observed that the lower ingroup concern expressed by the mobile and, to a lesser extent by the anticipators, was accounted for by their lower identification with their inherited group. Therefore, by distinguishing between non-mobile individuals based on their willingness to undertake mobility, we provided further insights about the process of social mobility. Despite recent findings illustrating the maintenance of identification with the low-status group and its coexistence with high levels of identification with the high-status group (Kulich et al., 2015), our research rather emphasizes an identity discount strategy as a privileged way for mobile individuals to cope with their status-inconsistent identity configuration. Nevertheless, little is known about the mechanisms leading social mobility anticipators, as well as mobile individuals to engage in such a coping strategy. Further investigations are thus needed in order to unravel such social dynamics and to identify precisely the conditions favoring the expression of such an identity discount.

In summary, it seems that individuals anticipating upward mobility follow the principle "I want – therefore I am". Indeed, they already start to dissociate from their low-status group, not only through their (more negative) attitudes toward it, but also through their level of identification with the inherited group, as characterized by the identity discount strategy uncovered in this research. This attitudinal and identity discounts allow them to reduce the dissonance they may experience as a result of the asymmetrical statuses of their different groups' membership. The felt dissonance could even be stronger than the one experienced by mobile individuals, because of the coexistence of their motivation to improve their condition and the unescapable nature of their inherited membership.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of 'Ethical code concerning research at the Faculty of Psychology and Educational Sciences at the University of Geneva, Ethical Commission' with written informed consent from all subjects. In all studies, participants ticked a box before starting the survey and after finishing it, indicating their informed consent and agreement to use their responses for research purposes. The protocol was approved by the 'Ethical Commission of the Faculty of Psychology and Educational Sciences at the University of Geneva'.

# AUTHOR CONTRIBUTIONS

fpsyg-08-01451 August 24, 2017 Time: 14:48 # 13

MC, CK, VI, and FL-C contributed to the study design. MC collected the data. MC performed the data analysis and interpretation under the supervision of CK, VI, and FL-C. MC drafted the manuscript, and CK, VI, and FL-C provided revisions.

# FUNDING

The work undertaken by the MC, CK, and FL-C on this paper was supported by the Swiss National Science Foundation grant

# REFERENCES


(N◦ 100014\_149197/1) for the project "Negotiating identities: Minority groups handling social status improvements".

# ACKNOWLEDGMENTS

The authors of this paper would like to thank Dr. Cristina Aelenei for her advice regarding the interpretation of statistical analyses, and the following students for their significant contribution during the process of data collection: Ekaterina Abramova, Priscilia Barras, Fabio Bezerra, Agathe Bidaud, Fatima Gazijeva, Abbas Kanani, Fabio Mason, Elodie Michel, Agathe Perrichon, Giulia Valsecchi, and Kevin Vezirian.



in low status groups. Eur. J. Soc. Psychol. 44, 563–577. doi: 10.1002/ejsp. 2046


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Chipeaux, Kulich, Iacoviello and Lorenzi-Cioldi. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Flexible Loyalties: How Malleable Are Bicultural Loyalties?

Andy Y. Chiou<sup>1</sup> \* and Brittany K. Mercado<sup>2</sup>

*<sup>1</sup> Business Management Department, Farmingdale State College (SUNY), Farmingdale, NY, USA, <sup>2</sup> Department of Management, Baruch College (CUNY), New York, NY, USA*

Biculturals are individuals who are acculturated in two cultures and have dual identities. Due to this, many early discussions on biculturalism argued that biculturals may have divided loyalties between their two cultural backgrounds and the identities derived from these backgrounds. This view is further highlighted given historical and contemporary debate regarding immigrants in the European and American political arenas. These concerns illustrate two possibilities. First, that biculturals have a preference for their home or host culture, identifying one as the in-group to express loyalty toward and the other as the out-group. Second, biculturals may alternate between who they identify as their in-group depending upon the circumstances. In a particular cultural environment, a given bicultural may feel greater degrees of loyalty toward that culture, while feeling different loyalties when immersed in a different cultural environment. To-date, few empirical studies have examined these two questions in detail. We proposed two hypotheses: First, biculturals will express higher levels of loyalty for a specific culture if they have been exposed to a prime congruent with that culture than if they have been exposed to a prime associated with a different culture. Second, the magnitude of preferences expressed for the two cultures will differ depending on the cultural prime. We experimentally investigated this phenomenon in a sample of Chinese-Americans (*N* = 136) using a computer simulated soccer game between the United States and China. This simulation was selected in order to avoid the controversial nature of an immigration or cultural conflict scenario. Past research has shown that support for the sports team of a given country is a form of expressing loyalty. Participants were randomly exposed to one cultural priming condition (American, Neutral, Chinese) using commentaries recorded in different languages: English, no commentary, and Chinese. Participants were then asked to what degree they would cheer for each team. Participants expressed more likelihood to cheer for the Chinese team than for the American team. However, our results indicate that cultural priming does influence the degree to which the participants express loyalty for the Chinese team over the American team in the form of rooting behaviors.

Keywords: biculturalism, bicultural identities, social Identities, loyalty conflicts, cultural priming

# INTRODUCTION

Many early discussions on biculturalism have argued that biculturals, or individuals who have internalized two cultures (Brannen and Thomas, 2010), are sometimes seen as marginal individuals who possess divided loyalties and are ambivalent toward either of their cultural identities (cf. LaFromboise et al., 1993). Although in recent decades research on biculturalism has greatly

#### Edited by:

*Fabio Lorenzi-Cioldi, University of Geneva, Switzerland*

#### Reviewed by:

*Liudmila Liutsko, Park of Biomedical Research in Barcelona, Spain Ilka H. Gleibs, London School of Economics and Political Science, UK*

> \*Correspondence: *Andy Y. Chiou chioua@farmingdale.edu*

#### Specialty section:

*This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology*

Received: *02 August 2016* Accepted: *06 December 2016* Published: *20 December 2016*

#### Citation:

*Chiou AY and Mercado BK (2016) Flexible Loyalties: How Malleable Are Bicultural Loyalties? Front. Psychol. 7:1985. doi: 10.3389/fpsyg.2016.01985* expanded our understanding of bicultural cognition and identity organization (Hong et al., 2000; Benet-Martínez et al., 2002), while bringing to question the very concept of biculturals as marginal individuals (Pilar and Udasco, 2004), less research has been conducted on the extent to which biculturals identify with either of their cultural backgrounds. This question is of practical importance given issues prevalent in the United States, such as the forced relocation of Japanese-Americans during the Second World War, the anti-Muslim sentiment faced by Muslim- and Arab-Americans (Krieg, 2015), and allegations of espionage faced by Asian-Americans (Purdy, 2001; Asher-Shapiro, 2015; Stewart, 2016) in the twenty first century. That these types of suspicions have so frequently fallen on biculturals in the United States has caused concern and spurred investigation (Asher-Shapiro, 2015).

Despite our current understanding of bicultural cognition and identity, the question of whether biculturals have mixed loyalties still remains. This is critical given increasing cultural complexities not only in the United States, but also in the rest of the world. Biculturals, and by extension multiculturals, will increasingly be placed in positions where they may be seen as needing to choose between their ethnic or the mainstream culture, either at work (e.g., cross-country business negotiations or resource allocation between subordinates) or in their social lives (e.g., support for various political issues). Therefore, to launch investigation into this important issue, we propose the following research question: Given the dual cultural backgrounds that biculturals possess, do they have a preference for their home/ethnic or host/dominant culture, identifying one as the ingroup and the other as the out-group when exposed to a specific cultural environment, or do they alternate between who they identify as their in-group depending upon the circumstances?

# BICULTURALISM

Biculturals are individuals who have gained fluency in two different cultures and internalized both; this fluency provides them with the knowledge of a particular cultural setting necessary to adjust their behavior in accordance with that cultural setting (Mok and Morris, 2009; Mok et al., 2010). These individuals are able to switch between different cultural world views depending on the cultural stimuli they receive (Hong et al., 2000), a phenomenon that extends to the switching of personalities (Ramirez-Esparza et al., 2006; Chen and Bond, 2010).

Bicultural individuals can gain fluency in two cultures through the acculturation process, broadly defined as the "cultural and psychological change that follows intercultural contact" (Berry et al., 2006, p. 305). Individuals who immigrate from one society to another face the challenge of interacting with the host culture while maintaining connections with their home culture. It is this acculturation process that allows biculturals to accumulate cultural knowledge and gradually create cultural identities in two different cultures, allowing them to become bicultural. It should be noted that current research suggests acculturation is not necessarily a unidimensional concept, in which the host culture gradually subsumes the home culture, but instead a bidimensional model, in which acculturation in the home and host cultures must be considered separately (cf. Tsai et al., 2000; Nguyen and Benet-Martínez, 2007). Biculturalism has traditionally been viewed as an integrated acculturation strategy (Nguyen and Benet-Martínez, 2007), with consensus in the literature that biculturalism relies on individuals acculturating in and identifying with both home and host cultures. Although an argument may be made that second generation biculturals who are born in a host country differ from first generation biculturals who emigrated from a home country to a host country (Berry and Sam, 1997; Phinney et al., 2001; Liebkind, 2003), current bicultural research has thus far demonstrated that these two groups of biculturals display similarities in terms of cognition (Mok and Morris, 2010; Mok et al., 2010) and identity (Miramontez et al., 2008).

Biculturalism can also occur in other forms, driven primarily by globalization. Arguments have been made that Hong Kong citizens also represent a form of biculturalism due to globalization and British colonization (Bond and Cheung, 1983; Bond, 1993; Hong et al., 2000). While Nguyen and Benet-Martínez (2013) note that Hong Kong biculturals are different due the dominant and heritage cultures being the same, i.e., Asians living in an Asian society, in comparison to immigrant biculturals in which the dominant and heritage cultures are different, e.g., Asians in American society, bicultural priming studies that have succeeded in eliciting different cultural frames (Hong et al., 2000) and identities (Brewer et al., 1987; Ng and Lai, 2009, 2011) in Hong Kong biculturals.

No matter what processes lead to biculturals, whether through immigration or globalization, research has informed us that all biculturals face the need to balance the requirements of their own ethnic culture and the host culture in a constant negotiating act (Nguyen and Benet-Martínez, 2007). This balancing of cultural identities may also be malleable. Extending earlier evidence that biculturals can be primed to see the world through different cultural lenses (cf. Hong et al., 2000), scholars have demonstrated that bicultural identities can be influenced via cultural priming. There are many ways in which culture can be primed, ranging from pronoun circling tasks, scrambled sentence tasks, subliminal tasks (cf. Oyserman and Lee, 2008 for a detailed discussion), to cultural icons (Hong et al., 2000) and language (Ramirez-Esparza et al., 2006; Chen and Bond, 2010). Social identity theory (SIT; Tajfel and Turner, 1979) and selfcategorization theory (SCT; Turner et al., 1987), jointly referred to as the social identity approach (SIA), provide insights into influences on the cultural identities of bicultural individuals. Individuals derive a social identity from groups they belong to, leading them to positively evaluate their own group compared to other groups (Tajfel and Turner, 1979). Of equal importance, individuals also possess multiple self-concepts which may be activated depending upon the specific situation; one particular aspect of an individual's self-concept may become more salient in a social context for which said self-concept exhibits the greatest fit (Turner et al., 1987; Hogg et al., 1995). That individuals can recognize multiple in-groups is a phenomenon which has been empirically established (Brewer et al., 1987; Hewstone et al., 1993; Brewer, 1999; Roccas and Brewer, 2002; Kang and Bodenhausen, 2015).

Verkuyten and Pouliasi (2006) further showed that depending on the cultural primes Greek-Dutch biculturals were presented with, they self-identified more as either Greek or Dutch. Greek primes elicited greater self-identification as Greek, while Dutch primes elicited greater self-identification as Dutch (Verkuyten and Pouliasi, 2006). Similar studies have been conducted on Hong Kong biculturals (Ng and Lai, 2011), Mexican-Americans (Devos, 2006), and Chinese-Australians (Liu, 2015), further empirically demonstrating the effects of cultural primes on cultural identification. The priming of individual social identities can also influence decision making, with individuals making decisions that are congruent with the values of the primed identity (LeBoeuf et al., 2010). One possible form in which individuals can engage in such decision making is group loyalty.

Group loyalty in a social identity context can manifest in many different ways, either affective, cognitive, or behavioral (Van Vugt and Hart, 2004). Affective manifestations can relate to the experience of positive emotions through group membership, cognitive manifestations can relate to overall trust in group members, and behavioral manifestations can relate to sacrificing of the self for the betterment of the group (Van Vugt and Hart, 2004). Because acculturation is the process by which individuals gradually identify with given cultures, we can expect biculturals to exhibit various manifestations of group loyalty with the acculturated cultures. Although the aforementioned studies (Devos, 2006; Verkuyten and Pouliasi, 2006; Ng and Lai, 2011; Liu, 2015) have demonstrated that the self-perceived cultural identity of biculturals can be made malleable depending on the cultural context they are exposed to Hogg et al. (1995) review of identity and social identity theory noted the importance of determining how individuals behave toward out-groups in social identity research. As such, little research has been conducted on how biculturals behave toward other individuals or cultural groups given the activation of a cultural identity. Additionally, no studies have yet to place biculturals in a situation where the two cultures they belong to are in direct competition with each other, thereby forcing biculturals to make a direct comparison of loyalties between the two cultural groups when they have been primed with a particular cultural prime.

For biculturals who identify with two different cultures, group loyalty can be complicated. Because they have two possible cultural identities with corresponding in-groups, they could potentially have loyalties toward two cultural groups. This dilemma is amplified in situations when biculturals may be forced to choose which of the two cultural groups to express loyalty toward. However, if the cultural identities of biculturals can be primed, then cultural primes should influence how biculturals express loyalties toward different cultural groups. Biculturals should reflect this malleable identification by expressing increased loyalty for the culture of the primed identity.

Hypothesis 1: Biculturals will express higher levels of loyalty for a specific culture if they have been exposed to a prime congruent with that culture than if they have been exposed to a prime associated with a different culture.

Hypothesis 2: The magnitude of preference that a bicultural exhibits toward one culture over another culture will be influenced by the cultural prime.

It should be noted that while acculturation may involve the accumulation of cultural knowledge and facilitate the eventual identification with a given culture, (cf. Nguyen and Benet-Martínez, 2007), cultural priming will not have an effect on the level of acculturation in a given bicultural. Cultural priming activates the cultural knowledge and identity that has been developed as a result of acculturation (Hong et al., 2000), and therefore does not effect the level of acculturation itself.

# METHODS

In order to determine whether biculturals could be primed to express loyalties toward different cultural groups, we designed an ostensibly low-stakes experimental study involving a soccer game. A sports game was chosen in order to avoid any sensationalism that may be associated with asking participants to respond to scenarios similar to the examples of discrimination discussed earlier, while still presenting the participants with two cultural groups engaging in competition. Sports psychology research has noted that individuals tend to root for athletes and sports teams that represent their identified in-group (Branscombe and Wann, 1992; Branscombe et al., 1993), a tendency that has been shown to be an expression of loyalty (Gwinner and Swanson, 2003). This design also served to mirror Branscombe and Wann's (1992) study, which used a boxing match between the United States and former Soviet Union from Rocky IV to assess participants' emotional reactions to two competing countries. A soccer game was selected for our study in order to avoid sports that may be strongly associated with a given culture (e.g., baseball with American culture or ping-pong with Chinese culture). The researchers were entirely self-funded for the entirety of this study. All experimental procedures were conducted in accordance with the relevant Institutional Review Board and in a manner congruent with the ethical guidelines of the American Psychological Association.

# Participant Recruitment

The participants for this study were recruited from an undergraduate research participation pool of a university located in a large metropolitan city in the northeastern United States. The recruitment materials posted on the participant pool website requested self-identified Chinese-Americans who were fluent in Chinese-Mandarin to participate in a study on decision making.

A total of 188 participants signed up for the study. After filtering out incomplete responses, non-Chinese-American participants, and participants who did not finish the survey between 10 and 30 min, 136 responses remained. Incomplete responses were identified automatically by the online survey website. Non-Chinese-American participants were identified through a combination of self-identified ethnicity and the cultural background of both parents. Responses were filtered for response time as analysis of the response times showed wide variance, ranging from less than 1 min to over 60 h. This filter was utilized to ensure that participants did not proceed with the tasks too quickly, which would mean insufficient time viewing the stimulus materials, and also to ensure that participants did not take too much time and return to the study after a lengthy interruption, which would compromise the priming effect. Mean age was 21.69 years (SD = 3.07); 51.1% of the participants were female. Most of the participants were born overseas (73.68%), with 24.81% having been born in the United States, and 1 participant declined to disclose the country of birth. Of the participants born overseas, the average number of years spent in the United States was 10.68 years (SD = 6.85).

# Procedure

All participants completed study materials online, participating at times and locations of their convenience. The survey website was programmed in a manner that provided a controlled experimental environment in which participants were exposed to the cultural priming materials, described in more detail in the cultural priming section, for a set amount of time. The content and order of the study materials were presented to the participants as follows.

#### Cultural Priming

Participants were randomly assigned to one of three cultural priming conditions: American, Chinese, and a Neutral priming condition for comparison. A combination of graphical and linguistic priming methods was used in this study in order to reinforce the cultural priming effect. The graphical priming materials were drawn from Mok's bicultural studies (**Figure 1**; Mok and Morris, 2012, 2013). Within each condition, each image was shown to participants for a minimum of 5 s before participants were allowed to proceed to the next image, in order to ensure that the participants properly processed the contents of each image. The minimum time was enforced by a timing function within the survey software. Given the possibility that language itself may serve as a prime (Ramirez-Esparza et al., 2006; Oyserman and Lee, 2008; LeChuga and Wiebe, 2009; Chen and Bond, 2010), in this study the cultural icons were paired with culture-congruent textual and audio languages. For the American priming condition all materials were presented in English text. For the Chinese priming condition, all materials were presented in Traditional Chinese-Mandarin text. For the intended neutral priming condition all materials were presented in English text due to the necessity of presenting participants with experiment text. Given the general environment surrounding the study, English was selected as the default language to use.

#### Scenario

Following the cultural prime, participants were informed they would be shown a simulation of a soccer game and be asked to answer a few questions about the audio commentary that had accompanied the game and the game itself. The scenario was adapted from the methodology used in Branscombe and Wann's (1992) study. We presented each participant with a short 5-min video clip of a soccer game involving an American team and a Chinese team. The video clip was compiled from recordings of a simulated soccer match between the United States and China from FIFA 11, a soccer video game. The soccer match was run as an automated simulation between the two teams, utilizing the video game's artificial intelligence (AI) with no input from the researchers. The resulting video clip consisted of passes between the AI players of the respective countries, interceptions, and attempts at goal shots; however, no scoring shots were included in the clip in order to provide the perception of an inconclusive game. The cultural priming was further reinforced through the usage of voiceover commentary for the video clip recorded by native speakers of each language. Both commentators read from a pre-written script that described the same actions displayed in the

video clip. Although all participants were shown the exact same video graphics, the audio commentary that accompanied the video clip differed between priming conditions. For the American priming condition the audio commentary was in English, while for the Chinese priming condition the audio commentary was in Chinese-Mandarin. For the Neutral condition, no audio commentary was provided.

#### Expression of Loyalty

Following the soccer game, we assessed participants' expression of loyalty for each of the teams engaged in the simulated match by measuring their rooting behaviors. Due to the possibility that participants may root for both teams despite the cultural priming, we assessed rooting behavior for the participants toward both teams. Participants were asked to indicate to what extent they would root for the American and Chinese teams. This was measured by two items: "I would cheer for the US team," to assess rooting behavior for the American team, and "I would cheer for the Chinese team," to assess rooting behavior for the Chinese team. Participants responded on a 5-point Likert scale, with 1 denoting "Extremely Disagree" and 5 denoting "Extremely Agree." The two items were presented in randomized order.

#### Behavioral Acculturation

To assess participants' levels of acculturation with American and Chinese cultures, Tsai et al. (2000) General Ethnicity Questionnaire (GEQ) was adopted for this study. The GEQ is a 39-item, 5-point Likert scale (1 = Strongly Disagree, 5 = Strongly Agree) measure of acculturation that can be tailored to specific cultures, with items such as "I listen to American [Chinese] music," and "I celebrate American [Chinese] holidays" (GEQ-American, α = 0.87, GEQ-Chinese, α = 0.85 in the present sample). All participants completed two versions of the GEQ: American and Chinese, presented in randomized order. The acculturation measures were presented after the presentation of the main independent and dependent variables so as to not inadvertently prime the participants. Analysis afterwards showed no significant differences in acculturation between the two cultural priming conditions, indicating that, as expected, results obtained from the acculturation measures were not influenced by the cultural primes.

#### Demographics and Background Variables

At the conclusion of the study, participants were asked to complete a demographic questionnaire, including items about their cultural backgrounds and their parents' country of origin. Participants were also asked to provide information on the types of sports they typically watched, whether they were familiar with soccer, and whether they supported any particular soccer teams.

# RESULTS

Participants in this study were overall more acculturated with Chinese culture (M = 3.78, SD = 0.43) than American culture (M = 3.25, SD = 0.42), t(136) = −10.31, p < 0.05, Cohen's d = −0.88, in line with past studies on biculturals that have also shown similar acculturation patterns (Tsai et al., 2000; Benet-Martínez et al., 2002; Benet-Martínez and Haritatos, 2005; Cheng et al., 2006; Miramontez et al., 2008; Zou et al., 2008). Therefore, despite the differences found in the levels of acculturation in American and Chinese cultures, the sample was bicultural due to being acculturated with both cultures to some extent. The correlations for the variables investigated, along with the means and standard deviations for each, are presented in **Table 1**.

Participants were overall not very familiar with soccer (M = 1.99 on a 5-point Likert scale, SD = 0.90, with "2" on the scale being "A little."), and out of those that expressed support for specific soccer teams, none indicated support for American or Chinese soccer teams. There were also no significant differences between sex in terms of familiarity with soccer. Therefore, no participants were filtered out due to preconceived loyalties toward the soccer teams from either country.

Given the 3 (between subjects priming condition: American, Neutral, and Chinese) × 2 (within subjects repeated measures: Cheering for US team, and Cheering for Chinese team), we conducted a mixed ANOVA analysis. Analyses conducted while controlling for American and Chinese acculturation, sex, and familiarity with soccer did not change the significance or direction of the results, therefore results from analyses procedures most consistent with established methodology found in literature, controlling for acculturation and familiarity with soccer, are reported. Additionally, given discussions on the common usage of demographics characteristics as a proxy for


*N* = *136. p* < *0.05 for values in bold.*

acculturation (cf. Nguyen and Benet-Martínez, 2007), we did not use country of birth as a control variable, and instead used level of acculturation. Initial results showed a significant Prime × Cheering interaction, F(2, 130) = 4.89, p = 0.01, η 2 p = 0.07, indicating that the cultural priming did have an effect on the cheering. In order to better understand the results, we further examined the simple main effects. Univariate analysis on whether the cultural priming changed the level cheering for each team demonstrated small effects (η 2 <sup>p</sup> <sup>=</sup> 0.04 for both cheering outcomes; Cohen, 1988); however, these results were not statistically significant. Therefore, Hypothesis 1 is not supported. However, multivariate analysis comparing what effects the cultural priming had on the differences between the two cheering measures showed significant differences. Results showed that the cultural priming increased the magnitude of the preference participants displayed toward the Chinese team over the American team. Specifically, while the participants did not express significantly different cheering for either team under the American priming condition, the participants did express significantly lower levels of cheering for the American team (M = 3.13, SD = 0.67) compared to the Chinese team (M = 3.53, SD = 0.86) under the Neutral priming condition, F(1, 130) = 6.10, p = 0.02, η 2 <sup>p</sup> <sup>=</sup> 0.05. This difference was magnified under the Chinese priming condition, F(1, 130) = 48.29, p = 0.00, η 2 <sup>p</sup> <sup>=</sup> 0.28, with the participants expressing significantly lower level of cheering for the American team (M = 2.91, SD = 0.72) compared to the Chinese team (M = 3.75, SD = 0.79). Therefore, Hypothesis 2 is supported. **Table 2** presents the means and standard deviations for the dependent variables under all priming conditions, and **Figure 1** presents the results graphically. Implications of the findings are discussed in the following section.

# DISCUSSION AND IMPLICATIONS

The results in this study partially supported the proposed hypotheses in that the different priming conditions did not influence how the Chinese-American participants in this study rooted for each of the soccer teams, Chinese or U.S., however the different priming conditions did influence the difference by which the Chinese-American participants in this study cheered for the Chinese over the U.S. soccer teams. These findings indicate that the degree to which biculturals express loyalty for each of the respective cultures when they are in direct competition with each other can be externally influenced by


priming. This shows that bicultural loyalties can be malleable to a certain degree.

Most notably, cultural primes did have an effect on the magnitude of the difference by which biculturals express loyalty toward their host vs. the home culture. When their host cultural identity is activated, biculturals do not differ in their expressed loyalty, while the activation of the home identity causes biculturals to express the highest degrees of loyalty toward their home culture compared to the host culture, while under a supposedly culturally neutral situation in which no specific identity is being activated, their expression of home loyalty is slightly reduced. Essentially, biculturals do not feel differential loyalties when their host cultural identity is primed. However, when their host cultural identity is not activated, then their loyalty for the home culture becomes significantly higher.

Returning to societal implications, our findings suggest a Chinese-American living in the United States and exposed to American culture and English in daily life may not feel large disparities in terms of loyalties toward China or the United States, although they would be slightly more loyal to China. However, when Chinese cultural elements are present in the environment and thereby activating the Chinese identity, then the individual would experience heightened loyalties toward China. It is not exactly clear why this is the case. It is possible because our participants largely originated from China, they are innately loyalty to the Chinese team over the U.S. team regardless of the cultural prime. The primary effect cultural primes have on this loyalty is the degree to which they would express loyalty toward one team over another. In contrast, biculturals born in the United States may innately support the U.S. team over the Chinese team. The likely explanation is that the participants, due to exhibiting higher levels of Chinese acculturation than American acculturation, also had higher levels of cultural identification with Chinese culture than American culture even though this identification can still be manipulated by cultural primes. However, due to the limited sample size of the current study we were not able to conduct these post-hoc analyses. Future research may investigate this phenomenon further and also explore whether differences in behavioral acculturation, which were used as control variables in the current study, and cultural identification may have an impact on the expression of loyalty.

There may also be other possible moderators in this phenomenon. First and foremost is bicultural identity integration (BII). Research on BII show that how cultural stimuli activate cultural knowledge and identities in biculturals can be moderated by whether biculturals view their dual cultural identities as being conflicting and distant from each other (Benet-Martínez et al., 2002). A bicultural who feels more conflicted, or Low BII, responds contrastively to cultural stimulus, e.g., an American cultural prime activates the Asian knowledge frame and identity, while a bicultural who feels less conflicted, or High BII, responds congruently to cultural stimulus, e.g., an American cultural prime activates the American knowledge frame and identity. In the context of this study, a Low BII bicultural might cheer more for the Chinese team when exposed to an American prime, while a High BII bicultural might cheer more for the American team when exposed to an American prime.

Another possible moderator is that of in-group threat. Specifically, out-group derogation is a result of threat to the in-group, with individuals derogating lower-status groups when faced with a threat to the in-group, a concept known as downward compensation (Wills, 1981). In the context of this study it is possible that a Chinese-American bicultural would derogate against a team from a culture perceived as being lower-status when the Chinese team is being threatened. However, the current study design, which focuses on cheering for a given team, would need to be adjusted to reflect that individuals could also react negatively toward a competing team. This study design would be particularly interesting given findings that suggest intergroup bias is driven more by love for the in-group rather than derogation toward the outgroup (Halevy et al., 2008, 2012), suggesting that downward compensation is engaged in order to make the in-group easier to love.

Another possible moderator is that of personal or collective discrimination. Specifically, research suggests that individuals who belong to lower status groups exhibit higher degrees of intergroup bias (Leonardelli and Brewer, 2001). Therefore, in the context of the current study, depending on the status differential between the two competing sports teams, there may be a larger differential between cheering for the in-group team than the out-group team.

Although bicultural research has progressed rapidly in recent years, many questions regarding bicultural cognition, identity, and social identity remain to be answered. The results obtained from this study lend one more piece toward the puzzle of bicultural social identity and how biculturals express their loyalty toward different cultural groups, while highlighting several factors that still merit further research.

## REFERENCES


## ETHICS STATEMENT

This study was approved by the institutional review boards of Baruch College/CUNY and SUNY Farmingdale State College and found to be exempt per 45 CFR 46.101(b)(2). Participants were presented with an electronic version of the consent form. Participants read the consent form and indicated through check boxes whether they consented. Those who consented continued with the study, while those who did not consent were thanked for their time and did not continue with the study. Undergraduate students were involved in this study. Students who chose not to participate in the research were allowed to do so at any time during their participation. Students were recruited via a departmental Research Participant Pool. Students in the Research Pool may choose to fulfill their research credit requirements through two methods: participation in research studies, or writing short papers. Participation in research is not a requirement, and withdrawal from research studies in no way harms the student's grades or standing.

# AUTHOR CONTRIBUTIONS

AC contributions to manuscript: Literature review, initial hypothesis development, study design, data collection, data analysis, and writing BM contributions to manuscript: Hypothesis development, study design, data collection, data analysis, writing, and editing.

# ACKNOWLEDGMENTS

The authors would like to thank Louis Lipani for his assistance during the conceptual development and experimental design phases of the manuscript, and to both Louis Lipani and Bowen Shi for lending their voices as commentators for the sports clips.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Chiou and Mercado. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Multiple Identity Approach to Gender: Identification with Women, Identification with Feminists, and Their Interaction

#### Jolien A. van Breen<sup>1</sup> \*, Russell Spears <sup>1</sup> , Toon Kuppens <sup>1</sup> and Soledad de Lemus <sup>2</sup>

<sup>1</sup> Department of Social Psychology, University of Groningen, Groningen, Netherlands, <sup>2</sup> Departamento de Psicología Social, Universidad de Granada, Granada, Spain

Across four studies, we examine multiple identities in the context of gender and propose that women's attitudes toward gender group membership are governed by two largely orthogonal dimensions of gender identity: identification with women and identification with feminists. We argue that identification with women reflects attitudes toward the content society gives to group membership: what does it mean to be a woman in terms of group characteristics, interests and values? Identification with feminists, on the other hand, is a politicized identity dimension reflecting attitudes toward the social position of the group: what does it mean to be a woman in terms of disadvantage, inequality, and relative status? We examine the utility of this multiple identity approach in four studies. Study 1 showed that identification with women reflects attitudes toward group characteristics, such as femininity and self-stereotyping, while identification with feminists reflects attitudes toward the group's social position, such as perceived sexism. The two dimensions are shown to be largely independent, and as such provide support for the multiple identity approach. In Studies 2–4, we examine the utility of this multiple identity approach in predicting qualitative differences in gender attitudes. Results show that specific combinations of identification with women and feminists predicted attitudes toward collective action and gender stereotypes. Higher identification with feminists led to endorsement of radical collective action (Study 2) and critical attitudes toward gender stereotypes (Studies 3–4), especially at lower levels of identification with women. The different combinations of high vs. low identification with women and feminists can be thought of as reflecting four theoretical identity "types." A woman can be (1) strongly identified with neither women nor feminists ("low identifier"), (2) strongly identified with women but less so with feminists ("traditional identifier"), (3) strongly identified with both women and feminists ("dual identifier"), or (4) strongly identified with feminists but less so with women ("distinctive feminist"). In sum, by considering identification with women and identification with feminists as multiple identities we aim to show how the multiple identity approach predicts distinct attitudes to gender issues and offer a new perspective on gender identity.

Keywords: gender, multiple identities, social identity, group membership, identification with women, identification with feminists, femininity, stereotypes

#### Edited by:

Anat Bardi, Royal Holloway, University of London, United Kingdom

#### Reviewed by:

Toni Schmader, University of British Columbia, Canada Tegan Cruwys, The University of Queensland, Australia

> \*Correspondence: Jolien A. van Breen j.a.van.breen@rug.nl

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

Received: 10 November 2016 Accepted: 02 June 2017 Published: 30 June 2017

#### Citation:

van Breen JA, Spears R, Kuppens T and de Lemus S (2017) A Multiple Identity Approach to Gender: Identification with Women, Identification with Feminists, and Their Interaction. Front. Psychol. 8:1019. doi: 10.3389/fpsyg.2017.01019

# INTRODUCTION

Since the 1980s there has been increasing attention to the complexities of gender identity, acknowledging that, like many other social identities, gender has a strong cultural component, and is not a straightforward biological fact (Unger, 1979; Marecek et al., 2004). Here we examine women's attitudes toward gender group membership, and argue that these attitudes are governed by multiple identities: identification with women and identification with feminists. We contrast this multiple identity approach with other notable multicomponent approaches to gender identity and argue that the multiple identity approach is simple, while allowing for some new nuances in gender identity compared to previous models. Importantly, this approach helps us understand why being feminine and feminist are not mutually exclusive.

We do not consider here the personal, social and biological factors that determine an individual's gender identity, but rather study women's attitudes toward the socially shared aspect of gender group membership. What does it mean to be a member of the social category of women? An important aspect of the reasoning we present here is that an individual is not entirely free to construct the meaning of group membership as they please. Instead, the meaning of group membership is constructed at the societal level and to a large extent socially shared (Moscovici, 1988; Crocker, 1999). We are interested in how people respond to the social construction of a group to which they belong. We believe that considering identification with women and identification with feminists as separable components of gender identity can offer interesting new perspectives on attitudes toward gender group membership.

The idea that gender identity is multidimensional is reflected in many different models (Condor, 1986; Cameron and Lalonde, 2001; Egan and Perry, 2001; Becker and Wagner, 2009), and an important question arising from such approaches is how the dimensions combine and interact. Many models (Condor, 1986; Henderson-King and Stewart, 1994; Cameron and Lalonde, 2001) discuss evidence that high identification with women can be combined with different gender ideologies (e.g., traditional, progressive, feminist). However, if the gender dimensions are seen as independent, then this means that it should also be possible for the same (feminist) ideology to be combined with both high and low identification with women. Yet, few models discuss this option. One influential model that has explicitly conceptualized gender identity as composed of two independent dimensions is the Gender Identity Model (GIM, Becker and Wagner, 2009). The GIM aims to explain endorsement of sexism and support for collective action, and distinguishes between (1) identity content, a preference for traditional vs. progressive gender roles, and (2) identity strength, measured as identification with women. That is, though the GIM postulates two independent dimensions, only one of these dimensions is a content dimension (traditional vs. progressive), while the other, identification with women, reflects identity strength. In the current studies we propose that identification with women not only reflects identity strength but also has implications for the content of gender identity. That is, our approach incorporates content for both dimensions.

Specifically, we suggest that the content associated with identification with women centers on group characteristics and attributes: what does it mean to be a woman in terms of one's characteristics, traits, interests and values? For instance, key group attributes may include being warm and caring (Fiske et al., 2002; Chen et al., 2004). Although identity content is likely to be socially shared to some degree, individuals can differ in the extent to which they accept or internalize society's view of the group, which is reflected in their degree of identification (Ellemers et al., 2002). There is evidence that those who identify strongly with their group are more likely to self-stereotype, and consider themselves more typical of the group (Turner et al., 1987; Spears et al., 1997, 2001; Leach et al., 2008). Chen et al. (2004) showed that, when asked to list 5 traits that are most typical of women as a group, those who were strongly identified with women listed the same traits as those who were less identified with women, providing evidence that this perception was socially shared. However, those who were strongly identified with women were more likely to say that (positive) traits that defined the group also defined themselves (Chen et al., 2004), than those were less committed to women as a group. Based on these previous findings, we suggest that the content associated with identification with women is socially constructed around group characteristics. Those who are highly identified with women place high importance on traits and characteristics that society considers gender-typical, which we expect to translate to increased tendencies to self-stereotype, and increased perceptions of femininity, compared to those who are less strongly identified with women.

Alongside the characteristics associated with the group, the meaning of group membership also includes the place of the group within the larger social system (Livingstone et al., 2009). What does it mean to be a woman in terms of relative status, social (in) equality, and disadvantage? We argue that attitudes toward such (politicized) identity content are reflected in identification with feminists. In line with this notion, previous research has shown that identification with feminists is related to increased perceptions of sexism in society (Henderson-King and Stewart, 1994), discontent with current power distributions and the status quo (Reid and Purcell, 2004), and increased involvement in collective action (Liss et al., 2004; Nelson et al., 2008; Yoder et al., 2011). Based on these previous findings, we argue that the content of feminist identification is socially constructed around disapproval of the disadvantaged social position of women as a group. An individual's degree of identification with feminists reflects the importance they place on these issues. Those who are strongly identified with feminists have internalized the values of feminism, reject the gender status quo, and consider women to be disadvantaged in comparison to men.

In sum, we propose that identification with women and identification with feminists reflect attitudes toward different components of the social construction of gender. If we think of identification with women as relating to what the group is, then we can think of identification with feminists as relating to how the group is doing in relation to other groups. The level of identification with each of these identities reflects the extent to which a person has accepted and internalized the content associated with that identity. In line with the notion that identification with women and identification with feminists are separable components of gender identity, previous research has found that the correlation between them is very small (Roy et al., 2007).

One benefit of a model in which identification with women and identification with feminists are largely independent, but also associated with specific identity content, is that different combinations of the identities allow for additional nuances in gender identity content. For instance, this perspective allows for high identification with women, without assuming that this will necessarily lead to politicization. Relatedly, those who identify strongly with feminists may differ in their identification with women, which we expect to translate (inter alia) to differences in the importance they place on "femininity." Thus, our multiple identities approach explicitly allows for the possibility that femininity (related to identification with women) could co-exist with feminist identification. Such a distinction in the role of femininity is supported by the gender literature and in the feminist movement: Some branches of feminism emphasize femininity as a domain of positive distinction from men (e.g., feminism of difference, Gilligan, 1977), while others downplay femininity (Butler, 2002). Thus, in this approach femininity and identification with feminists are not necessarily mutually exclusive.

A further consequence of considering identification with women and feminists as distinct, is that they can have conflicting or opposing effects on attitudes toward certain gender issues, such as when an issue relates to attitudes toward group characteristics and group relations. For instance, radical collective action aims to improve the social position of women, and should therefore be positively related to identification with feminists. However, radical collective action may also be negatively related to identification with women, to the extent that radical action is considered gender-atypical behavior for women (Eagly and Steffen, 1986; Hercus, 1999). Additionally, identification with women and identification with feminists may interact in predicting support for radical collective action, so that the positive relationship between identification with feminists and support for radical collective action is stronger amongst women who are less highly committed to typically feminine characteristics (lower identification with women). Likewise, stereotypes are often used to legitimize the intergroup inequality (Jost and Kay, 2005; Rudman and Glick, 2008), and as such endorsement of gender stereotypes is likely to be negatively related to identification with feminists. At the same time, however, gender stereotypes reflect information on what is considered "gender-typical" behavior and can provide differentiation from outgroups (Brewer, 1991; Mlicki and Ellemers, 1996). Thus, when identification with women is low, low attachment to femininity and reduced tendencies to self-stereotype may strengthen the effect of identification with feminists on their disapproval of stereotypes.

In sum, in the current paper we propose a multiple identities approach to gender. Importantly, this approach allows both identification with women, and identification with feminists to reflect content, while keeping a simple 2-factor structure. In Study 1, we examine the hypothesis that identification with women and identification with feminists represent separable dimensions of gender identity. We expect that identification with women predicts attitudes toward group characteristics (e.g., femininity) and identification with feminists predicts attitudes toward the social position of the group (e.g., gender inequality). In Studies 2–4, we examine the utility of this multiple identities approach in predicting differences in gender attitudes. Specifically, we expect that identification with women and identification with feminists interact in predicting support for collective action and perceptions of gender stereotypes. All studies reported here were approved by the relevant ethical committees, and conducted in accordance with the Helsinki declaration.

# STUDY 1

In the first study we examine the central predictions of the multiple identities approach. This study uses a correlational design, to examine the hypothesis that identification with women and identification with feminists will be relatively independent (i.e., not, or only weakly correlated). Secondly, we expect that identification with feminists will predict views on social relations, such as gender equality, and identification with women will predict views on group characteristics, such as perceived femininity.

# Method

### Participants

Ninety-one female students from the University of Groningen participated in exchange for course credit. The mean age was 20.8 years, ranging from 18 to 48. The majority of participants were German (53%) or Dutch (33%). The remaining 14% indicated another nationality, with 4% indicating non-Western nationalities. Given a multiple regression model with the two identification variables entered as predictors, this sample can detect small-to-medium effect sizes (R 2 change <sup>≈</sup> 0.09) with a power of 1 − β = 0.80 when α = 0.05 (G∗Power, see Faul et al., 2007).

#### Independent Variables

#### **Identification with women**

Identification with women as a group was measured by 4 items (α = 0.77) adapted from Doosje et al. (1995; also see de Lemus et al., 2015). I identify with this group; I have strong ties with this group; This group is an important part of my self-image; Being a member of this group is an important part of how I see myself. These items are easily cast in terms of feminism, allowing us to measure identification with women and identification with feminists with the same items.

#### **Identification with feminists**

Identification with feminists was measured using the same scale as for identification with women, substituting the word "women" for "feminists" (4 items, α = 0.94).

#### Dependent Variables

Each of the measures included in this study used 7-point Likert scales, ranging from "not at all" to "very much", with the exception of the self-identification measure, which was categorical.

#### **Attitudes to group characteristics**

Leach identification scale. We included the Leach et al. identification scale (Leach et al., 2008, α = 0.87). This scale is composed of five subscales: centrality of group membership, satisfaction with group membership (4 items, e.g., "I am pleased that I am a woman"), solidarity with the group (3 items, e.g., "I feel solidarity with women"), perceived homogeneity of the group (2 items, e.g., "women have a lot in common with each other"), and self-stereotyping (2 items, e.g., "I am similar to the average woman"). Some items of the centrality subscale were also present in the measure of identification with women. Those items were not repeated, and therefore the centrality subscale is not analyzed separately.

Perceived femininity. Two items measured perceived femininity of the self: "I am a feminine woman" and "I enjoy doing things that are considered typically feminine" (Leaper and Van, 2008, α = 0.66).

#### **Attitudes to group position**

Perceived disadvantage. Three items (α = 0.65, adapted from Cameron and Lalonde, 2001) were used to create a "perceived disadvantage" scale. These items were "I believe that women are disadvantaged compared to men in today's society," "If we do nothing, women will continue to be disadvantaged compared to men" and "I have experienced sexism in my daily life."

Ambivalent sexism scale. The ambivalent sexism scale (Glick and Fiske, 1996) consists of the subscales hostile sexism (11 items, α = 0.92), and benevolent sexism (11 items, α = 0.89). The scale includes items such as "Women should be cherished and protected by men" (benevolent), and "Most women interpret innocent remarks as sexist" (hostile).

Modern sexism scale. The extent to which people perceive sexism in society was measured by the modern sexism scale (Swim et al., 1995) consisting of 8 items (α = 0.82). The scale includes items such as "Society has reached the point where women and men have equal opportunities for achievement."

Attitudes to the feminist movement. The Attitudes to the Feminist movement Scale (Fassinger, 1994) assesses attitudes toward feminism with items such as "Feminist principles should be adopted everywhere." The scale consists of 10 items (α = 0.74).

#### **Self-identification**

The final question asked participants to self-identify as a nontraditional woman, a traditional woman, a feminist or "I don't know" (Gurin and Markus, 1989; Cameron and Lalonde, 2001). This measure was included to distinguish issues related to labeling as a feminist, from issues related to the content of attitudes (Zucker and Bay-Cheng, 2010).

#### Procedure

This study was conducted using Qualtrics. At the start of the questionnaire, participants provided written informed consent and reported demographic information (including gender). Scales were presented in the order described above, items within scales were randomized. It took participants an average of 20 min to complete the study. At the end of the study, participants read a debriefing, and were thanked for their participation.

#### Analytical Strategy

Using multiple regression analyses in which identification with women and identification with feminists are simultaneously entered as mean-centered predictors, we examine the hypothesis that identification with women predicts attitudes toward group characteristics, and identification with feminists predicts attitudes toward the social position of the group.

# RESULTS

# Identification with Women and Feminists

On average women identified strongly with women (M = 5.71, SD = 0.74; 7-point scale), while identification with feminism was substantially lower (M = 3.33, SD = 1.38; 7-point scale). The correlation between identification with women and identification with feminists was small (r = 0.18, p = 0.101), indicating that these are relatively distinct constructs. Given this finding, we examine the content associated with these identities in more detail. Specifically, we hypothesized that identification with women predicts attitudes toward group characteristics (e.g., femininity) and identification with feminists predicts attitudes toward the group's social position (e.g., sexism). The correlations between the different variables are shown in **Table 1**.

#### Hypothesis Test

#### **Attitudes toward group characteristics**

In line with our hypothesis, attitudes related to group characteristics were predicted by identification with women, but not identification with feminists (see **Table 2**). Specifically, higher identification with women was associated with higher self-rated femininity [B = 0.24, SE = 0.03, t(88) = 7.42, p < 0.001, R 2 change = 0.40]. Those who were more strongly identified with women were also more likely to self-stereotype [B = 0.21, SE = 0.04, t(88) = 4.77, p < 0.001, R 2 change <sup>=</sup> 0.21] and more satisfied with being a group member [B = 0.62, SE = 0.11, t(88) = 5.79, p < 0.001, R 2 change <sup>=</sup> 0.28]<sup>1</sup> .

#### **Attitudes toward the group's social position**

Further, as hypothesized, attitudes related to the group's social position were predicted by identification with feminists, but not identification with women (see **Table 2**). Specifically, higher identification with feminists was related to increased perceptions of modern sexism [B = 0.39, SE = 0.13, t(88) = 2.99, p = 0.004, R 2 change <sup>=</sup> 0.10]. Likewise, high identification with feminists was associated with higher perceptions of disadvantage for women

<sup>1</sup>Adding the interaction term of identification with women and feminists did not substantially change these results.


TABLE1|CorrelationtableofStudy


TABLE 2 | Attitudes predicted by identification with women, and identification with feminists in Study 1.

[B = 0.15, SE = 0.03, t(88) = 4.38, p = 0.001, R 2 change <sup>=</sup> 0.19] and endorsed less hostile sexism [B = −0.05, SE = 0.02, t(88) = −2.20, p = 0.031, R 2 change <sup>=</sup> 0.05]. Note, however, that the effect of identification with feminists on hostile sexism is quite small given the size of the sample used in this study. Finally, as would be expected, identification with feminists predicted more positive attitudes to the feminist movement, B = 0.05, SE = 0.01, t(88) = 5.05, p < 0.001, R 2 change <sup>=</sup> 0.23<sup>2</sup> .

Solidarity with the group was predicted by both identification with feminists [B = 0.05, SE = 0.01, t(88) = 3.46, p = 0.001, R 2 change <sup>=</sup> 0.08] and identification with women [<sup>B</sup> <sup>=</sup> 0.18, SE = 0.02, t(88) = 6.64, p < 0.001, R 2 change <sup>=</sup> 0.30], such that solidarity with women as a group was highest amongst those who identified strongly with both women and feminists. Endorsement of benevolent sexism, and perceived homogeneity of the group were not affected by either identification with women or feminists (ts < 1).

#### Additional Measures

The measure of self-report identification showed that 49% of the participants identified themselves as non-traditional women, 18% indicated that they thought of themselves as traditional women, only a very small percentage (5%) identified as feminists, and 28% indicated that they did not know. Thus, more than a quarter of women could not or would not classify themselves. Although the percentage of women explicitly identifying as feminists was very small (5%), identification with feminists distinguished those who self-labeled as feminists from those who did not [χ 2 (3) = 14.36, p = 0.002]. Importantly, the different self-identification categories could not reliably predict attitudes toward gender issues (femininity, satisfaction, modern sexism, and disadvantage) (Wald's Z < 1.37, ps > 0.241). That is, correspondence between categorical self-identification and attitudes toward gender issues is limited, confirming the discrepancy noted by previous work (Zucker and Bay-Cheng, 2010).

## Discussion

In this study, identification with feminists and identification with women showed only a small correlation (consistent with Roy et al., 2007). Moreover, there was evidence that identification with women correlates with attitudes toward group characteristics, and identification with feminists correlates with attitudes to the group's social position. These findings support predictions of the multiple identities approach which permits content for both identities. The difference between the identities is the type of content they incorporate.

Results of this study confirmed the relative independence of the two identities, suggesting that identification as a feminist can exist alongside a sense of personal femininity, a pattern reflected in high identification with women and feminists. These women

<sup>2</sup>Adding the interaction term of identification with women and feminists did not substantially change these results.

also showed the highest solidarity with the broader group of women. At first sight, the combination between satisfaction with group membership associated with identification with women, and perceptions of disadvantage associated with identification with feminists, may seem contradictory. However, these concerns may be reconciled by a desire to accord more status and value to typically feminine attributes, tasks and interests: maintaining a focus on femininity, while at the same time resolving disadvantage. In fact, it could be argued that if feminism implies defending the notion that femininity is not inferior to masculinity, then feminism does not undermine femininity, but rather affirms it.

It is worth noting that in this study, only a very small number of women (4%) self-labeled as feminists. This finding is in line with findings of previous research showing that women are reluctant to self-identify as feminists, even though they may hold feminist attitudes (Aronson, 2003; Zucker and Bay-Cheng, 2010). Such under-use of one category means that only the remaining three categories are used to self-categorize, which limits the variance of such categorical measures and suggests that it is preferable to measure identification with women and feminists as continuous variables.

Further, it is noteworthy that in this study, the mean of identification with women is above the mid-point of the scale, which means that "low identification with women" in this study is relative, rather than absolute. Indeed, there may be many different issues that affect the absolute mean levels of identification in a certain sample. For instance, making salient inter-group competition can increase levels of in-group identification reported (David and Turner, 1999). More specific to the gender context, however, the finding that the mean of identification with women is above the mid-point of the scale might be explained, in part, by the fact that belonging to the category of women is not purely chosen, but "ascribed" by others (based, largely, on biological indicators). In other words, in the case of identification with women, even a woman who is dissatisfied with her group membership and does not consider herself typical of the group, likely still considers herself a woman. Identification with feminists, on the other hand, is more similar to a chosen identity or an opinionbased group (Bliuc et al., 2007; McGarty et al., 2009). Such considerations might inspire a "baseline" level of identification with women, as reflected in somewhat higher overall means.

Taken together, results of Study 1 suggested a relatively clearcut division of attitudes as either relating to group characteristics or the group's social position. However, many gender issues are more complex than this, and have implications for group characteristics as well as the group's social position. In such a case, we may expect both identification with women and identification with feminists to play a role in determining attitudes to such an issue, through additive or interactive effects. Studies 2–4 further explore the utility of the multiple identities approach in predicting attitudes to gender issues that may relate to concern for group characteristics as well as concern for the group's social position.

# STUDY 2

In Study 2, we examine the utility of the multiple identity approach in predicting attitudes to gender issues that have a bearing both on concern for group characteristics and the group's social position, focusing specifically on collective action. Collective action is aimed at confronting disadvantage and producing social change (Van Zomeren and Iyer, 2009), and in the current study we distinguish between radical and moderate forms of collective action (Tausch et al., 2011).

In the context of gender, it has been shown that identification with feminists has a positive relationship with collective action (Liss et al., 2004; Nelson et al., 2008; Yoder et al., 2011). Those who identify strongly with feminism perceive that women are disadvantaged in society, and as such they wish to change the status quo. When considering identification with women, there is reason to expect that it will not have a strong relationship with collective action, as collective action does not relate directly to group characteristics. Thus, we expect that high identification with women does not necessarily lead to increased support for collective action (Henderson-King and Stewart, 1994). However, in the case of radical collective action, we may expect that identification with women will have a negative effect on support for this type of action. Radical collective action is often defined as collective actions that involve some degree of aggression, anger, or even violence (Tausch et al., 2011), traits that are oppositional to social definitions of femininity (Eagly and Steffen, 1986; Hercus, 1999; Fiske et al., 2002). Based on this line of reasoning, we might also expect an interaction between identification with women and identification with feminists when considering radical collective action. Only women who are strongly identified with feminists are likely to consider radical action to improve the social position of women, meaning that support for radical action is low when identification with feminists is low, irrespective of identification with women. However, the motivating influence of identifying as a feminist for radical action will only lead to actual support amongst those who are relatively unconcerned about radical action being atypical and uncharacteristic for the group (i.e., when identification with women is low). That is, higher identification with women might dampen the effect of strong identification with feminists on support for radical action. If this is the case, we would expect support for radical action only amongst those women who identify with feminists but not women. Study 2 examines this possibility.

In sum, using a correlational design, this study examines the hypothesis that support for collective action is affected by both identification with women, and identification with feminists. We expect that support for (radical and moderate) collective action is stronger amongst those who are more highly identified with feminists, but less so with women.

# Method

#### Participants

One hundred and twenty one female participants were recruited amongst students of the University of Granada, Spain. Age ranged from 18 years old to 50 years old, with an average of 19.75. Participants took part in exchange for course credit. Given a multiple regression model with the two identification variables and their interaction entered as predictors, this sample can detect small-to-medium effect sizes (R 2 change <sup>≈</sup> 0.065) with a power of 1 − β = 0.80 when α = 0.05 (G∗Power, see Faul et al., 2007).

#### Design

Data for this study were collected as part of a larger experiment (de Lemus et al., in preparation) with a 2 × 2 betweenparticipants design. Identification with women and identification with feminists were measured alongside the manipulated factors, and the effect of the identification variables on support for collective action is the focus of the current study. As such, this study uses a correlational design.

#### **Manipulation**

Data for this study were collected as part of a larger experiment which included a 2 × 2 between-participants manipulation. The first manipulated factor exposed participants to either stereotypical or counter-stereotypical gender roles through pictures showing men and women in kitchen or office settings. Stimuli for this exposure phase were pictures of men and women, appearing in three different contexts (kitchen, office, and a neutral outdoor setting). The same 6 persons (3 women and 3 men), with an emotionally neutral face, appeared in the different contexts. Participants in the stereotype condition were presented with 90% of the women appearing in a kitchen, and 90% of the men appearing in an office; whereas those in the counterstereotype condition were presented with 90% of the men appearing in a kitchen, and 90% of the women appearing in an office (counter-stereotypical exposure group). Interspersed with the (counter-) stereotypical pictures were neutral trials (N = 16) in which men and women appeared outdoors. Participants were presented with 160 trials in total during the exposure phase.

After the exposure phase participants completed an evaluative decision task. Participants were required to classify target words as either positive or negative. Each target word was preceded by a picture prime. The primes used were the second manipulated factor: Half of the participants saw stereotypical gender roles as primes, whereas the other half of the participants completed the task with male and female faces as primes. That is, in both cases the primes conveyed gender information, but for half the participants the primes also invoked gender role information. The evaluative decision task consisted of 4 blocks of 64 trials.

These two manipulated factors created 4 experimental between-participants conditions: stereotype exposure and faces primes, stereotype exposure and role primes, counter-stereotype exposure and faces primes, counter-stereotype exposure and role primes. Crucially for the current study, however, the manipulated factors did not affect support for collective action, either on their own, or in interaction with the identification variables, as described in the "analytical strategy" section below.

#### Independent Variables

#### **Identification with women and identification with feminists**

Identification with women and identification with feminists were measured in the same way as in Study 1 (4 items each; α = 0.78 and α = 0.95, respectively).

#### Dependent Variables

A complete list of the dependent variables included can be found in the Supplementary Materials. Below we describe only the measures of interest for this study.

#### **Support for collective action**

Support for moderate collective action was measured by 6 items (α = 0.68), focusing on actions like signing a petition, joining a peaceful public demonstration, or lobbying for women's rights. Support for radical collective action was measured with 5 items (α = 0.76), focusing on actions like attacks on sexist institutions, blackmailing, or hacking into e-mail accounts (Tausch et al., 2011). Support for each action was rated on an 11-point scale from not at all to very much. All items referred to the action being taken in order to "reduce gender inequality." Thus, it was clear that the objective of both types of action was the same, only the form differed.

#### **Perceived efficacy**

Perceived efficacy of women as a group was measured with three items (α = 0.82) adapted from van Zomeren et al. (2008). The scale includes such items as "Together, women can achieve their aims." This was used as a control variable in the analyses.

#### Procedure

Participants provided written informed consent, were assigned to one of four conditions, and completed the manipulation. Participants then completed a paper-and-pencil questionnaire, with the measures of central interest, identification with feminists, identification with women and support for collective action at the end. After completing all measures, participants read a funneled debriefing and were thanked for their participation.

#### Analytical Strategy

Because the measures of interest in this study were taken after a manipulation we examined the effect of the manipulated factors on identification with women, identification with feminists and collective action intentions, but no effects were found (Fs < 2.5, p > 0.116). However, identification with women [t(120) = 4.77, p < 0.001], identification with feminists [t(120) = 4.66, p < 0.001] and support for moderate collective action [t(120) = 5.02, p < 0.001] were all related to perceived group efficacy. Therefore, group efficacy is controlled for in the analyses presented below.

Using multiple regression analysis, we examine the hypothesis that both identification with feminists and identification with women affect support for collective action. Specifically with regards to radical collective action, we expect an interaction between identification with women and identification with feminists. Therefore, identification with feminists, identification with women, and their interaction are entered into the regression model as mean-centered predictors.

## Results

#### Identification with Women and Feminists

As in Study 1, participants identified strongly with women as a group (M = 5.82; SD = 0.88), and less with feminists (M = 3.63; SD = 1.58). Again, identification with feminists and identification with women were not significantly correlated (r = 0.12, p = 0.193). The correlations between the different variables are shown in **Table 3**.

#### Hypothesis Test

Moderate and radical collective action were weakly but significantly related (r = 0.22, p = 0.013). Support for moderate action was higher (M = 8.28) than support for radical action (M = 2.46). Support for moderate collective action was predicted by identification with feminists [B = 0.21, SE = 0.08, t(120) = 2.73, p = 0.007, R 2 change <sup>=</sup> 0.05]: those who identified more strongly with feminists were more likely to support moderate collective action. There was no effect of identification with women on support for moderate collective action (t < 1.31). Support for radical collective action was positively predicted by identification with feminists [B = 0.35, SE = 0.09, t(120) = 4.01, p < 0.001, R 2 change = 0.12], while identification with women negatively predicted support for radical action [B = −0.34, SE = 0.15, t(120) = −2.20, p = 0.030, R 2 change <sup>=</sup> 0.035], though this effect was small in size. The interaction between identification with women and identification with feminists did not reach significance (t < 1.24, p > 0.218)<sup>3</sup> . These effects illustrate that support for radical collective action is higher when identification with feminists is high, and identification with women is low. This pattern was the result of additive effects rather than an interaction, and as such provides partial support for our hypothesis.

# Discussion

This study replicates findings from Study 1 that identification with women and identification with feminists constitute separable dimensions of gender identity. Additionally, results from this study show that those who identify more strongly with feminists are more likely to support both moderate and radical collective action strategies aimed at increasing equality between the groups. This is in line with results from Study 1, which suggests that identification with feminists is related to attitudes toward the group's social position (inequality, sexism, relative status). Identification with women on the other hand did not predict support for moderate collective action, and there was some evidence that it negatively predicted support for radical collective action. This indicates that high identification with


<sup>3</sup>These results are not substantially changed when controlling for the manipulated factors.

women does not automatically translate to increased support for collective action, and when collective action seems to contradict social definitions of femininity (radical), higher identification with women is associated with somewhat reduced support for such actions. In addition to these additive effects we also considered the possibility of an interaction between identification with women and identification with feminists, but there was no evidence for this.

In sum, Study 2 shows that support for moderate collective action increases with identification with feminists, but is not related to identification with women. Support for radical collective action is highest amongst those women who identify strongly with feminists but not women, due to additive effects of identification with women and identification with feminists.

# STUDY 3

Study 3 examines another domain expected to relate to both identification with women and identification with feminists: gender stereotypes. Study 1 showed that identification with feminists is related to concern for the societal position of women. As stereotypes are often used to legitimize the gender hierarchy (Jost and Kay, 2005; Rudman and Glick, 2008) they can be seen as unfair and disadvantageous for women. Therefore, it is likely that those who are strongly identified with feminists find gender stereotypes more problematic than those who are less strongly identified with feminists. At the same time, gender stereotypes provide information about which behaviors are considered typical and appropriate for the group (Prentice and Carranza, 2002), and provide a basis for differentiation from out-groups (Spears et al., 1997), in this case, men. Given that Study 1 showed that identification with women is related to attitudes toward group characteristics, it is likely that those who are strongly identified with women find gender stereotypes less problematic than those for whom identification with women is lower. Thus, we might expect additive effects of identification with women and identification with feminists on perceptions of gender stereotypes. However, we might also expect identification with women and identification with feminists to interact. Specifically, we argue that the effect of identification with feminists on critical attitudes toward gender stereotypes will be stronger amongst those for whom group characteristics are less important to their identity (lower identification with women). In other words, because those who are less strongly identified with women attach less importance to typical group characteristics and attributes, identification with feminists more easily leads to criticism of gender stereotypes, because there is no conflict between the two motivations. This line of reasoning suggests that identification with feminists leads to critical attitudes toward stereotypes, particularly for lower levels of identification with women.

This study uses an experimental design to examine the hypothesis that attitudes toward gender stereotypes are predicted by identification with women, identification with feminists, and their interaction. Identification with women and feminists are measured continuously, as in Studies 1 and 2. Attitudes toward gender stereotypes are assessed with a direct self-report measure, as well as an indirect measure. The indirect measure of attitudes toward gender stereotypes exposes participants to a (within-participants) manipulation in which two women express different views of gender stereotypes: one speaker is critical of gender stereotypes, while the other speaker endorses gender stereotypes. The dimension of interest is differences in participants' agreement with one speaker over the other.

# Method

#### Participants

A community sample of 201 female participants was recruited through ProlificAcademic. Of these, 59% were from the United Kingdom, 37% were from the United States, and 4% had other nationalities. Age ranged from 16 years old to 68 years old, with a mean age of 30.6 (SD = 10.758 years). Eight participants were excluded because their completion times exceeded the mean completion time by more than 3 SD, indicating that they had not completed the study in one sitting. Six participants were excluded because they failed the attention check. Three further participants indicated that they had trouble understanding the questions, and were also excluded. The final sample included 184 participants. Given a multiple regression model with the two identification variables and their interaction entered as predictors, this sample can detect small-to-medium effect sizes (R 2 change <sup>≈</sup> 0.043) with a power of 1 − β = 0.80 when α = 0.05 (G∗Power, see Faul et al., 2007).

#### Independent Variables

#### **Identification with women and identification with feminists**

Identification with women and identification with feminists were measured in the same way as the previous studies (α = 0.87 and α = 0.97, respectively).

#### **Manipulation**

We created a within-participants manipulation that presented participants with a conversation between two women. The manipulated factor is the attitudes expressed by each of these women: one speaker criticizes gender stereotypes, the other endorses them. Each speaker made 2 arguments. The antistereotype speaker argues that stereotypes are problematic because they legitimize and exacerbate disadvantage faced by women. The pro-stereotype speaker argues that stereotypes in themselves are not always negative. Thus, we created a withinparticipants manipulation with 2 levels (anti-stereotype vs prostereotype). As a dependent variable we then measured the extent to which our participants agreed with each of the speakers (see details below). In sum, this measure was designed as an indirect measure of participants' views of gender stereotypes.

#### Dependent Variables

#### **Ratings of speakers**

After reading the manipulation, participants rated the speakers on how much they agreed with them, how considerate, friendly and intelligent they found them, and how much they liked them. Ratings on these dimensions were highly correlated (rs > 0.7) and taken together as an indicator of participants' positive attitudes toward the speaker. We expected that ratings of the speakers would be affected by the interaction between identification with women and identification with feminists, such that higher identification with feminists leads to a preference for the antistereotype speaker over the pro-stereotype speaker, and that this relationship becomes stronger for lower levels of identification with women. This measure was analyzed with multiple regression analysis. As our hypotheses focus on preferences for one speaker over the other, we created a difference score reflecting differences in ratings of the speakers by subtracting ratings of the antistereotype speaker from ratings of the pro-stereotype speaker. That is, a positive difference score represents a preference for the pro-stereotype speaker, and a negative difference score represents a preference for the anti-stereotype speaker. The analysis focused on predicting these differences between the ratings of the two speakers from the identification variables, and their interaction<sup>4</sup> .

#### **Perceptions of stereotypes**

As a second, more direct, measure of perceptions of stereotypes, participants saw a list of pre-tested statements reflecting descriptive (N = 10, α = 0.93), and prescriptive stereotypes of women (N = 4, α = 0.91) (Rudman, 1998; Eagly and Karau, 2002). Examples included "women are less aggressive than men" (descriptive), and "women should be more caring than men" (prescriptive). For each of these items, participants rated how problematic they found the statement. Preliminary analyses revealed that participants found prescriptive stereotypes significantly more problematic than descriptive stereotypes [Mdifference = 1.31, t(183) = 13.16, p < 0.001] and therefore descriptive and prescriptive items were analyzed separately. This measure was analyzed with multiple regression analyses in which identification with women, identification with feminists, and their interaction were entered as predictors. We expected that the identification variables will interact, such that women find stereotypes more problematic when they are more strongly identified with feminists, and that this relationship is stronger for lower levels of identification with women.

In addition to these central measures, we included measures of perceived femininity of the self (α = 0.87), perceived disadvantage for women (α = 0.93), Modern Sexism (α = 0.77), hostile sexism (α = 0.94), and benevolent sexism (α = 0.92). These measures were included to replicate findings of Study 1, and they were measured as described above. Finally, some exploratory measures were included, which are described in the Supplementary Materials.

#### Procedure

Data was collected through Qualtrics. Participants accessed the study through the ProlificAcademic website. At the start of the study, participants provided written informed consent, completed demographic information (including gender), as well as the measures of identification with feminists and identification

<sup>4</sup>We use multiple regression analysis here to maintain consistency across measures and across studies. However, we also repeated this analysis using repeated measures ANCOVA. In the RM ANCOVA, the identification variables and their interaction were entered as covariates and the different Speakers as the withinparticipants factor. Results of the RM ANCOVA were equivalent to outcome of the multiple regression analyses (described below), as these two analytical techniques are based on the same procedure (Edwards, 1985; Judd et al., 2011).

with women, and the replication measures. They then read the manipulation text and rated the speakers and arguments, followed by the measure of attitudes toward stereotypes. At the end of the study, participants read a debriefing and were thanked for their participation.

#### Analytical Strategy

We assess the correspondence between findings of this study and those of Study 1 using multiple regression analysis. Predictors are identification with women, and identification with feminists. When evaluating our hypotheses regarding the effects of the manipulation, and perceptions of gender stereotypes we include the interaction between identification with women and identification with feminists in the multiple regression model as a third predictor.

# Results

#### Identification with Women and Feminists

Identification with women was above the mid-point of the scale (M = 4.93, SD = 0.91; 7-point scale), while identification with feminists was below the mid-point of the scale (M = 3.37, SD = 1.53; 7-point scale). The correlation between identification with women and identification with feminists was somewhat higher than in previous studies, r = 0.25, and this correlation was significant (p = 0.001). **Table 4** shows the correlations between the different variables.

#### Correspondence with Study 1

Overall, findings of this study correspond largely to the results of Study 1. Like in Study 1, identification with women predicted attitudes toward group characteristics, and identification with feminists predicted attitudes toward the social position of the group. The statistical information for these findings is presented in **Table 5**. Specifically, as in Study 1, those who were more strongly identified with women saw themselves as more feminine than those who were less strongly identified with women. Moreover, as in Study 1, stronger identification with feminists was associated with higher perceptions of modern sexism in society, and disadvantage for women as a group, as well as reduced endorsement of hostile sexism. Benevolent sexism was predicted by additive effects of identification with women and identification with feminists. Stronger identification with feminists was associated with lower endorsement of benevolent sexism, while stronger identification with women was associated with higher endorsement of benevolent sexism.




TABLE 5 | Attitudes predicted by identification with women, and identification with feminists in Study 3.

#### Hypothesis Test

#### **Effects of the manipulation**

The manipulation exposed participants to an anti-stereotype speaker and a pro-stereotype speaker. Overall, the pro-stereotype speaker was given more positive ratings than the anti-stereotype speaker [Mdifference = 0.61, t(183) = 7.60, p < 0.001]. The preference for the pro-stereotype speaker over the anti-stereotype speaker was particularly strong amongst those who are highly identified with women [B = 0.30, SE = 0.09, t(183) = 3.21, p = 0.002, R 2 change <sup>=</sup> 0.054]<sup>5</sup> . That is, those who are highly identified with women rated the pro-stereotype speaker more positively than the anti-stereotype speaker. However, there was no evidence that ratings of the speakers were affected by the interaction between identification with women and identification with feminists (t < 1), and as such our hypothesis was not supported.

#### **Perceptions of stereotypes**

Participants indicated how problematic they found prescriptive and descriptive stereotypes of women. Results are depicted in **Figure 1**.

For prescriptive stereotypes (see **Figure 1** top), there were additive main effects of identification with feminists and identification with women. Prescriptive stereotypes are perceived as more problematic at higher levels of identification with feminists, B = 0.28, SE = 0.07, t(183) = 4.25, p < 0.001, R 2 change <sup>=</sup> 0.09. Prescriptive stereotypes are perceived as less problematic at higher levels of women's identification [B = −0.28, SE = 0.11, t(183) = −2.56, p = 0.011, R 2 change <sup>=</sup> 0.03].

For descriptive stereotypes, there was an interaction between identification with women and identification with feminists [B = −0.13, SE = 0.06, t(183) = −2.05, p = 0.042, R 2 change <sup>=</sup> 0.02]. Decomposition of the interaction showed that women who are more strongly identified with feminists are more critical of gender stereotypes [B = 0.35, SE = 0.06, t(183) = 5.78, p < 0.001, R 2 change <sup>=</sup> 0.15]. This effect of identification with feminists is stronger when identification with women is low, [B = 0.47, SE = 0.09, t(183) = 5.09, p < 0.001, R 2 change <sup>=</sup> 0.12] than when identification with women is high [B = 0.13, SE = 0.08, t(183) = 1.53, p = 0.128, R 2 change <sup>=</sup> 0.01]. This effect is depicted in the **Figure 1** bottom. An alternative breakdown of the interaction revealed another significant simple slope which showed that, when identification with feminists is high, identification with women has a dampening effect on critical attitudes to gender stereotypes [B = −0.52, SE = 0.14, t(183) = −3.76, p < 0.001, R 2 change <sup>=</sup> 0.06], an effect that is not present when identification with feminists is low (t < 1). In **Figure 1** (bottom panel), this effect is evident from the fact that the lines representing lower vs. higher identification with women diverge more strongly at higher levels of identification with feminists.

These findings support the hypothesis that the interaction between identification with women and identification with feminists predict critical attitudes toward descriptive gender stereotypes. Those who are highly identified with feminists but not women are particularly likely to consider stereotypes problematic. For prescriptive stereotypes, a similar pattern appeared as a result of additive effects.

## Discussion

Study 3 replicated results from Study 1 in a community sample. Identification with women was related to attitudes toward group characteristics, while identification with feminists was related to attitudes regarding the group's position. These findings support the multiple identities approach in showing that identification

<sup>5</sup>Aside from the differences between the speakers, there were main effects of both identification variables on evaluations of the speakers (regardless of the arguments they put forward): higher identification with feminists (B = 0.11, SE = 0.03, t(183) = 3.53, p = 0.001, R 2 change <sup>=</sup> 0.05), and higher identification with women (B = 0.23, SE = 0.05, t(183) = 4.24, p < 0.001, R 2 <sup>c</sup>hange <sup>=</sup> 0.09) led to more positive ratings. Finally, there was an interaction between identification with women and identification with feminists [B = −0.07, SE = 0.03, t(183) = −2.12, p = 0.035, R 2 change <sup>=</sup> 0.02], such that women who were low identifiers on both dimensions gave especially low ratings. As these effects are independent of the arguments given by the speakers, scope for interpretation is limited.

with women and identification with feminists are distinguishable components of gender identity.

Aside from replicating earlier studies, this study also showed some novel findings. Specifically, in line with our hypothesis, results showed that women find gender stereotypes more problematic at higher levels of identification with feminists, and lower levels of identification with women. This pattern appeared as a result of additive effects for prescriptive stereotypes, and as an interaction for descriptive stereotypes. These findings illustrate that, in line with the multiple identities reasoning, different combinations of the two identities can lead to distinct gender attitudes. Given these findings, it is worth noting that the fact that some women do not object to (some forms of) stereotypes need not mean that they accept the lower status implications associated with gender stereotypes. They may support the notion of "typically female" activities and interests, but still object to the idea that these imply lower status. For instance, they may argue that typically female traits such as warmth should be valued more.

Interestingly, several other recent studies have found evidence for interactive effects of identification with women and feminists on gender attitudes (e.g., in this issue Leicht et al., 2017; van Breen et al., in preparation). One way of understanding such interactions, and conceptualizing how the two identities may be combined, is by thinking of the different combinations as theoretical "identity types" or identity profiles. In such a taxonomy, the first group includes women who are not strongly identified with either women or feminists (and as such are relatively "low identifiers"). Secondly, there are those who identify strongly with women but not feminists ("traditional women"; see Condor, 1986). In addition, there are two feminist profiles: those who are highly identified with feminists and women (whom we might call "dual identifiers", see in this issue Leicht et al., 2017), and those who identify strongly with feminists, but less strongly with women (whom we refer to here as "distinctive feminists"). These different identification "profiles" are not necessarily fixed or absolute categories, but rather should be seen as a way of conceptualizing different approaches to integrating the identities. We return to this conceptualization in more detail in the General Discussion.

Although the interaction was in line with our hypotheses and findings from other studies, the small size of the effect means that more power is needed to assess its reliability. Therefore, Study 4 will aim to replicate the interaction. One additional limitation of the current study is that the measure of perceptions of stereotypes asked only about how problematic participants found the statements. Participants might find certain statements problematic for different reasons. For instance, they may find stereotypes problematic because they are untrue, but they may also consider stereotypes problematic precisely because they are true. Additionally, it is worth noting that the manipulation of perceptions of stereotypes did not produce the expected effects. Study 4 examines these issues in more detail.

# STUDY 4

This study aimed, firstly, to replicate the findings of Study 3, and to refine the measure of perceptions of stereotypes. While Study 3 asked only how problematic participants found the statement presented, the current study also asked how true participants found the statements. In all other respects, the design and measures of Study 4 were identical to those of Study 3. Based on the results of Study 3, we expect that women will find gender stereotypes more problematic at higher levels of identification with feminists, and lower levels of identification with women—as a result of interactive effects in the case of descriptive stereotypes, and as a result of additive effects in the case of prescriptive stereotypes.

Like Study 3, this study uses an experimental design to examine the hypothesis that attitudes toward gender stereotypes are predicted by the interaction between identification with women and identification with feminists. The experimental factor consists of a within-participants manipulation that exposes participants to a scenario in which two women discuss different views to gender stereotypes. Identification with women and feminists are measured on a continuous scale.

# Method

#### Participants

Participants were 200 female students at the University of Groningen. Age ranged from 17 years old to 31 years old, with a mean age of 19.7 (SD = 2.08). Three participants were excluded because they failed the attention checks. One participant had to be excluded because she completed the study twice. The final sample included 196 participants. Given a multiple regression model with the two identification variables and their interaction entered as predictors, this sample can detect small-to-medium effect sizes in the range of those found in Study 3 (R 2 change <sup>≈</sup> 0.04) with a power of 1 − β = 0.80, when α = 0.05 (G∗Power, see Faul et al., 2007).

#### Independent Variables

The independent variables were the same as in Study 3: identification with women, identification with feminists and the within-participants manipulation.

#### Dependent Variables

The dependent variables in this study were largely the same as in Study 3. Only those measures that were added or adapted are described below.

#### **Perceptions of stereotypes**

As in Study 3, we examined women's attitudes toward stereotypes indirectly through participants' evaluation of the conversation between the pro-stereotype and anti-stereotype speaker.

Moreover, as a more direct measure of attitudes toward gender stereotypes, we asked participants to indicate how problematic they found a set of descriptive and prescriptive stereotypical statements. We also added some new questions, asking participants how true they found the each of the descriptive (α = 0.86) and prescriptive (α = 0.80) stereotypes.

#### **Exploratory items**

We included five exploratory items to examine how participants perceive women who behave stereotypically. Examples include "women who use their femininity to get ahead are only putting themselves down in the long run (reverse coded)," and "women who use their femininity to get by are only making the best of difficult circumstances" (α = 0.62). We also included 2 items examining women's views on gender differentiation. Items were "the fact that women are different from men should be a point of pride," and "women should try to disprove the idea that women are different from men" (reverse coded) (α = 0.52). Results for these measure are described in the Supplementary Materials.

#### Procedure

Data was collected through Qualtrics <sup>R</sup> . Participants accessed the study through the University of Groningen website. Participants first provided written informed consent and subsequently completed the measures in the same order as in Study 3. The new measure of gender differentiation and the exploratory items were completed at the end of the study. After completing all tasks, participants read a debriefing and were thanked.

#### Analytical Strategy

As in Study 3, we assess the correspondence between findings of this study and those of Study 1 using multiple regression analysis, with identification with women and identification with feminists entered as predictors. When evaluating our hypotheses regarding the effects of the manipulation, we include the interaction between identification with women and identification with feminists as a third predictor in the multiple regression model. For the measure of "perceptions of stereotypes," we also include the interaction term, and control for the perceived truth of the stereotype.

## Results

#### Identification with Women and Feminism

On average women identified strongly with their gender in-group (M = 5.41, SD = 1.04; 7-point scale), while identification with feminism was substantially lower (M = 3.35, SD = 1.51; 7-point scale). The correlation between identification with women and feminism was similar to that in Study 3 at r = 0.27 (p < 0.001). **Table 6** shows the correlations between the different variables.

#### TABLE 6 | Correlation table for Study 4.


#### Correspondence to Previous Studies

As in previous studies, identification with women predicted attitudes toward group characteristics, and identification with feminists predicted attitudes toward the social position of the group. The statistical information for these findings is presented in **Table 7**. Higher identification with women led to higher perceptions of femininity. Those who identified strongly with feminism perceived more modern sexism in society, endorsed less hostile sexism, and less benevolent sexism.

#### Hypothesis Test

#### **Effects of the manipulation**

As in Study 3, the manipulation of a conversation between a pro-stereotype and anti-stereotype speaker produced few theoretically interesting effects. Higher identification with women [B = 0.13, SE = 0.04, t(195) = 3.22, p = 0.001, R 2 change = 0.05] and feminists [B = 0.09, SE = 0.03, t(195) = 3.40, p < 0.001, R 2 change <sup>=</sup> 0.05] led to more positive ratings being given, regardless of the arguments put forward by the speakers. Unlike in Study 3, identification with women did not predict preference for either speaker (and neither did identification with feminists or the interaction, all ts < 1). In sum, our hypotheses regarding the manipulation were not supported.

#### **Perceptions of stereotypes**

For prescriptive stereotypes, those who perceived the stereotypes as more true also perceived them as less problematic [B = −0.67, SE = 0.09, t(195) = −7.95, p < 0.001, R 2 change <sup>=</sup> 0.24]. Moreover, women who were strongly identified with feminists found prescriptive stereotypes more problematic than those who were not so strongly identified with feminists [B = 0.10, SE = 0.05, t(195) = 2.09, p = 0.038, R 2 change <sup>=</sup> 0.02]. Unlike in Study 3, there were no effects of identification with feminists on attitudes toward prescriptive stereotypes [t(195) = −1.02, p = 0.309]. These findings are depicted in the top panel of **Figure 2**.

Descriptive stereotypes, too, were rated as less problematic by those who perceived them as more true [B = −0.32, SE = 0.08, t(195) = −4.02, p < 0.001, R 2 change <sup>=</sup> 0.07]. As in Study 3, the interaction between identification with women and identification with feminists predicted how problematic people found these stereotypes [B = −0.12, SE = 0.04, t(195) = −2.60, p = 0.010, R 2 change <sup>=</sup> 0.031]. Decomposition of the interaction showed that, TABLE 7 | Attitudes predicted by identification with women, and identification with feminists in Study 4.


at lower levels of identification with women, identification with feminists is an increasingly strong predictor of critical attitudes toward stereotypes. Put differently, the effect of identification with feminists is stronger when identification with women is lower [B = 0.23, SE = 0.08, t(195) = 3.03, p = 0.003, R 2 change = 0.042], as compared to higher (t < 1). This effect is depicted in the bottom panel of **Figure 2**. The alternative breakdown of the interaction revealed another significant simple slope: identification with women has a dampening effect on critical attitudes to gender stereotypes when identification with feminists is high [B = −0.33, SE = 0.11, t(195) = −3.02, p < 0.003, R 2 change <sup>=</sup> 0.04], but not when identification with feminists is low (t < 1). In **Figure 2** bottom, this effect is evident from the fact that the lines representing lower vs. higher identification with women diverge more strongly at higher levels of identification with feminists.

In sum, results for the descriptive stereotypes confirmed our hypothesis, and replicated the interaction demonstrated in Study 3. Importantly, these patterns appear while controlling for the perceived truth of the stereotype.

# Discussion

Overall, the results of Study 4 correspond largely to those of Studies 1–3. As in Studies 1 and 2, identification with feminists reflected attitudes toward group relations, while identification with women reflected attitudes toward group characteristics. Importantly, the interaction from Study 3 was replicated, which showed that women who are more strongly identified with feminists are more critical of gender stereotypes, and this effect of identification with feminists is stronger when identification with women is lower. These findings regarding the interaction are in line with other recent work from our lab, which has examined responses to implicit gender stereotypes (van Breen et al., in preparation). Results in that line of studies show that a combination of stronger identification with feminists and lower identification with women as a group leads women to resist exposure to implicit gender stereotypes, for instance through persistence in counter-stereotypical performance domains. However, as in Study 3, the interaction effect was small. Therefore, in section Pooled Analysis, we further explore the reliability of the pattern. Study 4 also revealed some novel findings: perceptions of the problematic nature of gender stereotypes could not be explained by differences in the perceived truth of gender stereotypes.

Although those who are highly identified with feminists, but not with women ("distinctive feminists") are most conspicuous in the results, theoretically speaking women who are highly identified on both dimensions (the "dual identifiers") are also interesting. These women are feminists, but do not necessarily object to descriptive gender stereotypes. This finding may be due to the fact that stereotypes can provide differentiation from the out-group (i.e., men, see Brewer, 1991; Mlicki and Ellemers, 1996) which might be a desirable feature for those who are strongly identified with the group.

As in Study 3, the manipulation did not produce the expected effects in this study. Overall, participants agreed more with the arguments of the pro-stereotype speaker than the arguments of the anti-stereotype speaker. It may be the case that the antistereotype speaker was perceived as "too radical." The antistereotype arguments were phrased quite prohibitively, such as "women should not behave stereotypically, as it reinforces the disadvantage women face." Participants may have disliked this, and therefore favored the pro-stereotype speaker. An additional limitation of the manipulation was that both speakers expressed disapproval of women's low status position, and as such both speakers could be said to be feminists. Indeed, there is some evidence that lower identification with feminists was associated with lower agreement with the speakers overall (see Supplementary Materials). The disapproval of the low status position of women was kept constant, rather than varied, because the measure was designed to focus on perceptions of stereotypes as harmful or not. If we had also varied speakers' views on women's disadvantage, the conversation would have become very complex. Already there was some evidence that participants found it difficult to remember details of the conversation, and as such we considered it undesirable to further complicate the manipulation.

# Pooled Analysis

Both Study 3 and Study 4 showed evidence that the interaction between identification with women and identification with feminists affects attitudes toward descriptive gender stereotypes. However, given the small size of the effect, we considered it worthwhile to assess this interaction in a post-hoc analysis with more power. As Study 3 and Study 4 had the same design, we can use Integrative Data Analysis (IDA; Curran and Hussong, 2009) to maximize power and evaluate the underlying pattern of the interaction. In this analysis, we pool the data from Studies 3 and 4 to assess whether, as in the individual studies, there is evidence that identification with women and identification with feminists interact to predict perceptions of descriptive gender stereotypes in the larger sample. Given a power of 1 − β = 0.80 and α = 0.05, the pooled sample (N = 380) can detect effect sizes of R 2 change <sup>≈</sup> 0.02 and above.

The pooled sample was analyzed with multiple regression analysis, in which, as before, identification with women, identification with feminists, and their interaction are entered as continuous predictors. Additionally, we added a dummy variable reflecting the Study from which each data point was derived, to control for the influence of the different samples.

Results showed that perceptions of descriptive stereotypes were affected by the interaction between identification with feminists and identification with women [B = −0.12, SE = 0.04, t(379) = −3.18, p = 0.002, R 2 change <sup>=</sup> 0.023]. Breakdown of the interaction showed the same patterns as those described above. Firstly, at lower levels of identification with women, identification with feminists is an increasingly strong predictor of critical attitudes toward stereotypes: the effect of identification with feminists is stronger when identification with women is low [B = 0.36, SE = 0.06, t(379) = 6.11, p < 0.001, R 2 change <sup>=</sup> 0.09], compared to when identification with women is high [B = 0.12, SE = 0.05, t(379) = 2.43, p = 0.016, R 2 change <sup>=</sup> 0.01]. The alternative breakdown of the interaction revealed another significant simple slope: identification with women has a dampening effect on critical attitudes to gender stereotypes when identification with feminists is high [B = −0.46, SE = 0.09, t(379) = −5.09, p < 0.001, R 2 change <sup>=</sup> 0.06], but not when identification with feminists is low (t < 1).

These findings confirm that the interaction found in Studies 3 and 4 is reliable when assessed in the pooled data set. Taken together, the different patterns that make up the interaction show that critical attitudes toward descriptive gender stereotypes are strongest amongst women who are strongly identified with feminists, but less so with the broader group of women.

# GENERAL DISCUSSION

The studies presented here provide insight into how identification with women and feminists predict different attitudes toward gender as a social category. We now review the results of the studies in the light of the multiple identities approach, and evaluate its utility in predicting attitudes toward gender issues.

# The Multiple Identities Approach

The multiple identities approach proposes that attitudes toward gender as a social category are determined by two distinct dimensions of gender identity: identification with women, reflecting attitudes toward the characteristics associated with the group, and identification with feminists, reflecting attitudes toward the social position of the group. This central prediction of the model is confirmed across the four studies reported here, in student samples as well as a community sample. That is, the studies confirm that identification with women and identification with feminists represent distinguishable aspects of gender identity, and as such, that gender identity is not unitary (Condor, 1986; Henderson-King and Stewart, 1994; Becker and Wagner, 2009). Moreover, results show that identification with women is related to attitudes toward group characteristics, such as femininity and self-stereotyping. These "group characteristics" need not be thought of as essentialist traits, but rather as part of a culturally shared understanding of the social category of "women" (Devine, 1989; Rudman and Glick, 2008). Identification with feminists, by contrast, is related to attitudes toward the group's social position, such as support for collective action and perceptions of sexism.

#### Combining Identification with Women and Feminists

If we consider gender identity in the light of the multiple identities approach, this gives rise to the question of how the dimensions may be combined. The multiple identities approach suggests that, when a certain issue has a bearing on both group characteristics and the group's social position, attitudes toward such an issue will be affected by both identification with women and identification with feminists. Indeed, studies 2–4 showed that issues such as support for radical collective action and perceptions of gender stereotypes are affected by both identification with women and identification with feminists, manifested as additive or interactive effects. The finding that particular combinations of identification with women and feminists lead to differences in attitudes toward gender issues is not only in line with the multiple identities approach, but also corresponds to other recent work from our lab (van Breen et al., in preparation), as well as the findings of Leicht et al. (2017, this issue).

The combinations of different gender identities can be thought of in terms of different conceptual groups or "prototypical types" of gender identifiers. In fact, several theorists have found it helpful to discuss the possibility of gender identity "subgroups" to address the question how different aspects of gender identity relate to one another (Condor, 1986; Gurin and Markus, 1989; Becker and Wagner, 2009). In our approach, the first possible combination includes those whose identification with both women and feminists is relatively low ("low identifiers"). Low identifiers navigate gender group membership by giving priority to social identities outside the gender context, as they dislike being viewed in terms of gender (Barreto et al., 2010). Secondly, there are those who identify strongly with women but not feminists ("traditional women"). Traditional women value typically female gender roles (Condor, 1986), but they disavow feminist concerns about the social position of women. There are two feminist subgroups: those who are highly identified with feminists and women ("dual identifiers"; see Leicht et al., 2017 in this issue), and those who are highly identified with feminism, but not women (whom we have called "distinctive feminists"). Dual identifiers can be described as preferring integrative identity management strategies that unite their commitment to women as a group with their commitment to feminism. For instance, they may be willing to take on leadership positions (Leicht et al., 2017; this issue), but prefer more feminine styles when they do so (Olsson and Walker, 2004). Distinctive feminists, on the other hand, navigate gender group membership by giving priority to feminist issues over their identification with women. For instance, they may disavow feminine beauty ideals because they perceive them as contributing to women's objectification (Murnen and Smolak, 2009). It is important to note that even though "distinctive feminists" do not identify highly with women, this does not mean that they are "anti-women" (see Becker et al., 2011; Cichocka et al., 2013). Rather, they disavow the (current) social construction of the group.

Importantly, this taxonomy does not represent fixed or absolute categories, but rather a way of conceptualizing different approaches to integrating the identities. Indeed, we see gender identity as dynamic and context-dependent. Given that the social construction of identity plays a large part in our approach, arguably the most important contextual factor is the nature of the social construction. Different cultures may construct gender differently, and this may in turn affect attitudes to specific gender issues. Additionally, an individual's commitment to the different identities may develop over time, for instance through personal experience. Likewise, research on social influence has shown that making salient an intergroup context can shift individuals' attitudes toward those of more radical minorities within the ingroup (David and Turner, 1999). As we used cross-sectional data we did not examine this dynamic component of multiple identities in the current study, but we believe this is a fruitful area for future research.

In sum, the different combinations of high vs. low identification with women and feminists can be thought of as reflecting different strategies for managing multiple gender identities. Some women prioritize one dimension over the other (traditional women; distinctive feminists) while others seek to integrate them (dual identifiers).

# Advantages of the Multiple Identities Approach

The multiple identities approach has several advantages that are worth highlighting. Firstly, the fact that identification with women and identification with feminists represent separable components of gender identity allows for different kinds of identity content, which is crucial when attempting to model something as diverse as attitudes toward gender group membership. One consequence of this is that identification with feminists and femininity are not mutually exclusive: a woman may embrace both femininity and feminism. As noted above, this issue is also reflected in feminist discourse (Gilligan, 1977; Butler, 2002). A further consequence of the two independent dimensions is that some women are highly identified with women as a group, but do not hold politicized identities. Indeed, our findings on collective action confirm that high identification with women does not automatically increase politicized attitudes (Henderson-King and Stewart, 1994).

As identification with feminists can function independently of identification with women, identification with feminists can also be part of men's gender identity (e.g., Digby, 2013). Preliminary results of applying the multiple identities approach to men's gender identity<sup>6</sup> show that, as amongst women, identification with men as a group correlates with perceived masculinity and self-stereotyping, while stronger identification with feminists increases perceived prevalence of sexism. However, the relationship between the identities is somewhat different amongst men: for men the factors are negatively correlated; those who identified more strongly with men, and felt more masculine, were less likely to identify with feminism (see also Burn et al., 2000; Lemaster et al., 2015). In sum, the possibility of applying the multiple identities approach to men's gender identity allows us to assess how men's attitudes toward gender group membership differs from women's, as well as where similarities lie. Though further work is needed on this front, we consider this a strength of the model.

The distinction between group characteristics and the group's social position may also play an important role in how people think of identities outside the gender context, such as ethnic group membership. For instance, we can think of the multiculturalist approach to ethnic diversity as appreciating group differences while also addressing political disadvantage (Verkuyten and Brug, 2004), suggesting that, as the multiple identities approach argues, both attitudes to group characteristics and perceptions of the group's social position play a role in how social group membership is constructed.

A further methodological strength of this approach is its concise measure of identification, using eight items in total to measure identification with women and identification with feminists. The two identification variables were measured with the same items, apart from the fact that the word "women" was replaced by "feminists." Thus, these gender identity dimensions are shown to be independent, even when the measures are very similar. Therefore, the lack of correlation between identification with women and identification with feminists is a conservative test of the independence of the dimensions.

A limitation of the current study is its correlational nature, preventing inferences about causal direction. For instance, the relationship between identification with feminists and perceived sexism might arise because identification with feminists leads to increased sensitivity to sexism (Major et al., 2003) or, conversely, increased exposure to sexism might lead to increased identification with feminism (Henderson-King and Stewart, 1994). In fact, it is likely that both these processes play a part in identity development. A further limitation is the reliance on student samples in Studies 1, 2, and 4. Students are likely to hold more progressive attitudes than the general population, and therefore it may be more possible for the same person to identify with both women and feminists amongst students than it is in the general population. That is, it is possible that in other populations there would be a negative correlation between the two identities. However, in Study 3, which used a community sample, there was no evidence for such a negative correlation. Findings from Study 3 instead tended toward a positive correlation between identification with women and identification with feminists. Nevertheless, the reliance on student samples is an additional limitation of this study.

# Conclusions

This study develops the multiple identities approach to gender identity, in which identification with women and identification with feminists are orthogonal components of gender identity, which together predict attitudes toward gender group membership. Identification with women predicts attitudes toward group characteristics, such as perceived femininity and self-stereotyping, while identification with feminists predicts attitudes toward the group's social position, such as sexism and disadvantage for women. Different combinations of identification with women and feminists give rise to four conceptual identity profiles: low identifiers, traditional women, distinctive feminists, and dual identifiers. Importantly, the multiple identity approach helps to explain differences in gender attitudes, notably that: (1) Strong identification with feminists does not preclude a sense of being feminine; (2) Strong identification with women as a group does not automatically increase politicized attitudes; and (3) Critical attitudes toward gender stereotypes are most pronounced amongst feminists who are less strongly identified with women. Taken together, findings from these studies suggest that considering identification with women and identification with feminists as multiple identities can provide valuable new insights into attitudes toward gender group membership.

<sup>6</sup>Further details of this work can be obtained from the first author.

## ETHICS STATEMENT

The studies reported in this article conform to APA ethical guidelines, and have been approved by the relevant local ethical committees. The studies conducted at the University of Groningen have been approved by the Ethical Review Board of the University, which bases its approval on guidelines from the professional code of the NIP, and the Personal Data Protection Act. More information can be obtained from http://www.rug.nl/ research/heymans-institute/organization/ecp/ or the first author.

# AUTHOR CONTRIBUTIONS

All authors contributed to study conception and design. Acquisition and analysis of data was conducted by JB and SdL.

# REFERENCES


JB drafted the manuscript, RS, TK, and SdL gave feedback for its revision.

# FUNDING

This work was supported by Grant no. PSI2016- 79971-P from the Spanish Ministry of Science and Technology (AEI/FEDER, UE) awarded to SdL.

## SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fpsyg. 2017.01019/full#supplementary-material


van Breen et al. Multiple Identity Approach to Gender

diffuse forms of system justification. J. Pers. Soc. Psychol. 88, 498–509. doi: 10.1037/0022-3514.88.3.498


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 van Breen, Spears, Kuppens and de Lemus. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# National Identification Counteracts the Sedative Effect of Positive Intergroup Contact on Ethnic Activism

#### Adrienne Pereira\*, Eva G. T. Green and Emilio Paolo Visintin

Laboratory of Social Psychology, Institute of Psychology, Faculty of Social and Political Sciences, University of Lausanne, Lausanne, Switzerland

Positive intergroup contact with socially and economically advantaged national majorities has been shown to reduce ethnic identification among minorities, thereby undermining ethnic minority activism. This finding implies that ethnic identity is the relevant social identity driving ethnic minorities' struggle for equality. We argue that the study of the "sedating" effect of positive intergroup contact for minorities should be more nuanced. The existence of multiple and sometimes interplaying social identities can foster a reinterpretation of the meaning of "ethnic" activism. This study therefore examines how the interplay of ethnic and national identities shapes the sedating effect of contact on minority activism. We expect national identification to buffer the sedated activism resulting from reduced ethnic identification. That is, the mediation from intergroup contact to reduced ethnic activism through weakened ethnic identification is expected to be moderated by national identification. With survey data from Bulgaria, we investigated support for ethnic activism among Bulgarian Roma (N = 320) as a function of their contact with the national majority as well as their degree of ethnic and national identification. The predicted moderated mediation was revealed: a negative indirect relationship between contact and activism through decreased ethnic identification occurred among Roma with low national identification, whereas no sedating effect occurred among Roma identifying strongly as members of the Bulgarian nation. We discuss the meaning of national identification for the Roma minority, who experience harsh discrimination in countries where they have been historically settled, as well as convergence of these findings with work on dual identification. We highlight the role of interacting social identities in mobilizing resources for activism and the importance of adopting a critical view on ethnic discourse when studying activism in both traditional and immigrant minorities.

Keywords: intergroup contact, minority activism, ethnic identification, national identification, Roma

# INTRODUCTION

Minority activism aims at modifying norms or practices established by a majority group (Moscovici, 1976). In the case of ethnic minority activism, unequal treatment and discrimination based on the ethnic categorization of individuals is challenged. Ethnic categories and the related social identities develop when a group of individuals share a common ancestry, physical traits

#### Edited by:

Natasza Kosakowska-Berezecka, University of Gdansk, Poland ´

#### Reviewed by:

Anna Włodarczyk, Universidad Católica del Norte, Chile Manuela Barreto, University of Exeter, UK

> \*Correspondence: Adrienne Pereira agiroud2@gmail.com

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

> Received: 24 October 2016 Accepted: 14 March 2017 Published: 10 April 2017

#### Citation:

Pereira A, Green EGT and Visintin EP (2017) National Identification Counteracts the Sedative Effect of Positive Intergroup Contact on Ethnic Activism. Front. Psychol. 8:477. doi: 10.3389/fpsyg.2017.00477

or values differentiating them from others (Smith, 1991). Most countries are ethnically diverse in this sense, and ethnic identities usually differentiate subordinate minorities from dominant majorities within superordinate nation-states (see Staerklé et al., 2010). The ethnic identities of national minorities (or majorities) are indeed in part relational, stemming from interdependent comparisons, and frequently, unequal treatment.

Notwithstanding the criticism addressed to integration policies that focus on prejudice reduction rather than addressing structural inequalities, recent research has shown that members of ethnic minorities who experience positive contact with members of the advantaged, or dominant, majority display attenuated ethnic activism (e.g., Dixon et al., 2007), partly because of reduced ethnic identification (e.g., Wright and Lubensky, 2009). Building on the literature revealing this "irony of harmony" resulting from positive intergroup contact (Saguy et al., 2009; see however Kauff et al., in press), our goal is to examine whether the sedating effect of positive contact on ethnic activism via reduced ethnic identification is buffered by national identification of minority members. As national identities are central in the contemporary world (Reicher and Hopkins, 2001), we argue that the synergy between ethnic and national identifications in ethnic minorities is at play in intergroup encounters. The contribution of this research is twofold. First, we provide novel insights that speak to the recent integration of two research traditions—work on social identity predictors of activism and work on demobilizing effects of intergroup contact (see also, e.g., Çakal et al., 2011). Second, we examine the perspective of the Roma minority, historically one of the most severely rejected ethnic minorities in Europe (e.g., Heath and Richards, 2016), yet hardly studied in social psychology. We conducted a cross-sectional survey in Bulgaria, a multicultural society composed of ethnic minorities among which Roma are the second largest.

# Positive Intergroup Contact and Minority Activism

There is ample evidence that positive intergroup contact improves intergroup attitudes (Pettigrew and Tropp, 2006). In particular, individuals from advantaged groups who have positive contact experiences with members of disadvantaged social groups show less prejudice and negative emotions as well as greater support for egalitarian policies (e.g., Pettigrew and Tropp, 2011). However, the consequences of positive intergroup contact are different for members of disadvantaged groups. Recent research has revealed that positive contact with members of socially and economically advantaged groups is associated with attenuated support for egalitarian policies. For example, studies conducted in post-apartheid South Africa have shown that quantity of positive contact with Whites was related to Blacks' decreased support to compensatory and preferential policies aiming to ensure racial equality (Dixon et al., 2007; Çakal et al., 2011) and reduced behavioral intentions in favor of the Black minority (such as signing a petition or participating in anti-discrimination projects; Çakal et al., 2011). Similarly, Tropp et al. (2012) revealed that ongoing friendships with Whites were associated with a progressive decline of support for ethnic activism among African- and Latino-American college students. In short, unless the struggle for equality is carried out by the advantaged majority, positive contact with majority members can demobilize minorities, and unintendedly result in the status quo of power relations.

Scholars have explored the psychological processes underlying this demobilization of minority members. For example, Saguy et al. (2009) found a relationship between experiencing positive contact with Israeli Jews and reduced support for social change among Israeli Arabs (i.e., improvement of their position in Israel). The relationship was mediated by improved attitudes toward Jews, by increased perception that Jews treat Arabs fairly, as well as by decreased awareness of structural inequalities (see also Çakal et al., 2011; Tropp et al., 2012 for similar findings showing decreased perceptions of discrimination). Furthermore, using self-reports of past interracial contact, Wright and Lubensky (2009) demonstrated that interactions with Whites before entering university reduced endorsement of ethnic-based collective action both in Africanand Latino-American students. Crucial to the present study, the sedating effect of positive intergroup contact was mediated by reduced ethnic identification. Tausch et al. (2015) found a similar disidentification process among Latino-American students resulting from present interethnic friendships with Whites. In the current study, we thus also expect that positive intergroup contact with the Bulgarian national majority is related to reduced ethnic activism among the Bulgarian Roma minority through diminished ethnic identification.

The rationale for the sedative effect of contact put forward by Wright and Lubensky (2009) derives from the idea that collective action and prejudice reduction are two incompatible routes in disadvantaged groups' struggle for social equality. Indeed, with ethnic identification driving ethnic activism, social protest implies recognizing social disadvantages and motivation to improve the status of the ingroup. Positive intergroup contact, in turn, results in lowered attention to inequalities (Saguy et al., 2009) and weakened salience of group categories (see Brown and Hewstone, 2005 for a discussion on category salience during intergroup contact). Experiencing positive contact with more advantaged individuals makes intergroup boundaries seem more permeable. Hence, the advantaged social identity becomes relevant and the disadvantaged (ethnic) identity less relevant for disadvantaged individuals. Consequently, members of disadvantaged groups reinterpret their social identity as mirroring a common ingroup shared with the advantaged group or as a dual identity with elements of the disadvantaged and advantaged identity (Gaertner and Dovidio, 2000). According to Wright and Lubensky (2009), in such identity configurations the subordinate ethnic identity is no longer strong enough to drive social protest, since it is dominated by or at least predisposed toward the advantaged group. Building on this research, in the current study, we suggest that the motivation to enhance the subordinate group's position can remain in new identity reconfigurations following intergroup contact experiences and thereby allow for activism.

Dynamic reconfigurations of identity are not a new idea. Social identity theory (Tajfel and Turner, 1979) and the literature on social stigma (e.g., Major and O'Brien, 2005) have described a range of social identity adjustments that are crucial for the well-being of stigmatized groups members (Bobowik et al., 2014). Outgroup derogation (e.g., prejudice) is one strategy. Individual upward mobility from a discriminated group to another more privileged one is another (see Tausch et al., 2015). Furthermore, recent studies on negative identity management have proposed that disadvantaged ethnic minority members cope by "navigating" multiple group identities (e.g., Curtin et al., 2016). Multiple identities are described in those studies as resources, not replacing, but rather repositioning and reinterpreting the disadvantaged (ethnic) identity in light of other group memberships. In particular, members of ethnic minorities can experience a psychological overlap between their exclusive (i.e., ethnic) and inclusive (i.e., national) identities. Some studies have found that this particular configuration of their "collective" identity drives support for minority activism (e.g., Çakal et al., 2016; Curtin et al., 2016). Note that such identity reconfigurations communicate societal changes minority members wish to see (Smith et al., 2015), for example a normative change in ethnic (or racial) groups differentiation. In the present study, we consider the interplay of ethnic and national identifications when examining the relationship between contact and activism.

# National Identity and Its Mobilizing Effect for Ethnic Minorities

National identification has been conceptualized as a relational construct, which "provides a means of appealing to a group of people [...] within a given territory" (Reicher and Hopkins, 2001, p. 26). As national subgroups, ethnic minorities can thus identify with, that is feel attached to, both ethnic and national groups. Furthermore, "native" or traditional ethnic minorities may strongly identify with the nation considered their ancestral homeland (Sibley and Liu, 2007). Indeed, the strength of ethnic minorities' national identification has been shown to vary across countries (e.g., Staerklé et al., 2010).

As ethnic minorities can simultaneously identify with an ethnic and a national group, we argue that when interpreting the sedating effect of positive intergroup contact on minority activism via reduced ethnic identification national identification should be accounted for. Recent research supports such reframing. For example, national identification can increase minority members' expectations to be treated fairly and help them believe in social change (Chryssochoou and Lyons, 2011). As a result, when unequal treatment is experienced, national identification fuels minority members' disappointment (Klandermans et al., 2008). National identification also fosters feelings of entitlement to political rights among immigrant minority members (Klandermans et al., 2008; Scuzzarello, 2015) as well as claims of country "ownership" among established minorities (Brylka et al., 2015). Additionally, minorities can strategically navigate identities (e.g., adapt their statements of identification) as a function of the audience they are communicating with. Ethnic minority members may make more national identification claims before a host country audience compared to an audience of ethnic peers (Barreto et al., 2003). Furthermore, a particular overlapping of ethnic and national self-definitions labeled dual identity (i.e., identifying as both a member of the ethnic minority and of the national group) has been shown to uniquely predict support for ethnic minority activism (e.g., Simon and Ruhs, 2008; Simon and Grabow, 2010; see also Verkuyten, 2016, for the effects of normative contexts in this process). For example, in the US, Glasford and Dovidio (2011) found that ethnic minority members exposed to a dual identity representation (as both American and member of an ethnic group) were more motivated to address disparities during intergroup interactions compared to participants exposed to a common ingroup representation (as solely American). The mobilizing effect of ethno-national identification stresses an important feature of minority activism: when minority group members become active, they do so because their minority identity is defined in terms integrating the "more inclusive societal context in which this struggle has to be fought out" (Simon and Klandermans, 2001, p. 319; see also van Zomeren, 2016).

To summarize, national identification is a dynamic identity process observed among minority members, alongside ethnic identification. Our reading of research on intergroup contact, negative social identity management strategies, and collective identity explanations of group activism (see also Çakal et al., 2011) suggests that the synergy of ethnic identification—involving separation and grievances—and national identification—bringing entitlement—more accurately reflects the collective identification of ethnic minority members and should thus be considered when examining the sedating effects of positive contact. The goal of the current study is thus to examine whether the negative effect of positive intergroup contact on ethnic activism via reduced ethnic identification is buffered by the national identification of minority members.

We make a novel theoretical contribution by examining national identification as a buffering factor of the demobilization process resulting from intergroup contact. Reasoning in terms of identity (re)configurations allows considering identification as a dynamic process rather than a fixed ethnic identity (see Tajfel and Turner, 1979). As prior research has shown the importance of multiple identities for endorsement of social change, in particular the combination of ethnic and national identities, we suggest that national identification counteracts the sedative effects of positive contact. Based on the outlined theoretical arguments, in our research in Bulgaria, we examined the moderating role of national identification in the relationships between contact, ethnic identification and ethnic activism. In particular, we hypothesize a moderated mediation pattern underlying the sedative effect of intergroup contact (see **Figure 1**): the mediation pattern from intergroup contact to reduced Roma activism through Roma's weakened ethnic identification should be moderated by identification with the Bulgarian nation. Identification as a member of the Bulgarian nation should thus buffer the sedative effect of experiencing positive contact with members of the Bulgarian national majority. We expect that the indirect effect from intergroup contact to activism through ethnic identification

emerges only for Roma with low national identification, or at least is stronger for them than for Roma with high national identification.

# The Context of the Present Study: Roma in Bulgaria

We also make novel empirical contributions to intergroup contact and collective action literature by examining the perspective of the Roma minority in Bulgaria. Roma are an understudied group, despite being historically and currently the most severely rejected ethnic minority in Europe (Heath and Richards, 2016). The study of negative identity management among Roma is thus particularly important. However, others perceive Roma as an ethnicity more than Roma themselves (Csepeli and Simon, 2004). In other words, the Roma ethnicity is formed by discrimination experiences and attributed stereotypes as least as much as by a particular appearance, language or ancestry (Kligman, 2001). Moreover, Roma are spread all over Europe forming small national minorities that self-identify with the various national groups or traditions (Marushiakova and Popov, 2007). Accordingly, Roma minorities are an interesting case for the study of collective identity predictors of activism, because both the abstract and externally defined ethnic identity and the various nation-specific minority identities can be potentially mobilized.

In Bulgaria, the Roma minority has experienced assimilation policies before, during and after the communist regime (Marushiakova and Popov, 2008a). Such policies have certainly also affected the self-identification of Roma across generations. Indeed research has shown that Bulgarian Roma and ethnic Bulgarian adolescents do not differ in their level of attachment to the nation (Dimitrova et al., 2014). After the political transition to democracy, Bulgaria declared itself a multicultural country and highlighted the different cultural and religious groups in its constitution. Compared to neighboring countries also with large Roma minorities (e.g., Hungaria, Romania), contemporary Bulgaria is the only country with a national strategy for integration of Roma people in its legislation (European Commission against Racism and Intolerance, 2014). Yet, the upsurge of ethnic nationalism in Eastern Europe along with economic difficulties throughout the European Union have reached Bulgaria and constitute a renewed threat to Roma inclusion. Roma communities are scapegoated (see, e.g., Petkov, 2006), subjected to ethnic characterization and systematic othering (Kostova et al., 2011), and relegated to a sub-proletarian class (Vassilev, 2004). In this context, support for Roma activism against persisting ethnic discrimination is a highly relevant, albeit sensitive, enterprise. These observations pointing to an integrative normative context coupled with ethnic stigmatization call for a nuanced perspective when studying the relationship between contact with the Bulgarian majority and Roma activism (see also Kauff et al., in press for cross-national evidence). Roma's identity management and adjustments related to ethnic and national identity need to be considered (see Reysen et al., 2016 for a similar approach of the Roma issue).

# MATERIALS AND METHODS

# Participants

Three-hundred-twenty self-declared Roma from Bulgaria participated in this study.<sup>1</sup> A two-stage sampling procedure was used, which consisted of first determining sampling points in both urban and rural areas, and then seeking eight respondents stratified by age and gender from each sampling point. Respondents were recruited by a Bulgarian survey agency in two districts of the country: Montana (north-west) and Stara Zagora (center). The final sample consisted of 162 females and 158 males respondents (Mage = 43.30 years, SD = 16.66, range = 15–83 years old) nested in 40 clusters (i.e., sampling points). All respondents reported Bulgarian as their first language

<sup>1</sup>Our target sample size (determined by the fieldwork budget as well as statistical considerations) was 300 Bulgarian Roma respondents. We anticipated model designs with 10–15 parameters and bore in mind suggestions for sample sizes of 20 times the number of parameters (e.g., Kline, 2015). The final sampling point grid allowed for a sample of 320 Bulgarian Roma respondents in the entire survey involving national majority and minority members.

with the exception of one participant who reported Romani. Regarding educational level, 15.3% of Bulgarian Roma had never been to school or not completed primary education. The large majority (60.3%) of respondents had primary or lower secondary education. The proportion of respondents having an upper secondary education was 23.1%, whereas 1.2% had a university degree. Note that those with upper secondary and university education were slightly overrepresented in our sample as compared to official figures (Pamporov and Kabakchieva, 2012). We also asked participants to define what was their own or their family's current economic situation on a scale ranging from 1 (We have enough money for our needs and are able to save) to 5 (We have to cut back on consumption and we don't manage on our earnings). The subjective economic situation of Roma in our sample was modest (M = 4.03, SD = 0.97), with only 7.8% or respondents reporting having enough for their needs.

## Procedure and Measures

The data used in this study are part of a research project examining social psychological processes underlying interethnic attitudes and prejudice of both the Bulgarian majority and of the two largest ethnic minorities (Roma and Turks) in contemporary Bulgaria. The survey questionnaires were designed in English and then translated into Bulgarian using a backtranslation method. They were administered face to face by professional interviewers, who were members of the national majority. This survey was carried out in accordance with the recommendations of the Code of Deontology of the Swiss Psychological Society and of the American Psychological Association, and in compliance with the Law for Protection of Personal Data in Bulgaria. Respondents were provided with the necessary information for informed consent, as well as guaranteed anonymity and right to withdraw from the survey at any time.<sup>2</sup>

**Table 1** summarizes means, standard deviations, and correlations for the final variables used in the subsequent analyses.

<sup>2</sup>The questionnaire for the Roma sample included other measures that were not considered in the current study. The order of questionnaire items was the following: ethnic and national identification, ideological orientations, intergroup contact with and prejudice toward ethnic Bulgarians, intergroup contact with and prejudice toward Bulgarian Turks, social distance, status perceptions, perceived discrimination, acculturation orientations and expectations, political behavior and ethnic activism, and socio-demographics.



<sup>a</sup>Standardized score; ∗∗p < 0.01; ∗∗∗p ≤ 0.001.

We assessed close positive contact with three items (see Voci and Hewstone, 2003; Pettigrew et al., 2007) addressing the quantity and quality of Bulgarian Roma's close contact with ethnic Bulgarians ("How many ethnic Bulgarians do you know well?" and "How often do you experience these encounters with ethnic Bulgarians you know well as pleasant?") and extended contact ("How many people you know have ethnic Bulgarian friends?"). Quantity of contact (direct and extended) was rated on a scale ranging from 1 (None) to 4 (Many), whereas quality of contact was rated on a scale ranging from 1 (Never) to 5 (Always). The respondents reported both a lot of close contacts with ethnic Bulgarians (M = 3.38, SD = 0.95) as well as knowing many Bulgarian Roma having ethnic Bulgarian friends (M = 3.40, SD = 0.85). Since the quality item was filtered for the respondents reporting no contact with ethnic Bulgarians, we considered these respondents "never" having positive contact and replaced the missing values (n = 34) accordingly (i.e., with 1). Respondents perceived their encounters with ethnic Bulgarian as generally positive (M = 3.65, SD = 1.24). Extended contact was strongly correlated with quantity (r = 0.73, p < 0.001) and quality of contact (r = 0.64, p < 0.001). Moreover, a principal component analysis yielded a single factor explaining 80% of variance. Thus, the three items were considered as representing a single concept, that is, close positive contact with ethnic Bulgarians. Due to different response scales, the three items were standardized prior to computing a close positive contact score (α = 0.88; see also Pettigrew et al., 2007 for an intergroup friendship index using both direct and extended contacts).

Ethnic and national identification were measured with three parallel items ("Do you often think of yourself as a member of the Bulgarian nation/Roma?" "Is being part of the Bulgarian nation/Roma important to you?" "Do you feel close to other members of the Bulgarian nation/Roma?," e.g., Sidanius et al., 1997). The scale ranged from 1 (No, not at all) to 5 (Yes, very much). Both identification scales reached good or adequate reliability (for national identification α = 0.84, for ethnic identification α = 0.69). Both high level of ethnic and national identification were reported in the sample though ethnic identification was significantly higher than national identification, t(319) = −6.77, p < 0.001. The two identification indices were moderately correlated.

We assessed ethnic activism with four items (α = 0.85; see collective action scale of Mähönen and Jasinskaja-Lahti, 2015) addressing the willingness to improve the position of Roma in Bulgaria by different political contributions (vote for a candidate defending Roma's rights; defending the rights of Roma in public debate; defending the rights of Roma in situations where you notice discrimination; taking part in cultural events organized by Roma).<sup>3</sup> The five-point scale ranged from 1 (No, not at all) to 5 (Yes, very much).

<sup>3</sup>The item on the willingness to participate in cultural events organized by Roma could be seen as measuring a social creativity strategy (i.e., coping) rather than as an active resistance strategy (Becker, 2012). However, the item was strongly correlated with the other ethnic activism items (rs > 0.44, ps < 0.001). Furthermore, a principal component analysis confirmed that the four items loaded onto a single factor explaining 69% of variance.


X, independent variable; Me, mediator variable; Mo, moderating variable; Y, dependent variable. Gender, age, educational level and perceived economic situation of the respondents were controlled for in the analyses.

†p = 0.054; <sup>∗</sup>p < 0.05; ∗∗p ≤ 0.01.

fpsyg-08-00477 April 6, 2017 Time: 15:32 # 6

# RESULTS

The analytic strategy was to first attempt to replicate the sedating effect of intergroup contact on ethnic activism through reduced ethnic identification, and then to test whether national identification can buffer this effect. Preliminary analyses revealed that, due to the clustered nature of the sampling (i.e., with eight respondents from the same sampling point), data is non-independent (ICC of ethnic activism = 0.50). We therefore tested path models using the Mplus Complex command, which allows accounting for the non-independence and non-normality of observations (Muthén and Muthén, 1998–2010, p. 533).<sup>4</sup> Ethnic and national identification indices were centered as their interaction was modeled in the second step of the analysis. Gender, age, educational level, and perceived economic situation of the respondents were controlled for.

First, we assessed the mediating role of ethnic identification in the relationship between close positive contact and ethnic activism. As expected, we replicate the sedating effect of contact. The left panel (A) of **Table 2** shows that contact was negatively associated with ethnic identification, which in turn was positively associated with ethnic activism. The indirect effect of contact on ethnic activism through ethnic identification was significant (B = −0.04, SE = 0.02, p = 0.044).<sup>5</sup> However, also an unexpected positive relationship between positive close contact and ethnic activism was observed. The only significant effect of the control variables was a positive relationship between educational level and ethnic activism (B = 0.11, SE = 0.05, p = 0.044).

Next, we tested our moderated mediation hypothesis. Close positive contact was entered as the independent variable (X), ethnic identification as mediator (Me) and ethnic activism as the dependent variable (Y). The direct path from close positive contact to ethnic activism was also estimated. National identification was specified as the variable moderating (Mo) the relationship between ethnic identification and ethnic activism. The findings are summarized in the right panel (B) of **Table 2** and depicted in **Figure 2**. Again the sedating effect of contact was revealed: close positive contact was related to reduced ethnic identification, which in turn was positively associated with ethnic activism. National identification was positively related to ethnic activism. Importantly, national identification moderated the relationship between ethnic identification and activism. As predicted, the indirect effect of close contact on ethnic activism mediated by reduced ethnic identification was buffered by national identification. The sedating effect was significant for low national identifiers (B = −0.06, SE = 0.03, p = 0.042), while not for high national identifiers (B = −0.005, SE = 0.01, p = 0.678). Positive contact sedated activism thus only for Bulgarian Roma who reported weak self-identification as a member of the Bulgarian nation.

The direct path from close contact to ethnic activism remained marginally significant and positive in the moderated mediation model. None of the control variables yielded significant effects. Finally, we also examined whether national identification moderates the path between close positive contact with the majority and ethnic identification (i.e., the first stage of the mediation), or the direct path from close positive contact to ethnic activism. Additional moderated mediation analyses revealed that neither one of these paths was moderated by national identification. This suggests the specificity of interplaying ethnic and national identification that predict support for minority activism.

# DISCUSSION

The aim of the present study was to disentangle the sedating effect of positive contact with the majority on ethnic minority activism, by considering not only the reduction of ethnic identification, but the synergy of ethnic and national identification. Our chief contribution was to show that national identification buffered this sedating process: the demobilization effect of contact through reduced ethnic identification was

<sup>4</sup>We also conducted the same analyses using the PROCESS macro (Hayes, 2013) and tested indirect effects with bootstrapping. The result patterns were identical. However, PROCESS does not take into account the nested structure of data, necessary for the current sample.

<sup>5</sup>The indirect effects, estimated with Mplus, are calculated as the product between the regression coefficient of the IV on the mediator and the regression coefficient of the mediator on the DV.

found only among individuals with low levels of national identification.

The revealed buffering effect of national identification clearly confirms the need to consider more than one identity when studying minority activism (see also Curtin et al., 2016). This finding also suggests the existence of a non-conforming style of national identification among ethnic minority members. As Moscovici (1976) proposed, minorities use productive – as opposed to sedative – forms of conformity and alignment with the social context. This is in line with research on the effects of dual identity on ethnic activism (e.g., Simon and Grabow, 2010), which shows that the motor of ethnic minorities' struggle is a collectively endorsed commitment to the higher societal level, which in turn results in an entitlement for claims in the name of the subordinate ethnic group. Furthermore, we chose to focus on national identification among other potentially relevant interplaying identities (e.g., religious, opinion-based, gender) for several reasons. The demonstration that Roma identify with the nation contributes to contesting the stereotype of a "nomadic" and stateless minority fundamentally different from other ethnic minorities. Indeed, the Roma are often seen as "forever pilgrims," defined as a non-territorial and transnational ethnic community (Vermeersch, 2003). Roma should be rather compared to Indigenous people with regard to attachment to the national homeland: though their ethnic group identity derives from family lineages rooted in a homeland, national majorities consider them as unassimilated in the modern nation state. Furthermore, national identification can provide this kind of internally segmented minorities the psychological and political resources to challenge intergroup disparities (see Çakal et al., 2016). In contrast to Indigenous people, Roma have personal experiences of state paternalism under the communist rule, or at least they have been exposed to nostalgic discourses about the less marked ethnic segregation back then as compared to the present. These considerations suggest that national identification is an important component of Roma collective identity.

Our results also revealed an unexpected positive direct relationship between positive close contact and ethnic activism, which was not moderated by national identification. Several studies have indeed highlighted that positive interethnic contact does not exempt experiencing negative emotions due to threat of persistent interethnic prejudice (e.g., Shelton et al., 2005). Contact with the advantaged group provides opportunities for personally assessing intergroup injustice and leading minority members to realize their disadvantaged position. For example, Poore et al. (2002) revealed that Inuits' perception of groupbased discrimination increased as a function of their contacts with the North-American culture. This result does not contradict our main finding of national identification buffering the sedative effect of contact, but simply suggests other parallel consequences of intergroup encounters.

# The Interplay of Ethnic and National Identification in the Bulgarian Context

The findings of this research require reflecting on the Bulgarian Roma's national identification that implies a common ingroup shared with ethnic Bulgarians. The demobilizing effect of contact has been originally studied in segregated societies (e.g., South Africa, USA, Israel). In those contexts, bringing people back together involves the development of a common ingroup identity that can overlap with national identity, and the abandon of ethnic (or racial) categories. In contrast, post-socialist Bulgaria has experienced an inverse normative shift, from implicit interethnic differences to more explicit interethnic recognitions. During the communist regime in Bulgaria, interethnic contact took place within an ethnonationalist ideology affirming equality of citizens whatever their ethnic origin, and state policies were implemented ensuring the public integration of Roma citizens (Marushiakova and Popov, 2008b). After the transition to democracy and to market economy, the Bulgarian nationalist discourse and constitution highlighted multiculturalism. Ethnic minority members were encouraged to develop a strong identification as a member of the

nation "colored" with ethnicity, as Pettigrew (2010) commented about the specificity of interethnic relations in Bulgaria. Today, the inherited ethno-nationalist model is reinterpreted through political and media discourse describing Bulgarian poverty and economic difficulties as having a specific "racial" origin: Roma (see Kligman, 2001). One must note though that Bulgaria is praised for having avoided any major interethnic clash.

Accordingly, Bulgarian Roma's feeling of national belonging may not reflect a by-product of their positive relationships with the majority, but rather suggest a reminiscence of, or even a battle for, access to a national group from which they are materially and symbolically excluded (Pereira and Green, 2017). This interpretation echoes research in the field of coping with social disadvantage. According to system justification theory (Jost et al., 2004), members of socially disadvantaged groups engage in a struggle for social change only when the unfavorable nature of their self- or group-image "overcomes the strength of system justification needs and tendencies" (p. 887). This claim supports our interpretation, since the image of the Roma as a national subgroup is unquestionably unfavorable. The negotiation of both ethnic identification—stigmatized but nevertheless protected by the Bulgarian constitution—and identification as a marginalized member of the nation results in an unfavorable collective identity. Likewise, Barreto and Ellemers (2009) proposed that identifying with an under-represented or culturally threatened group accentuates rather than sedates the perception of social disadvantage. Finally, Hopkins and Blackwood (2011) demonstrated that stigmatized minority members (i.e., Muslims in Britain), who have internalized the denial of their citizenship, sometimes actively claim citizenship as a form of resistance against others' negative assumptions about their identities. Future research should substantiate these interpretations of the combative nature of national identification among Roma people in Bulgaria and in other East European countries with somewhat different legislative frameworks regarding ethnic minorities.

Moreover, the sedating effect of positive contact with the majority revealed in this study (among low national identifiers) may be due to the legacy of implicit interethnic boundaries combined with exclusion from the current legitimate national group (i.e., ethnic Bulgarians). In a study conducted in England among marginal citizens of Roma origin, Casey (2014) found a tendency to favor the status quo, that he argued was a strategy for maintaining Roma ethnic identity and cultural traditions that have survived thanks to social exclusion. Additionally, it is likely that contemporary Bulgarian Roma experience some degree of political cynicism (due to corruption, prolonged discrimination, etc.), known to impede activism (Klandermans et al., 2008). Further studies should unravel whether political cynicism is relevant in Roma minorities' struggle for social equality. More generally, to better understand what predicts ethnic minority activism, we must consider how intergroup contact between minorities and majorities (along with the normative context in which it takes place) constructs the interethnic issue at stake in the first place. Eastern European countries provide a fascinating field to study how minorities (and majorities) use their ethnic and national belonging to adapt to recent social change, and eventually readjust and react to the new emerging inequalities.

Some limitations of our study must be acknowledged. First, the wording of the national identification measure (i.e., being member of the Bulgarian nation) might have triggered the interplay of sub- and superordinate identities. This wording was, however, the best alternative in this context: asking Roma respondents to what extent they felt "Bulgarian" could come across as either insulting (i.e., as Bulgarian Roma have no other homeland, they may perceive that the interviewer sees them as authorized or unauthorized immigrants) or misleading (i.e., some Roma subgroups claim to be "ethnic" Bulgarians because of their lineage with proto-Bulgarian people living on the national territory before the modern state constitution). Second, one should not underestimate the bias in self-reporting intergroup contact and support for ethnic activism, in particular as the interviewers were members of the national majority. Confronted to a non-Roma audience, survey respondents may have made national identification claims (see Barreto et al., 2003). Third, the support for ethnic activism in our sample does not reflect the actual implementation of Roma activism in Bulgaria. Other predictors of engagement in activism should be accounted for in future research on Roma activism. In addition to support for collective activism, more individual responses to social disadvantage or the combination of individual and collective strategies should be considered (see Becker et al., 2015). Finally, no firm causal conclusions regarding the relationships between variables can be drawn from our cross-sectional design. Still, a large number of experimental studies have established that collective identity and intergroup contact are important predictors of activism increasing our confidence in the current findings. In our view, studying activism among a harshly stigmatized, yet understudied group such as Roma offsets the caveat of being unable to make firm causal claims.

# CONCLUSION

This study contributes to the scarce body of knowledge about Roma's activism (see, e.g., Vermeersch, 2014; Reysen et al., 2016 for other recent studies). We revealed that Roma activism in Bulgaria was shaped by contact with the advantaged national majority as well as by an interplay of ethnic and national identification forming a collective Bulgarian Roma identity that was mobilized to support social change. We demonstrated that the previously evidenced sedating effect of contact on activism due to a reduced ethnic identification can be counteracted by identifying with the Bulgarian nation. Our results and interpretations stress the adaptive (and instrumental) role of identification to both a subordinate ethnic and a superordinate national group, which need to be considered within the specific societal structures (van Zomeren, 2016). Rather than being fixed realities, ethno-national identities are collective identities, construed through communication and interactions, that vary across contemporary nations.

# AUTHOR CONTRIBUTIONS

fpsyg-08-00477 April 6, 2017 Time: 15:32 # 9

AP, EG, and EV developed the study concept and designed the research. AP and EV performed the data analysis and interpretation under the supervision of EG. AP and EG drafted the manuscript, and EV provided critical revisions.

# REFERENCES


# FUNDING

This work was funded by a joint grant from the Swiss National Science Foundation and the Ministry of Education, Youth and Science of Bulgaria (IZEBZ0 142998) awarded to EG and Yolanda Zografova.


Moscovici, S. (1976). Social Influence and Social Change. London: Academic Press.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Pereira, Green and Visintin. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# 'The More We Stand For – The More We Fight For': Compatibility and Legitimacy in the Effects of Multiple Social Identities

#### Maria Chayinska<sup>1</sup> \*, Anca Minescu<sup>2</sup> and Craig McGarty<sup>3</sup>

<sup>1</sup> Department of Psychology, University of Milano-Bicocca, Milan, Italy, <sup>2</sup> Department of Psychology, University of Limerick, Limerick, Ireland, <sup>3</sup> School of Social Sciences and Psychology, Western Sydney University, Sydney, NSW, Australia

This paper explores the expression of multiple social identities through coordinated collective action. We propose that perceived compatibility between potentially contrasting identities and perceived legitimacy of protest serve as catalysts for collective action. The present paper maps the context of the "Euromaidan" anti-regime protests in Ukraine and reports data (N = 996) collected through an online survey following legislation to ban protests (March–May, 2014). We measured participants' identification with three different groups (the Ukrainian nation, the online protest community, and the street movement), perception of compatibility between online protest and the street movement, perception of the legitimacy of protest, and intentions to take persuasive and confrontational collective action. We found evidence that the more social groups people "stood for," the more they "fought" for their cause and that identifications predicted both forms of collective action to the degree that people saw the protest and the online movement as compatible with each other and believed protest to be legitimate. Collective action can be interpreted as the congruent expression of multiple identities that are rendered ideologically compatible both in online settings and on the street.

Keywords: multiple social identities, perceived compatibility, perceived legitimacy of protest, collective action, political activism

# INTRODUCTION

A very relevant issue to address when examining the dynamics of grass-roots collective action is what type of commitments drive individuals' behavior and how the real-world structural context conditions collective efforts to attain social change. Although collective action is routinely understood as efforts by members of a disadvantaged social group to overturn an injustice, the concept itself suggests the need to look beyond a single, nominal social category membership as the seed of dissent toward contested and multifaceted political agency and, thus, multiple politicized collective identities as a potential explanation of the drivers of social movements for social change. Did the Russian revolution of 1917 establish the dictatorship of the proletariat? Are efforts 100 years later to "make America great again" directed to the benefit of all (US) Americans? Perhaps, but it also seems plausible that in these, and many other cases, that there is a number of salient social categories that may be relevant at the same time for either a community or the same individual actor. The multiplicity of actors, political agendas and group identities are likely to achieve higher

#### Edited by:

Soledad de Lemus, University of Granada, Spain

#### Reviewed by:

Paulina Górska, University of Warsaw, Poland Andrew L. Stewart, Clark University, USA

#### \*Correspondence:

Maria Chayinska m.chayinska@campus.unimib.it

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

> Received: 27 October 2016 Accepted: 11 April 2017 Published: 26 April 2017

#### Citation:

Chayinska M, Minescu A and McGarty C (2017) 'The More We Stand For – The More We Fight For': Compatibility and Legitimacy in the Effects of Multiple Social Identities. Front. Psychol. 8:642. doi: 10.3389/fpsyg.2017.00642

mobilization power in certain contexts, and social psychological models of collective action should be able to account for effects of such multiple identities.

The present research is designed to answer specific questions about how identification with distinct social groups coheres to underpin engagement in coordinated collective action. In particular, we aim to understand the process through which multiple social identities of self are translated into synchronized political action as well as conditions under which people opt for different forms of collective action. We seek to understand this phenomenon in relation to the wave of political activism in Ukraine starting in 2014.

Psychological research (as highlighted in this Topic) has demonstrated that people belong to a number of social groups and affiliations that can be potentially mobilized and politicized and has posited the question of whether these multiple commitments of self can lead to a synchronized expression (e.g., Cruwys et al., 2016; Curtin et al., 2016). Despite the increasing interest the underlying mechanisms of the expression of multiple identities remain unspecified.

The matter is complicated further because social movements may reflect not only multiple agendas but multiple methods. Some of these methods involve building support by persuading potential supporters to join a movement whereas others involve disrupting or even destroying opposition. Scholars have sought to understand the causes of extreme, non-normative and violent collective action (e.g., Moskalenko and McCauley, 2009; Thomas and Louis, 2014; Thomas et al., 2014; Becker and Tausch, 2015; Jiménez-Moya et al., 2015; Shuman et al., 2016) by distinguishing them from moderate, normative, and peaceful action. We appreciate that all of these distinctions have merit for various purposes In this study, we rely on the distinction between persuasive action –as a form of protest with the primary purpose of influencing/persuading third parties (or even opponents) to share a political goal and confrontational action, conceived as a form of protest that confronts opponents with direct action that may disrupt their activities. The advantage of this distinction is that labels for action such as "non-normative," "unlawful," "violent," and "extreme" are subject to locally applicable definitions that are often within the power of authorities to define. This is generally problematic where those authorities are themselves the targets of action, but is specifically problematic where forms of protests are outlawed during the course of a campaign. In Tunisia in early 2010 street protests were both illegal and very uncommon (McGarty et al., 2014). Protests continued to be illegal right through to the point that the Ben Ali regime was overthrown but they had become common right across the country by early January 2011. In Ukraine in 2013/14 public demonstrations, most famously in the Maidan Square in Kyiv, had become regular and heavily supported events, but in January of 2014 they were declared to be illegal, prompting a new wave of intensified protests.

A recognition of context in promoting and constraining the expression of social identities has prompted analysis of the perception of political opportunities in relation to the anticipated outcomes of protest efforts (e.g., Williams, 2004; Reicher and Haslam, 2013; van Stekelenburg and Klandermans, 2013). However, little attention has been paid to the role of the perception of legitimacy of protest in predicting different forms of collective action.

The present research readdresses these issues and suggests a framework for understanding the expression of multiple social identities situated in a specific historical context. The key objective of the present research is, therefore, to examine the mechanisms behind a synchronized expression of multiple social identities in explaining persuasive and confrontational collective action. In line with the social identity approach, we first propose that collective action can be explained to a greater extent by accounting for multiple social identities whose ideological contents are aligned rather than by focusing on a singular salient category membership (hence the title of this paper 'the more we stand for – the more we fight for'). In particular, based on self-expansion theory (Aron and Aron, 1996; Aron et al., 2004), we assume that people expand their self-concepts to include different identities of groups and communities they belong to, and this can occur without individuals necessarily incorporating or nesting one social identity into another. This psychological process, also referred to as the inclusion of other in the self, is thought to be achieved through an increasing overlap between the representations of self and social groups (e.g., Tropp and Wright, 2001).

Secondly, we argue that the ideological content of these identities need to be (or become) compatible with each other in order for them to drive collective action (see Bliuc et al., 2012; McGarty et al., 2014). We thus suggest that holding a shared (civic) vision based on the perceived compatibility of multiple identities provides solid psychological ground for engaging in collective action.

Another important consideration, in addition to the compatibility between identifications, is the degree to which the political opportunity structure (e.g., Tarrow, 1998; Meyer, 2004), that is, system-level constraints of individual-level intentions to take collective action, imposes a particular set of expectations regarding the ways in which those multiple identities may be expressed. In other words, if the norm that protest is a legitimate way of engaging in collective action is aligned with multiple identities, then the perception of protest as legitimate will help explain the effects of these identities on collective action. This is a particularly timely and contextually relevant operationalization, capturing people's perception of a key feature of the political opportunity structure in contested times of transition.

# Compatibility of Multiple Identities and Political Activism

The idea that collective action may be explained through politicization of multiple social identities has recently received more attention in collective action research (e.g., Curtin and McGarty, 2016; Curtin et al., 2016; Louis et al., 2016; building on earlier insights by Klandermans et al., 2008; Simon and Ruhs, 2008).

It has been argued that the psychological processes behind the simultaneous expression of multiple identities might involve the formation of opinion-based groups (see McGarty et al., 2009),

where the content of commitments (what 'we' stand for and what 'we' stand against, see Chayinska et al., in press), rather than strength and salience of social identification, has been shown to be the key factor to understanding politicization and action engagement. Smith et al. (2015b; see also Smith et al., 2015a) conceptualize this as the formation of an identity-norm nexus where people come to see shared views about how to change the world as an aspect of self. Qualitative analysis by Curtin et al. (2016) also revealed that individuals who experience marginalization and privilege at the same time, and arguably identify with advantaged and disadvantaged groups, tend to simultaneously express these multiple identities to the extent to which these identities may be subsumed under a broader identity category (i.e., interpretable as involving commitment to a common cause).

Turner-Zwinkels et al. (2015) have shown that politicization of social/personal identities is not merely a matter of increasing allegiance to multiple political agendas; it is the overlap in the normative content of these identities and a subjective internalization of their agendas through which the political becomes personal that predicts commitment and action. Similarly, Louis et al. (2016) contend that one of the reasons why activism in one domain (i.e., identification with Cause 1) might predict and facilitate the likelihood of activism in other domain (i.e., identification with Cause 2) is the ideological or normative alignment between these movements. According to these authors, it is therefore necessary to explicitly measure whether and how such a normative consensus leads to collective action.

Other scholars have highlighted that a meaningful interconnectedness of available multiple identities (e.g., Case et al., 2012; Greenwood, 2012) or so called 'identity-value fit' (e.g., Kutlaca et al., 2016) tends to facilitate their simultaneous expression, and that holding a number of social commitments, as opposed to a sparse social identity profile, is beneficial to life transitions. However, there appears to be one crucial condition: the multiple identities one holds need to be perceived to be compatible with each other (e.g., Benet-Martinez and Haritatos, 2005; Riketta and Nienaber, 2007; Iyer et al., 2009).

Taken together, these findings suggest that the more people perceive their multiple social identities to be compatible the higher the level of identity integration. Conversely, the perception of two or more identities being in opposition to each other, perhaps due to conflicting values and norms, signals a lower level of identity integration. We extend this line of research by suggesting that that expression of multiple politicized identities through collective action is more likely to occur when individuals do not have to make an 'either–or' choice between two or more commitments. In other words, when a high level of identity integration between multiple identities is present, thus when identities are perceived as more compatible, collective action is more likely to emerge.

Based on this literature review, we suggest that the greater the degree of normative or ideological compatibility between multiple social identities the more likely it is that they will lead to coordinated collective action for the same cause. Thus, the present study investigates the potential mediating role of perceived identity compatibility in the relationships between the identification with the online protest community and street movement and intentions to take persuasive and confrontational collective action. We conceptualize perceived compatibility between multiple identities as the extent to which their content (and the values assigned to it) are perceived at the individual level to be coherent and in congenial combination with one another. In other words, for multiple identities to be psychologically compatible, we assume, the identification with one social group must not be perceived as conflicting with identification with another group.

While we see ideologies as a perfectly viable basis for the formation of social identities (most obviously in relation to political groupings such as socialist and fascist) our focus here on the link between ideology and identity is chiefly in terms of the perceived compatibility of identities as they relate to participation in protest. Thus one pro-democracy, pro-European protester may hold an ideological commitment to non-violence or to obeying national laws (even when they are seen to be unjust) and another might believe that democratic ends justify violent means, or that 'bad' laws need to be broken. We turn to these matters now.

# Perceived Legitimacy of Protest

It has been widely accepted that the context within which politicized collective identities emerge plays an important role in the understanding of political collective action and its consequences. Past research has paid insufficient attention to the fact that the legitimacy of engaging in protest against authorities or for a particular cause is itself a very contested aspect of social structure, and therefore varies across political contexts. Although it is commonly taken for granted in liberal democratic settings that political structures accommodate the right for participation in protest, this is not true in most parts of the world throughout history. This pattern may potentially challenge the cross-cultural applicability of findings from Western democratic contexts to other contexts where transition between political regimes and democratization is an ongoing process and challenging reality.

Some political science research (e.g., Tarrow, 1998; Meyer, 2004; Corcoran et al., 2011) indicates significant links between democratization and protest such that a change in some dimensions of the political opportunity structure tends to affect an individual perception of the feasibility of protest. For instance, analyzing the data from the World Values Survey, Corcoran et al. (2011) have revealed that the perception of political institutions as open (a macro-level factor) affected individuals' sense of efficacy (a micro-level factor), which in turn was found to determine intentions to take collective action. Social psychological research illustrated these processes in the analysis of McGarty et al. (2014) looking at the protests against repressive regimes in North Africa in 2010 and 2011. In this context, protest came to be seen as feasible after striking novel images of anti-regime protest were recorded on camera phones, uploaded to social media video sharing sites (e.g., YouTube), and from there, broadcast through external satellite television networks (Al Jazeera) to citizens in Tunisia and Egypt. Arguably in this context online mobilization was not alienated from street protest but was a precondition for it: part of a broader global pattern that Castells (2012)

describes as the occupation of specific online spaces preceding the occupation of physical public spaces. Also, McGarty et al. (2014),' research captures the transition in people's perception that protest is "allowed," and the agency with which actors expanded this legitimacy of protest from the online to the street contexts.

Perceiving protest to be a legitimate political act is likely to result in collective action involving conventional, persuasive forms of action, but perhaps less so in the more confrontational forms of action (e.g., Simon et al., 1998; Thomas and Louis, 2014; Becker and Tausch, 2015). In the present study we examine whether perception of protest as a legitimate instrumental tool to achieve social change may also be rooted in the process that governs expression of multiple politicized identities. Specifically, we suggest that the ideological alignment (i.e., perceived compatibility) between different social identities along with perceiving protest as legitimate will explain how the relevant politicized identities will generate collective action. These processes are assumed to explain why and how multiple identities may align to predict engagement in collective action. In other words: 'the more we stand for' (multiple identities), 'the more we fight for' (increased collective action), because ideologically 'we' are fighting for the same goals (perceived compatibility) and because 'we' perceive our actions in protest as legitimate (perceived legitimacy of protest).

# Current Study

We tested these ideas in the context of the 2014 Euromaidan movement – an uprising against the refusal of the then Ukrainian national government to sign the Association Agreement with the European Union. After a set of pro-Western protests – the Russian-aligned majority of the Parliament of Ukraine passed a set of anti-protest laws that included measures limiting street assemblies and internet freedoms (Cohen, 2014). The new laws, criminalized all unauthorized meetings and gatherings in public places, and the online dissemination of "extremist information" (without providing a clear definition of 'extremist,' Centre for Civil Liberties, 2014). In the space of a few months, the political opportunity structure changed: the legitimacy of protests came to be contested in the midst of a political identity crisis of allegiances toward Ukraine, Europe or the Russian-led Customs Union among protesting Ukrainians. This is an especially intriguing context because of the legislative change. A growing social movement that sought to promote closer ties with Western Europe was confronted with new laws that made both street and online protest illegal. Obviously, however, street protests remained more detectable and punishable by authorities and 82 of street protesters were killed, more than 1,100 injured and 234 arrested in the period after the new laws were introduced (Ukraine Crisis: Timeline, 2014).

We captured this moment in this study, looking at whether and how participation in an online protest movement become an acceptable alternative to street protest, whether online activism may represent the legitimate continuation of the protest by other means in order to preserve the future of Ukraine, or whether online protest become an unsatisfactory and alienated substitute: expressing what Morozov (2009, 2011), Gladwell (2010) and others might deride as slacktivism, clicktivism or even in Morozov's terms "the net delusion" (see Schumann and Klein, 2015; Thomas et al., 2015).

We tested a model in which perceived identity compatibility and perceived legitimacy of protest mediate the relationships between multiple identities and collective action. We expected that, in the context of anti-government protest, people may find that there are more than one group or community that best represents their interests, and if they perceive that the values of these several groups are compatible (not conflicting), they will be likely to express their joint claims on behalf of those communities (the more we stand for – the more we find for).

We included three different social identifications as predictors of action: identification with the street protest movement, identification with the online protest movement, and Ukrainian national identification. We expected all three to be relevant predictors but the inclusion of national identification allowed us to address the possibility that identification with the single most relevant existing social category could provide an adequate (and parsimonious) account. Ethnic identification in terms of Ukrainian and Russian heritage represented other alternatives to measure single identities, and may seem obvious choices to external observers in view of recent dramatic conflicts in Ukraine. However, the civic ideology of the modern Ukrainian state (in which most participants would have been socialized) eschewed categorizations based on ethnicity in favor of a wider national identity category (see Prizel, 1998).

We expected that both persuasive and confrontational forms of collective action would flow from identification with the three different social identities (identification with Ukraine, identification with the online protest community, and identification with the Euromaidan street movement). Moreover, perceived compatibility and perceived legitimacy are expected to explain the effects of multiple identities on collective action. We generally also expected that the predictive power of the model including multiple identities and perceived compatibility and legitimacy will be stronger for persuasive than confrontational forms of collective action. This is because when considering persuasive collective actions, people are more likely to act out of a coherent ideological alignment between their multiple identities and the normative beliefs about these identities and about protest. When it comes to confrontational forms of collective action, this alignment between identities and normative beliefs might not be necessary. We tested these hypotheses with survey data collected during the 2014 protests.

# MATERIALS AND METHODS

# Participants and Procedure

Participants were approached through a public online survey posted to Facebook pages that were generally discussing political events in Ukraine. The data were collected between March 28 and April 30, 2014, (as soon as possible after the January 26 passage of laws that restricted people's right to protest led to larger protests in Ukraine). The questions of the survey focused on sociodemographics and attitudes toward current political issues. The items were available in separate Ukrainian and Russian versions

of the survey instrument. In order to guarantee coherence and validity of the questions, all items were translated from English to Ukrainian/Russian and back using a standard translation-backtranslation procedure (Brislin, 1970). Participants were required to be of Ukrainian nationality and aged over 18.

In total, the responses from 996 participants were used in the data analysis. The sample ranged in age from 18 to 77 (Mage 33.87 years, SD = 9.61) and comprised 51.7% women. Participants were highly educated (57.4% having graduated from university), 44.2% were employed full time, and 57% indicated Ukrainian as their first language. Some 72.7% reported that they completed this survey while in Ukraine, 24.7% – while living abroad (mostly in European countries, 15.5%, and in North America, 4.4%).

# Measures

#### Socio-Demographics

Participants indicated age, gender, ethnicity, country of birth, current residence, prior experience of living abroad, educational level, employment status, and mother tongue (i.e., Ukrainian, Russian, other).

#### Identification with Online Protest Community and the Street Movement

We measured self-expansion with the online protest community and with the street movement using a modified Inclusion-of-the-Other in-the-Self-Scale (the IOS-scale, Aron et al., 1992). The IOS task depicted five pairs of circles (numbered one to five), ordered by degrees of increasing overlap between the pairs. Self-expansion refers to a "fundamental human motivation to enhance potential self-efficacy (which is the ability to accomplish desired goals by attaining) greater material, social, and informational resources" (Aron and Aron, 1996; Aron et al., 2004). Participants were asked to indicate how close they felt toward online protest community and street movement, respectively, by selecting one of the five pairs of circles. Higher numbers are indicative of a smaller felt distance between oneself and others participating in the movement.

#### Identification with Ukraine

Six items from Leach et al. (2008) were used to measure identification with Ukraine. These and other measures below used five point Likert scales labeled from 1 (Strongly disagree) to 5 (Strongly agree). These items captured Leach et al. (2008) dimensions of centrality (e.g., "I often think about the fact that I am a part of the Ukrainian people"), satisfaction (e.g., "I am glad to be part of Ukraine"), and solidarity (e.g., "I feel solidarity with people in Ukraine") of identity, that comprise the second order dimension of group-level self-investment and are considered to be particularly important for collective political action. The items were averaged to form a composite measure of identification with Ukraine [Cronbach's alpha (α) = 0.95].

#### Perceived Compatibility

To measure perceived compatibility between the online protest community and the street movement we used four items adapted and modified from Riketta and Nienaber (2007): 'this online community is another platform for the street protest,' 'by becoming members of Online Protest Community people safeguard the very existence of the street protest,' 'in general, the mission statement of Online Protest Community fits well with the mission statement of the street protest,' and 'the ideas of Online Protest Community concerning interaction and cooperation correspond to the ideas of the street protest'), α = 0.79.

#### Perceived Legitimacy of Protest

Beliefs about legitimacy of protest were assessed using a 6 item scale: 'These people were wasting their time protesting (recoded),' 'I think protesting on the streets was a valid form of behavior in Ukraine,' 'Protesting changed nothing (recoded),' 'I think this was irresponsible behavior (recoded),' 'I think there should be more protests in Ukraine,' 'This was not typical Ukrainian behavior,' α = 0.71.

#### Persuasive and Confrontational Collective Action

Respondents were asked to indicate how willing they were to participate in 10 different offline collective actions. Principal components analysis yielded two components with eigenvalues greater than 1 that accounted for 55.45% of the variance. Loadings, after oblique rotation, revealed that relatively nonviolent, persuasive actions (e.g., 'voice group's claims in social network pages,' 'display symbolic attributes (flags, stripes) of my group, 'participate in marches and motorcades,' 'donate money for the cause of my group,' 'compile a blacklist (list for lustration, sanctions),' and 'participate in flash-mobs and art events organized to support the cause of your group') loaded primarily on the first component (41.10%); seemingly extremely confrontational actions (e.g., 'blockade activity of ideological opponents,' 'sneer at opponents' symbolic attributes (e.g., flags),' 'participate in mock political funerals,' 'sabotage political events of opponents') loaded on the second component (14.35%). The items were averaged to yield composites of individual's likelihood to engage in persuasive (α = 0.84) and extremely confrontational (α = 0.70) collective action. The two scales were moderately correlated (r = 0.592, p < 0.001).

# RESULTS

# Statistical Analyses

The preliminary analyses involved bivariate analysis and hierarchical multiple regression. In this step, predictor variables were centered when computing interaction terms to minimize colinearity. The main analysis involved a test of the mediational model. Data analyses were performed using SPSS 24 and Amos 24. Fit statistics, including χ 2 test (which can be affected by sample size), root mean square error of approximation (RMSEA), comparative fit index (CFI), and Akaike Information Criterion (AIC) were evaluated (Kline, 2011). The standardized paths between the variables included in the model were examined. The magnitude of effect sizes for the regression paths was determined as 0.10, 0.30, and 0.50 for small, medium, and large effects (Cohen, 1992). A p-value of less than 0.05 was considered to be statistically significant in all of the analyses.

# Preliminary Analysis: Do Multiple Politicized Identities Predict Collective Action?

Data screening was performed to ensure there were no violations of the assumptions. The descriptive statistics and bivariate correlations are presented in **Table 1**.

First, we performed a hierarchical multiple regression analyses to test whether identification with several politicized categories predicts collective action better than one salient identity and whether interaction terms should be included in a final model along with main effects<sup>1</sup> . Overall, the regression analyses indicated that multiple identities have additive positive effects on both types of collective action (on persuasive collective action, adjusted R<sup>2</sup> = 0.403, on confrontational: adjusted R<sup>2</sup> = 0.132) and that adding perceived compatibility and perceived legitimacy of protest significantly improved the explanatory power of both models, respectively, adjusted R<sup>2</sup> = 0.49, 1F(2,994) = 89.41, p < 0.001, and adjusted R<sup>2</sup> = 0.153. 1F(2,994) = 13.03. Details of the regression analyses are available from the corresponding author.

# Main Analyses: Do Compatibility and Legitimacy Mediate the Effects of Multiple Identities on Participants' Intentions to Engage in Persuasive and Confrontational Collective Action?

We tested a model in which perceived legitimacy and compatibility were considered as possible mediators of the effects of the three forms of identification on persuasive and confrontational action. Correlated error terms were allowed at each layer of the model. After the initial runs, we adjusted the models by setting two paths that had non-significant regression weights in the original models to zero, in particular the paths from the street movement identity to perceived compatibility (β = 0.03, p = 0.207) and from the Ukrainian identity to confrontational collective action (β = 0.01, p = 0.738). **Figure 1** shows the adjusted fitting model. The values for this final model fall within the cut-offs as advocated by Bentler and Bonett (1980) indicating good model's fit: χ 2 (2) = 1.594, p = 0.451, CFI = 1.000, RMSEA = 0.000 (confidence interval: Low = 0.000, High = 0.043), PCLOSE = 0.978, AIC = 67.594.

The final model shows that identification with the street movement was a significant direct predictor of both persuasive and confrontational action and that identification with the online protest movement was a direct predictor of only persuasive action. Ukrainian national identification was an indirect predictor of both forms through perceived legitimacy and compatibility.

# DISCUSSION

This study explored the mechanisms by which multiple identities predict collective action. First, we found empirical support to our hypothesis that collective action can be explained to a greater extent by accounting for multiple social identities with potentially aligned contents rather than by focusing on a singular salient category membership ('the more we stand for – the more we fight for'). Importantly, these relationships were found to be significant for both persuasive and confrontational forms of collective action.

The effects of identification with the online protest community are noteworthy. In particular, we found that both persuasive and confrontational collective action were predicted by identification with the online protest community due to increased perceptions of compatibility between the online and the street protest, but also due to the higher perception that protest is legitimate. These findings are intriguing as they contribute to the ongoing research on political participation through the Internet (e.g., Schumann and Klein, 2015; Thomas et al., 2015), which has been criticized as a low-cost and low-risk activism lacking commitment and social impact (e.g., Morozov, 2009; Gladwell, 2010). Our results indicate an alignment of identification with different groups, irrespective of the online-offline divide, and perceived compatibility between identifications and the perceived legitimacy of protest seems to equally and independently predict collective action.

In our case there was no evidence that online protest was seen to be a defective or unsatisfactory form of action even


OPC, online protest community; SM, street movement, all correlations p < 0.001.

<sup>1</sup>The preliminary analysis includedcentred interaction terms between each of the two identities (i.e., identification with the online protest community and with the street movement), and perceived compatibility, as well as perceived legitimacy of protest. In fact the only significant interaction term was in the model explaining persuasive collective action where the interaction of the effect of identification with the online protest community and perceived compatibility was significant: β = 0.08, p = 0.037. For reasons of parsimony and in view of the high power derived from the large sample size we do not consider these interaction effects in the main analyses.

alongside widely disseminated images of street protests that were globally distributed. It is important to bear in mind though that in Ukraine in 2014, as in other parts of the world, both online dissent and street protest were illegal. The Ukrainian government may have unwittingly increased the value of online dissent by banning it at the beginning of 2014, in the midst of the political crisis. Additionally, the online community of protestors offered a platform for the Ukrainian diaspora to become involved, whereas their participation in the street protests was logistically difficult if not entirely impossible. Future research may examine longitudinal changes in the relationships between multiple identities across various platforms of collective action, using individual-level analysis to track people's enduring participation in fighting for a common cause.

Secondly, our data revealed that, beyond the direct effect of multiple identities, the perceived compatibility between them adds to our understanding of people's engagement in collective action and explains the effects of identifications with the online community and with Ukraine. What "we stand for" and how we "stand for" the multiple communities we belong to (online, at the more abstract level of the national community), is of equal if not additional importance to our single memberships in any of them. Our study contributes to the theoretical discussion regarding the role of normative overlap between the agendas of different social groups in explaining cross-domain activism (e.g., Curtin et al., 2016; Louis et al., 2016) and long-term commitment to political causes (e.g., Smith et al., 2015b; Turner-Zwinkels et al., 2015). The results offer the interpretation that participation in collective action came to express national identification where Ukrainians saw protest as the right (legitimate) thing to do and where they perceived online and street protest to be compatible.

There was a seemingly paradoxical effect consistent with the perceived compatibility between street and online community identities mediating the effects of national identification on collective action but not the effects of street movement identification. In other words, the connection of the street Euromaidan movement with action was direct, so that supporters of the movement took to the streets. The connection of national identification with collective action, however, appeared to be indirect. In other words, the clear pathway between Ukrainian identification and collective action rested on perceiving the street movement and online movement were compatible with each other, because, we surmise that these two identities were themselves held to be valid expressions of Ukrainian identity (from the perspective of those supporters). We cannot know in retrospect whether that compatibility was a specific element that was relevant to the contested context of a nation that, at the time of data collection, was on the precipice of civil war or whether these patterns would obtain more broadly.

One possible way of thinking about these findings is that a capacity to form, synthesize, politicize and merge several opinion-based groups centered on short-/medium-term issues into a multi-goal campaign may serve as a key factor to understanding the processes behind coalition buildings and global activism. This also helps us understand the failures of mobilization: when networked campaigns use a vague idea-framing and related ideological noise that cannot justify involvement for a global cause, thus failing to bring people together. At the individual level, failure to cohere multiple group memberships into concerted collective action could be explained by exactly this lack of ideological overlap and missspecification of the identity-norm nexus (Smith et al., 2015b). Although the present study was not designed to explicitly measure a link between formation of identities with overlapping injunctive contents and coordinated collective action, we believe that the curvilinear nature of this relationship requires a further examination.

Finally, consistent with our expectations, we found that expression of multiple social identities through collective action was also explained by individual perceptions of legitimacy of protest. Specifically, our findings indicate that higher degrees of identification with all politicized identities led to increased perceptions of protest as a legitimate method for achieving social change, and thus to higher likelihood of engaging in persuasive and confrontational collective action. In fact, our theoretical analysis helped us to identify and test this intriguing puzzle within the context of Ukraine, in the immediate aftermath of the introduction of criminal penalties for political dissent. The revealed pattern is important as it suggests that recognizing both between-group and inter-personal variations in people's beliefs about protest (and incorporating the concept of perceived legitimacy of protest in collective action research) can help explain more general processes of choosing tactics from a spectrum of possibilities within a repertoire of contention. It is noteworthy to highlight that the effects of identifying with the street protest were only mediated by the perceptions of protest legitimacy and not by perceiving identity compatibility. Capturing people's perceptions of protest legitimacy is also a way of operationalizing people's engagement with the political opportunity structure, at times of political change and transition. This is much needed for developing a more dynamic theoretical model of the multiple links between identity and politics in constantly changing political environments.

Our findings raise other important questions: whether and under what conditions radicalization (confrontational political action) emerges from activism (non-violent political action)? To what extent do the tactics that one employs depend on political circumstances (e.g., legal criminalisation of dissent) and will variations in perceived legitimacy of protest produce similar patterns of collective behavior in both liberal and developing democracies? In other words, if variances in the perception of legitimacy of protest can help explain particular cases, can this conceptual approach generate testable models that hold across contexts? The answers are beyond the scope of this paper, but one factor may be due to the individual perception of political opportunities (e.g., Meyer, 2004) and, therefore, an elaborated conception of perceived legitimacy of protest that considers a broad range of conjunctural and issue-specific factors is recommended for future research.

To sum up, our findings support the idea that the expression of multiple politicized identities—their agency—can be understood to a greater extent when considering the political context and the rules of the game in which those identities are endorsed and internalized — that is, the surrounding ideological and political opportunity structure. However, it is important to advance our understanding of how various real or virtual communities, structured around non-contiguous spaces, may trigger confrontational (potentially radicalized) and persuasive collective behavior. Our models explained the latter to a greater extent, but not the former. Finally, we urge collective action research to continue to operationalize and test how the fluidity of the political opportunity structure affects the emergence of social and political identities, and the relationships of compatibility or opposition between these identities. A more complex framework capturing the diversity and multiplicity of identities (and relationships between them: such as perceived compatibility) as well as their relationship to the political context (the political background of legitimacy) will better equip us to understand and predict the paths to social change.

# Limitations

Reflecting on the external validity of our findings, we must exercise caution, due to the cross-sectional nature of our design, sample characteristics, and our use of self-report explicit measures of various politically sensitive issues. Therefore, although we obtained a large general community sample (in two languages), at a crucial time of the political crisis (shortly after the passing of laws that restricted people's right to protest), we cannot account for the potential selection bias in the sample, or for the powerful effect of 'history' happening at the time of the study. While we do not wish to assume causal links between, for example, perceiving protest as legitimate and engaging in certain types of collective action, we would still like to argue that it is important to capture the variation in people's beliefs about protest in a model predicting collective action. These variations will naturally be in tune with the changes in the political structure, and they are likely to have been particularly relevant for the Ukrainian setting. Further studies at different times in the development of a political crisis, and in contexts with variable degrees of democratization, will strengthen our empirical and theoretical ability to predict collective action.

Secondly, our results support the notion that the perceived compatibility between multiple identities is an independent predictor of collective action in addition to the combined effects of multiple politicized identities. This invites further refinements of the measures of compatibility and the three identification types. We assessed identification with three categories by using two different scales (i.e., item-based for national identification with Ukraine and the IOS pictorial measures for identifications with the online protest community and the street movement). This methodological discrepancy may account for the relative small covariance of these identifications. It is important to monitor how our understanding of the effects of multiple

identities on collective action may depend on the measurement type. At the same time, we would note that concerns about ecological validity should prevail over the exclusive reliance on conventional measures. It might have been much more intuitively easy for people to respond with a pictorial measure when thinking about their self-inclusion in communities that were new and emerging at the time of the study. At the same time, when assessing national identification, the more established measures are perhaps best to assess the depth and strength of people's group attachments.

Lastly, we operationalized perceived compatibility using several questions about the overlap between identity categories (i.e., referring to their ideological content). We did not explicitly measure the specific normative content and normative compatibility of the groups' political agendas. This measure might seem, on one hand, comparatively superficial. On the other hand, its predictive validity indicates that participants responded to these questions with the two communities (online and street protest) in mind. Once more, due to the emerging nature of these group identities and communities of protest, in the midst of the political instability and crisis, a more in-depth measure might have been both impractical and unnecessarily complicated. This leaves room for future research to test whether normative compatibility between multiple politicized identities explains other intergroup behaviors, beyond predicting collective action.

# CONCLUSION

Summing up, we propose that collective action in the 2014 Euromaidan protests can be interpreted as the congruent expression of multiple identities that are rendered ideologically compatible both in online settings and on the street. This study investigated multiple identities that are related to the specific political context of a country in transition to democracy, caught in months-long upheavals and street protests, at a time when online interactions allowed for increased transnational mobilization and involvement in politics. The questions were: how do people negotiate their identities with their country, the

# REFERENCES


online community of protesters and the street movement? Would these identities converge to support a concerted political agenda, thus increasing collective action intentions? Or would they be redundant in capturing people's feelings and engagement with the various groups? In addition, how do these multiple identities relate to the political opportunity structure where protest itself was classified as illegal by the government, in a country grappling with an emergent democratic culture? We found evidence that the more social groups people "stood for," the more they "fought" for their cause and that identifications predicted both forms of collective action to the degree that people saw the protest and the online movement as compatible with each other and believed protest to be legitimate.

We explained persuasive form of collective action to a greater extent compared to the confrontational form of action. Perhaps, negotiating multiple identities and looking for ideological alignment is a strategy that is more easily employed by those with moderate political agendas. Future research should explore how the dynamics between multiple identities (creating dissonance and lack of compatibility) might be employed to temper engagement in more confrontational or radical political actions.

# ETHICS STATEMENT

This study was carried out in accordance with the recommendations of Research Ethics Committee, University of Limerick with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Research Ethics Committee.

# AUTHOR CONTRIBUTIONS

This research is a part of a Ph.D. project of MC, developed under the supervision of AM in collaboration with CM. All authors contributed equally to the preparation of this manuscript.


perceptions in Ukraine. Group Process. Intergroup Relat. doi: 10.1177/ 1368430216682354


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Chayinska, Minescu and McGarty. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Bridges or Barriers? Conceptualization of the Role of Multiple Identity Gateway Groups in Intergroup Relations

#### Aharon Levy1,2 \*, Tamar Saguy<sup>1</sup> , Eran Halperin<sup>1</sup> and Martijn van Zomeren<sup>2</sup>

<sup>1</sup> The School of Psychology, Interdisciplinary Center Herzliya, Herzliya, Israel, <sup>2</sup> Heymans Institute for Psychological Research, University of Groningen, Groningen, Netherlands

The modern era of globalization has been accompanied by a massive growth in interconnections between groups, and has led to the sharing of multiple identities by individuals and groups. Following these developments, research has focused on the issue of multiple identities, and has shed important light on how individuals who hold these complex forms of identity feel and behave, and on the reactions they elicit from members of other groups. However, the potential of groups with such multiple identities (e.g., biracials, immigrants, etc.) to affect the intergroup relations between the groups that represent the respective sources of the different identities (e.g., Blacks and Whites, country of origin and country of residence, etc.) has not been examined to date. Accordingly, in this paper, we first systematically explore the potential of groups in which people identify with multiple social categories, or groups that are perceived as such by others, to play a role in intergroup dynamics. Next, we offer a theoretical framework outlining what functions groups of people with shared multiple identities may serve (as bridges or barriers) by proposing how their presence may facilitate or deteriorate intergroup relations. Finally, we present recent empirical research examining how groups of people with shared multiple identities can act as gateways and bridge the cleft between two separate groups that represent the respective sources of their different identities, and discuss the theoretical and practical implications for the field of intergroup relations.

Keywords: multiple identity, gateway groups, intergroup conflict, conflict resolution, dual identity, cross categorization, biracial, social identity complexity

# INTRODUCTION

Race, at least in the United States, has typically been treated as a dichotomy (e.g., Black or White), with individuals challenging this racial dichotomy being likely to become socially excluded or even penalized (Hickman, 1997; Davis, 2010; Khanna, 2010; Wagner et al., 2010). However, in recent years a clear shift seems to be taking place toward an increase in both the presence and influence of biracial identity. Indeed, over the past 15 years, the Black and White biracial population in the United States has tripled in size numbering over 2.5 million (U.S. Bureau of the Census, 2015, based on self-report), and the current estimate is that by 2050 one out of five Americans will be

#### Edited by:

Soledad de Lemus, University of Granada, Spain

#### Reviewed by:

Niklas K. Steffens, The University of Queensland, Australia Marion Chipeaux, Université de Genève, Switzerland Clara Kulich, Université de Genève, Switzerland

> \*Correspondence: Aharon Levy aharon.levy@post.idc.ac.il

#### Specialty section:

This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

> Received: 09 November 2016 Accepted: 13 June 2017 Published: 29 June 2017

#### Citation:

Levy A, Saguy T, Halperin E and van Zomeren M (2017) Bridges or Barriers? Conceptualization of the Role of Multiple Identity Gateway Groups in Intergroup Relations. Front. Psychol. 8:1097. doi: 10.3389/fpsyg.2017.01097

of mixed-race (Lee and Bean, 2004). Indeed, recent research even suggests the rise of a new multi-racial identity that is replacing the monolithic race identities in the United States census (Roth, 2005; Davenport, 2016). The million-dollar question is how these identity developments will affect the existing intergroup relations between Blacks and Whites in the United States, and how biracials will be perceived by others, and perceive themselves, now that they are becoming more and more a prevalent in social reality.

This million-dollar question, however, is not limited solely to the realm of race. Just as biracials are strategically situated between Blacks and Whites on a structural level of analysis, so are, for example, the Bosnian citizens of Serbia situated in the same kind of social overlap between Serbia and Bosnia. Thus, this group which has a mixture of Bosnian and Serbian national identities, may have similar influence on the international social dynamics between the two nations in conflict. Moreover, multiple identities exist not only in the same dimension (biracial, dual nationality), but also as a result of cross cutting identities such as nationality and religion. For example, the Muslim community in the Kashmir region of India share their national identity with India while at the same time share their religious identity with Pakistan, and as such may be able to affect the relations between these two conflicting countries as well. In fact, similar social structures can be found in many other contexts and levels of analysis as well (which will be elaborated on below), raising the same question of how the emergence of such a group with a mixed identity will influence intergroup relations, and what this implies for members of groups with such multiple identities.

This question of how multiple social identities affect intergroup relations remains largely unanswered by contemporary social psychology, mainly because research seems to have focused on how individuals who hold these complex forms of categorization feel and behave (Baysu et al., 2011; Gocłowska and Crisp, 2014), and on the reactions they elicit from members of the dominant group (González and Brown, 2006; Bodenhausen, 2010; Rodeheffer et al., 2012; Scheepers et al., 2014; Urbiola et al., 2017). This research has found for example, that multiple identification among minority groups (e.g., both German and Turkish) typically relates to better well-being (Sam and Berry, 2010; Nguyen and Benet-Martínez, 2013), and that for majority group members, endorsing the existence of multiple identities (as reflected in the notion of multiculturalism) promotes more positive outgroup attitudes (Wolsko et al., 2000; Scheepers et al., 2014).

However, despite the importance of these findings, they leave several key questions unanswered. More specifically, we seem to know relatively little about the potential of groups equipped with multiple identities to affect the intergroup relations between the groups that represent the respective sources of their multiple identities. Given their shared identity with both groups, groups of people with shared multiple identities can potentially bridge the cleft between two (or more) otherwise separate groups (e.g., if someone is your sister's sister than she is probably your sister as well). At the same time, under certain conditions these groups of people with shared multiple identities can also pose barriers to conflict resolution by raising issues of trust and betrayal. The main goals of this paper are therefore to outline the notion of such gateway groups (GGs) amidst intergroup conflict, and to situate it in a novel theoretical framework. This framework will outline what functions GGs may serve (as bridges or barriers), and how these functions may improve or deteriorate intergroup conflict both among those who perceive these groups from the outside, as well as among the members of the GGs themselves. Where relevant, we will further review data supporting elements of the theoretical framework proposed.

Of course, this is not just important theoretically, but also for practical reasons. When considering the potential positive effect of groups of people with shared multiple identities, Turkish immigrants in Germany for example, may be able to impact the relations between Turks and Germans in general by virtue of being perceived as identifying with both these entities. Similarly, biracials in the United States might have the ability to bridge relations between Blacks and Whites, and Arab citizens of Israel might likewise be able to influence the relations between Israel and Palestine. Thus, our approach focuses on the psychological perceptions, experience, and functions of GGs, both from the perspective of the groups in conflict and of the GG members themselves.

Despite the potential of GGs for improving intergroup relations, GGs might also serve as barriers as much as bridges, for instance when GG members feel their multiple identities to be in conflict with each other, or when groups in conflict do not trust GG members because they fear mixed and even shifting loyalties. This possible complexity may lead to a decrease in GG members' dual identification. At worst, the GG might even be perceived as a "fifth column" that undermines the position of either group in the broader conflict. This may have negative consequences for intergroup conflict in general, and presumably also for GG members' well-being.

Therefore, in this paper we will map the existing findings regarding cases in which the GGs serve their bridging function (which improves intergroup relations and has positive psychological consequences for all involved), and cases in which they serve their barrier function (which deteriorates intergroup relations and has negative psychological consequences for all involved). Below, we first define and conceptualize GGs, and review recent evidence regarding the positive and negative potential of such groups when their members perceive themselves as having multiple identities. Then, we present recent data for the positive potential of such groups being perceived as having multiple identities and consider possible setbacks that this may have. Finally, we discuss functions of GGs that have not been examined and develop relevant hypotheses.

# DEFINITION AND CONCEPTUALIZATION

We define GGs as groups characterized by unique social categorizations that enables them to be categorized as and identified with more than one group within the context of

intergroup relations. Importantly, the categorization of a group as a GG can stem from the way others perceive this group, or by how the group members experience their own group, or both. Due to the fractal nature of social categorization that can be analyzed on several different levels of analysis, there can be many different types of GGs. As mentioned above, GGs can exist on a national level (e.g., Israeli Arabs) and on a racial level (e.g., biracials). Additionally, GGs can be found on a larger global cultural level, countries such as Turkey or Albania can mediate between the western world, and the Muslim world which they are both identified with (Keyman, 2007); and on a smaller scale situated in the midst of a specific ethnic group (e.g., between Ashkenazi and Sephardic Jews; see Cohen et al., 2004).

# GG AND EXISTING MULTIPLE IDENTITY LITERATURE

The GG concept fits smoothly within the existing literature on multiple identities, although a significant distinction that can be made here is that of perspective. While existing frameworks are mainly rooted in the perspective of the dual or multiple identifier (i.e., focusing on how individuals cope with multiple identities), the GG concept can also refer to the perspective of social groups that view the GG from the outside. For instance, if the biracial community is perceived by the White community and the Black community as biracial, then the biracial GG may have an impact on this intergroup relation even if the biracial individuals do not necessarily subjectively identify with both groups simultaneously (and vice versa). This conceptualization enables us to study both perspectives systematically. Thus, we introduce a clear distinction between the type of perspective on the groups of people with shared multiple identities (i.e., does existing research deal with how these groups are perceived by others, or does it deal with how the members of these groups perceive themselves).

Furthermore, multiple identities can have different functions in the contexts of intergroup relations (i.e., as bridges or barriers) that have either positive or negative effects on the relations between the groups that represent the sources of those identities. The suggested framework is a 2 × 2 matrix based on the perspective on the groups of people with shared multiple identities (from within or without), and on the function of the groups of people with shared multiple identities (bridge or barrier)<sup>1</sup> .

# POSITIVE OUTCOME OF GROUP MEMBERS PERCEIVING THEMSELVES AS HAVING MULTIPLE IDENTITIES

Several studies have examined the positive outcomes that stem from individuals' identification with multiple social groups. Theory and research on the so-called social cure contends that identifying with multiple social groups is directly linked to improved health and well-being (Jetten et al., 2012, 2014; Steffens et al., 2016a,b). However, since our focus in this paper is on the function of multiple identities in the context of intergroup relations, we will focus specifically on the possible outcomes that stem from multiple identities and affect intergroup relations. In this section, we will briefly review relevant work done on dual identity, common ingroup identity, cross-categorization, and social identity complexity, and describe how these lines of thought explain the positive outcome of multiple identities in intergroup relations.

# Dual Identity

Arguably the most relevant line of work for the present purposes is theory and research on dual identity (Brown and Hewstone, 2005; Dovidio et al., 2009). Dual identity is a simultaneous identification with a distinct subgroup and a common superordinate group (e.g., the Latino minority in the United States that identifies simultaneously as Latino and as American; Dovidio et al., 2009). The central benefit of dual identification lies in allowing minority group members to feel connected to the dominant majority group, while maintaining their distinctiveness as a separate group simultaneously. In several empirical studies, this dual identification was found to be associated with numerous constructive outcomes including the well-being of the dual identifiers (Sam and Berry, 2010; Nguyen and Benet-Martínez, 2013), and inhibition of extremism (Simon and Ruhs, 2008). Such outcomes are typically explained by the notion that they fulfill individuals' need to identify with their original subgroup while still feeling connected to a larger whole (Brewer, 1991).

Additionally, following the logic of the common ingroup identity model (Gaertner et al., 1993), dual identities might signal to the respective groups that a superordinate identity, incorporating both groups, is possible. Such a common identity has been found as a useful tool for the reduction of intergroup prejudice. While a common ingroup identity might also have negative effects on disadvantaged groups, and seems very hard to maintain in the context of intergroup conflict (Saguy et al., 2009), the fleshing out of dual identity aspects can help maintain such a superordinate identity while counteracting the possible negative side effects (Dovidio et al., 2009). Thus, dual identity can induce the positive impact of a common group identity despite the described difficulties. For example, in a study by Hornsey and Hogg (2000), when university students were primed with both their superordinate identity (university students), and their subordinate identity (humanities or mathscience students) simultaneously, they displayed lower levels of bias toward their outgroup compared to when they were primed only with their superordinate or their subordinate identities separately. As we will later elaborate, the notion of GGs resonates with this dual identity construct in that the presence of a multiple identity GG can be both a reminder for its counterparts of the similarity between the two separate groups, while at the same time help maintain each group's distinction.

<sup>1</sup>This structure also corresponds with other reviews dealing with different aspect of groups with multiple identities, such as the work done by Kang and Bodenhausen on social cognition.

An interesting example of the potential of a dual identity when paired with a common ingroup identity can be found in a few recent studies examining collective action tendencies among disadvantaged-group members in the United States (Blacks and Latinos, see for example Ufkes et al., 2016). These studies have found that increasing only the salience of a common United States identity among Blacks and Latinos reduced intergroup bias but also resulted in lower collective action intentions, and thus led to the maintenance of existing social inequality. Increasing salience of dual identity, however, did not decrease collective action intentions, and was found to both reduce intergroup bias as well as challenge existing social inequality.

# Social Categorization Overlap

Another very rich source of information on the positive implications of individuals and groups perceiving themselves as having multiple identities is in the field of social category overlap. The work on cross-categorization, for example, addresses the multiple identities an individual holds, and their potential overlap. According to research on cross-categorization, given that individuals are members of several groups simultaneously (e.g., both Black and female), members of an outgroup on one dimension may be evaluated more positively if they are also ingroup members on another dimension (e.g., a Black women evaluating a White woman).This crossing of categories was found to reduce intergroup prejudice and discrimination among those holding the multiple identities (Brewer and Campbell, 1976; Deschamps and Doise, 1978; Migdal et al., 1998; Crisp and Hewstone, 1999; Hutter and Crisp, 2005).

Similarly, social identity complexity also deals with the overlap between different social identities (specifically the extent of such overlap) and posits that raising awareness to the partiality of overlap between social identities decreases the salience of social categories, and in turn raises tolerance for outgroups in general (Roccas and Brewer, 2002; Brewer and Pierce, 2005; Brewer, 2010; Brankovic, 2016 ´ ; Sønderlund et al., 2017). For example, in a study by Vasiljevic and Crisp (2013), participants who were primed with a multiple social categorization mindset, increased cognitive flexibility, displayed lowered prejudice toward a multitude of outgroups, fostered egalitarian values, and enhanced their trust toward outgroups. Here too, the GG notion can be seen as building on this existing construct, as the GGs are inherently characterized by identity overlap as we will describe below.

In sum, the literature described above, which is a culmination of decades of studying multiple identities in intergroup contexts, provides clear indication that when it comes to intergroup relations there are several positive functions multiple identification may have on those who hold it. Whether in facilitating a common superordinate identity, increasing cognitive flexibility, or reducing intergroup prejudice, both the multiple identifiers and their counterparts stand to gain from these groups identifying themselves with more than one social category. Having said that, we also believe that it is important to consider the potentially negative side of GGs' multiple identification, and asses the possible backlash of groups perceiving themselves as having multiple identities in the context of intergroup relations.

# NEGATIVE OUTCOME OF GROUP MEMBERS PERCEIVING THEMSELVES AS HAVING MULTIPLE IDENTITIES

Despite the promising potential of GGs in terms of group members perceiving themselves as having multiple identities (as described above), there is an important reservation to make in this regard, and that is the possible backlash the fleshing out of such multiple identities may have in the context of intergroup relations. In terms of individuals perceiving themselves as having multiple identities, a negative side effect that can possibly result from stressing the multiple identification of a group, can be the decreased well-being amongst group members themselves (Gaither et al., 2013). Some of the research in the field of dual and multiple identities have found that there can be a downside to holding multiple identities, especially in the context of intergroup conflict.

For example, in cases where intergroup relations are tense, multiple identifiers are sometimes forced to distance themselves from one of their ingroups, and even partake in ingroup derogation in the attempt to overcome disadvantage in cases of social inequality (Verkuyten and Reijerse, 2008; Derks et al., 2015, 2016; Kulich et al., 2015). Moreover, under some conditions, biracial individuals may feel anxiety that stems from interracial encounters while encountering both Blacks and Whites (Gaither et al., 2013).

Furthermore, when the identities multiple identifiers hold are perceived as incompatible, it may foster controversial or even destructive forms of political radicalism (Simon et al., 2013). Finally, research on dual identifiers in the context of intergroup conflict has shown, that the tension between the conflicting sources of the dual identity causes the dual identifiers to be marginalized by both counterparts. This double marginality may lead to the general exclusion of these groups of people with shared multiple identities, weaken their collective infrastructure, and at times even bring about collective identity crisis (Al-Haj, 2000; Berdahl and Moore, 2006; Lowrance, 2006; Pinson, 2008). Another relevant notion in this regard is that of intersectionality, which claims that individuals' identification with more than one discriminated group in an intergroup context may lead to marginalization of wrongdoings toward such a group, even compared to other discriminated groups that do not have multiple identities (Crenshaw, 1991). Although this is not a direct negative outcome of holding multiple identities, it only applies to those who hold more than one discriminated social identity, and thus also corresponds with the framework we are suggesting.

In sum, despite the positive potential of GGs holding multiple identities in the context of intergroup relations, this element also has the potential of putting such GGs between the proverbial identity rock and hard place. This suggests that perceiving oneself as having multiple identities might not always be so beneficial, and this may also be the case when being perceived as such by external groups.

# POSITIVE OUTCOME OF GROUPS BEING PERCEIVED AS HAVING MULTIPLE IDENTITIES

As important as it is to discover the outcomes of individuals' identification with multiple groups in the context of intergroup relations, it is equally important to understand the potential of the presence of GGs to affect the intergroup relations between the groups that represent the respective sources of their multiple identities. Even though, to our knowledge, no prior work has taken the approach we propose here with respect to the potential role of GGs in intergroup conflicts, the work described above on social categorization processes lays the groundwork and intrinsically corresponds with the GG notion we will describe below.

For instance, the presence of a multiple identity GG can be both a reminder for its counterparts of the similarity between the two separate groups, and a signifier of the group's distinction (Saguy et al., 2009). As such, multiple identity GGs can be utilized in order to highlight dual identities and foster common ingroup identities among those who perceive the GG as holding multiple identities. Additionally, since GGs can be seen as social groups in which the ingroup identity overlaps with the outgroup identity, the research on cross-categorization is highly relevant as well. The effects of intergroup prejudice reduction that cross categorization was found to have among group members that perceive their own identities as crisscrossing, may also take place in scenarios involving others that perceive the GGs as holding cross-cutting identities. Given that little is known about such scenarios, we present our line of thought below, and provide some first evidence of the positive potential that the presence of GGs can have in intergroup conflict.

# Extending the Existing Research

Besides the direct links between the GG notion and existing literature described above, in which this notion can be seen as an integral part of existing work, the GG notion can theoretically broaden the existing scope of the multiple identity literature as well. First, while dual identity has primarily been linked with hierarchically nested identities in the form of a superordinate (typically majority) and the subgroup identity (typically minority group. e.g., Turkish immigrants in Germany that are nested in the superordinate German identity while maintaining a separate Turkish identity), GGs also incorporates situations in which identities are not nested. For example, the biracial community in the United States. Does not necessarily have a clear hierarchically nested structure between its White and Black identity (i.e., neither racial identity encompasses the other). Thus, all dual identifiers can be seen as members of a GGs but not all GGs can be explained with the notion of dual identity.

Second, dual identification scenarios usually include three different social agents: two distinct social groups, and the dualidentity group. However, to our knowledge the existing literature only address two of these social agents: one of the two distinct groups, and the dual identifiers. For example, in the case of the Mexican minority in the United States. Most of the literature

addresses either the minority itself, or the White Americans and their interaction with the Mexican minority, but the Mexicans in Mexico are not addressed. The broader notion of a GG enables the incorporation of several different relevant groups for a more inclusive, realistic, and complex understanding of intergroup dynamics. For example, due to the fact that the GG viewpoint accounts for all three parties in this scenario, it would enable the explication of phenomena such as United States presidential candidates courting Mexican officials during a United States presidential campaign, as well as a more accurate modeling of the intergroup dynamic between Mexicans and Americans inside and outside of the United States.

Finally, whereas cross categorization deals with meshing identities from different dimensions (e.g., race and sex), GGs create an overlap between identities from a single dimension (e.g., the overlap of two different racial groups). Thus, the GG fills an important gap not fully covered by cross categorization, of the identity overlap between identities from the same dimension such as national identities overlapping in immigrant communities, or racial identities overlapping among biracials. Moreover, cross categorization usually requires a positive overlap between two identities in order to take effect (see example A in **Figure 1**). However, from an external perspective the existence of a GG might suffice in order to achieve the positive effects of cross categorization in scenarios not deemed eligible in the past, by creating an identity overlap outside of the conflicting parties and inside the GG itself (see example B in **Figure 1**). Take for instance, the intergroup relations between religious people and gay people. These two communities conceptually do not overlap, and therefore are not a natural candidate for cross-categorization. Nonetheless, the existence of a religious and gay community GG might be able to symbolize the necessary overlap needed to induce the positive cross categorization effect for both of these respective groups.

Taken together, the existing research on multiple identification can be used to predict that the presence of a GG may lead to positive outgroup orientations also among external groups that perceive it as such. Moreover, this may also enable the broadening of the dual identity prospect as described above, and extend the explanatory scope of the multiple identities literature as well. Based on this assumption we designed several studies in order to examine the potential of multiple identity GGs to improve

intergroup relations between the external groups representing the respective sources of the multiple GG identities. Thus, we studied how others perceive those in a multiple identity GG, and how that perception influences intergroup relations (Levy et al., 2017, under review).

# Perceived GG Empirical Studies Overview

In one recent paper (Levy et al., 2017), we first conducted a pilot study designed to simply compare the perceptions and attitudes toward an outgroup to the perceptions and attitudes toward a GG. Studies have found that under certain conditions biracials maybe simply perceived as Black, i.e., distinct outgroup, and not in the complex manner we have described (Peery and Bodenhausen, 2008; Rodeheffer et al., 2012). Accordingly, this study was meant to examine the premise that GGs with multiple identities are indeed perceived differently and more positively (or less negatively) than the distinct outgroup. The study was conducted in the Israeli-Palestinian conflict context among Israeli participants (ingroup), with the Palestinians as the outgroup, and the Arab citizens of Israel as the GG. This is a good example of a multiple identity GG because Arab citizens of Israel are affiliated with both the Israeli Jewish group with whom they share their citizenship as well as with the Palestinian group residing in the Westbank and Gaza, with whom they share their national identity, and thus can serve as a GG. As expected, the GG was evaluated more positively than the outgroup in almost every single indicator including: group stereotypes, perceived similarity with the ingroup, contact motivation with group members, support for aggression against the group, and feelings of anger and hate toward the group. These findings confirmed our assumption regarding the positive potential of GGs for improving intergroup relations. However, these findings may be restricted to this particular context, and needed replication (to safeguard external validity) and corroboration in terms of internal validity (through an experimental approach).

Therefore, the next study in this line of research was a correlational study meant to examine the correlation between the perception of the GG as holding multiple identities, and positive attitudes between the groups reflected in the GG. Based on the GG hypothesis, the positive attitudes between the groups that made up the multiple identities should be higher when they indeed viewed the intermediate GG as having multiple identities. To facilitate external validity, the correlational study was run in two separate contexts: The Israeli Palestinian context with the Arab citizens in Israel as the GG (similar to the pilot study), and the religious factions in the Israeli context with the Liberal Religious Jewish community as the GG<sup>2</sup> .

In line with the GG hypothesis, and in both contexts, the more participants viewed the intermediate group as having a dual identity, the more positive their attitudes toward the outgroup became<sup>3</sup> . Given these findings, we proceeded to test the GG hypothesis experimentally (to increase internal validity) and assess improvement of intergroup relations through an actual resource allocation task, which is a behavioral measure.

Specifically, we conducted two experiments. The first experimental study aimed to test whether the presence of a GG that clearly encompasses multiple identities, would improve intergroup behavior under highly controlled settings. On-line participants were first assigned to artificially created groups, based on personal inconsequential preferences (Tajfel, 1978), and the key outcome was the amount of resources they allocated to the outgroup vs. the ingroup. In the control condition, the groups were created in a dichotomous manner, reflecting a more traditional two-group context. In the experimental condition, the groups were created such that there was an ingroup, an outgroup, and a GG that shared attributes with both the ingroup and the outgroup, and was thus perceived as having a dual identity. According to the GG hypothesis, and to the findings from the Pilot and correlational studies, the perception of multiple identities (i.e., in the experimental condition) should improve intergroup attitudes and behavior, as compared to a control condition. The results of this study supported this prediction: The presence of a GG that encompasses multiple identities led to more positive intergroup attitudes and behavior. Participants in the GG condition, compared to those in the control condition, allocated more resources to the outgroup, had greater contact motivation, and showed higher tendency for equal division and a lower tendency for complete discrimination.

The second experimental study aimed to replicate the first one while adding two additional elements: First, the experiment was carried out in small groups as opposed to individually<sup>4</sup> . Second, this study was performed in the lab rather than on-line so that we would have better control over the participants' environment and thus a better ability to simulate the dual identity condition. The second experimental study replicated the results of the first in a more meaningful and interactive context by showing that the presence of a GG leads to more positive intergroup attitudes and behavior. As in the previous study, and in line with the GG hypothesis, participants in the experimental GG condition collectively allocated more resources to the outgroup, had a higher tendency for equal division, a lower tendency for complete discrimination, and showed greater contact motivation.

However, these experiments used relatively artificial groups, which makes generalization to the real world somewhat difficult.

<sup>2</sup>The Liberal Religious Jewish community in Israel represents a group that is situated in between two separate, often conflicting, categories: Secular Jews, who are a relatively standard westernized secular society, and Ultra-Orthodox Jews who are a zealous religious society which adamantly separates itself from the nonreligious Jewish world in all walks of life. On the one hand, Liberal Religious Jews still abide to Jewish religious law but are, on the other hand, immersed in their local secular cultures as well. This enables them to be identified with both Secular and Orthodox societies, and possibly act as a gateway between them. The participants

in the study were Secular (ingroup), and the outgroup was the Ultra-Orthodox community.

<sup>3</sup>The attitudes toward the outgroup in this study were measured by contact motivation with the outgroup, general feelings toward the outgroup, and also a hypothetical resource allocation between the ingroup and the outgroup.

<sup>4</sup> Studies have shown that intergroup interactions are generally more competitive than interpersonal interactions (Insko et al., 1992; Wildschut et al., 2003). This suggested that allowing individuals to make allocation decisions in groups would put them in a stronger intergroup conflict situation thus enabling us to increase conflict intensity without losing experimental control.


TABLE 1 | Empirical studies examining the positive effect of the presence of a perceived multiple identity GG on intergroup relations between its external counterparts (Levy et al., 2017, under review).

Therefore, we followed-up these studies with another set of two quasi-experiments that tapped into the issue of racism in the United States, and the presence of the biracial GG (Levy et al., under review). The prediction here was that if the presence of a GG were to have the same positive influence in the real world scenario as it did in the artificial group setting, it should reduce prejudice in the form of symbolic racism. To test this hypothesis, we presented the participants with sets of photographs of individuals from different races in the pretext of a memory exercise. In the control condition participants were presented with photographs of Black and White individuals, and in the experimental condition participants were presented with the same photographs but with the addition of biracial individuals' photographs as well. As expected, the results showed that the presence of a biracial GG significantly diminished racist perceptions, and also improved the attitudes toward outgroup victims of current intergroup conflict events<sup>5</sup> .

In a final study, we directly manipulated the level of perceived GG multiple identity in the Israeli-Palestinian context, by providing the Israeli participants with survey data showing that the majority of the Arab Israeli GG members saw themselves both as Israelis and Palestinians and did not find a contradiction in their dual identification. Then, we examined the effects of that manipulation on behavior and attitudes toward the Palestinian outgroup compared to participants in an empty control condition. Furthermore, this study explored mediating variables in order to try and shed some light on the underlying mechanism of the GG effects.

In line with previous studies, the results showed that participants in the experimental dual identity condition allocated more resources to the outgroup, and they also displayed decreased support for aggressive policies toward the outgroup. Importantly, this study also suggested initial evidence for underlying psychological mechanisms, as the presence of a dual identity led to reduced negative stereotyping of the dual identity group, reduced ingroup identification, and reduced anger toward the outgroup which all in turn predicted improved outgroup orientations. These findings are in line with the GG hypothesis, and suggest that perceiving an intermediary group as dually identified with both the ingroup and the outgroup should have positive effects on intergroup attitudes and behavior (See **Table 1**).

In sum, the studies described above are the first studies to provide empirical evidence for the positive potential multiple identity GGs can have on the external groups that perceive them as such. These studies show that across several different contexts the presence of a group with multiple identities led the groups that represent the sources of these multiple identities to harbor more positive attitudes toward one another and to display more positive intergroup behavioral patterns. These findings fit well with the positive effects of having multiple identities, as reviewed in the previous section, and suggest that there is much promise in the notion that GGs have a role to play when it comes to improving intergroup relations. However, here too it is important to consider the potentially negative side effects of GGs being perceived as having multiple identities in the context of intergroup relations.

# NEGATIVE OUTCOME OF GROUPS BEING PERCEIVED AS HAVING MULTIPLE IDENTITIES

Based on the theories and empirical evidence mentioned above, the positive potential of externally perceiving GGs as having multiple identities in intergroup relations is evident. On the other hand, it is likely that stressing the connection between a GG and the outgroup can easily become detrimental, especially

<sup>5</sup>As opposed to the previous studies in which the described effects were found across the board, in this study, participants that were not prone to racism (low SDO) did not differ as a function of whether biracials were present. However, among participants prone to racism (high SDO) we found a replication of the previous GG studies in which the presence of a biracial GG significantly diminished racist perceptions, and also improved the attitudes toward outgroup victims of current intergroup conflict events.

in the context of severe intergroup conflict. Other than the simple animosity that can arise from association with a negatively perceived outgroup, the link between the multiple identity GG and the proximal ingroup, which has so far been stressed as a positive attribute, might sometimes backfire and result in the multiple identity GG posing barriers to conflict resolution. Nonetheless, despite the intuitiveness of this possible backlash, to our knowledge there is no empirical work to date that has examined this negative aspect of groups being perceived as having multiple identities.

We therefore extend our line of thought here to suggest two possible ways in which GGs can be perceived as barriers rather than bridges. First, this is likely the case where those who perceive GG members to have a multiple identity expect them to side with the ingroup. Although this may seem possible at first sight, it implies that any perceived violation of that expectation of ingroup support may lead external groups to treat GG members with strong suspicion, and as potential traitors. Furthermore, while an outgroup member acting on behalf of the outgroup is predictable, for an ingroup member to act on behalf of the outgroup in the context of intergroup conflict is considered treason, and is perhaps one of the most socially deplorable acts possible. This notion corresponds with the literature on the 'black sheep effect' (Marques et al., 1988; Pinto et al., 2010) in which attempts by ingroup members that are perceived as a deviation from group norms are judged and punished even more harshly than similar actions originating from outgroup members. Second, in cases where those who perceive GG members to have multiple identities and expect them to side with the outgroup, such perceivers are likely to associate all the negative attributes linked to the outgroup with the GG as well. This perception is bound to frame the GG as a threat and simply amplify the existing conflict tension and animosity.

In sum, while there is no existing empirical evidence regarding the possible back-lash of GGs being perceived as holding multiple identities, it is likely to assume that such an outcome is possible under certain conditions, and depending on different possible affiliations of the GG.

# POSSIBLE EXPLANATORY MECHANISMS

So far, we have discussed the possible impact GGs might have on intergroup relations. However, we have not addressed the issue of why they have such impact. Several studies offer some insight into possible mechanisms. For example, Gocłowska and Crisp (2014) have examined three necessary conditions needed for dual identity to be able to help foster creativity among those who hold such an identity: (1) The dual identifiers need to develop a deep relationship with the two (or more) groups that they belong to, (2) the dual identifiers need to undergo the process of adaptation to living and functioning in a new group, but at the same time remain identified with their original culture, and (3) the dual identifiers need to experience some distance and dissonance between their host and home cultures. While this research focused on creativity which is not necessarily directly linked to intergroup relations, such variables may play a role in the intergroup context as well.

For instance, in terms of relationships with and adaptation to the different groups (condition 1&2 above), Hornsey and Hogg (2000) found that the salience of the different groups representing the multiple identities can affect intergroup bias, and that members of groups with shared multiple identities for whom both categories are salient exhibit lower levels of intergroup bias compared to those for whom only one of the groups is salient. In terms of distance and dissonance between identities (condition 3 above), research by Simon et al. (2013) has examined the aspect of dissonance between identities, or more specifically incompatibility of identities, and has found that incompatible identities among dual identifying immigrants in Germany led to elevated sympathy for political radical action. On the other hand, research by Chayinska et al. (2017) found that compatibility between multiple identities encourages and legitimizes collective action. Another element related to the dissonance between identities that may have intergroup implications, is the level of projection of a single identity on the superordinate inclusive identity which the GG are associated with. Based on the work done on the projection model (Mummendey and Wenzel, 1999; Kessler et al., 2010) the more the multiple identity GG resembles the identity projected on the superordinate identity, the better its chances of evading prejudice and negative emotions.

Additional aspects that have been found to influence the impact of multiple identification in the intergroup context are group status, size, and threat. Some of the positive effects associated with multiple identity in the intergroup context, such as intergroup bias reduction, have been found to only take place among minority groups (González and Brown, 2006), and in contrast, research in a different context has shown that dual identification was most efficient in reducing intergroup bias among high status dual identifying groups (Dovidio et al., 2009). Finally, research by Baysu et al. (2011) has found that in the presence of identity threat, dual identity can be detrimental, but in the absence of such threat dual identity is preferable to any other form of minority identification such as assimilation or separation.

In terms of the impact that multiple identity might have on external groups that perceive a GG as having multiple identities, the recent empirical studies mentioned above have begun to shed some light on the possible mechanism at hand. These studies have found that the presence of a dual identity led to reduced negative stereotyping of the dual identity group, as well as reduced ingroup identification (Levy et al., 2017). Based on these findings, it seems that the impact of the presence of multiple identity GGs on groups that perceive them as such, may be mediated by fostering a more complex perception of social categorization in general.

In sum, it seems that several different mechanisms can explain the potential impact of GGs on intergroup relations. These mechanisms include variables such as: Compatibility or similarity between the multiple identities held by the GG; the type of relationship the GG has with its counterparts; the status and size of the GG compared to its counterparts; the levels of threat felt by and from the GG; the social identification of both the GG and

its counterparts; and the manner in which the GG is perceived by the groups it interacts with. While the existing research refers to most of these variables, the literature still lacks a clear model that factors in these elements, and enables a clear prediction of a GG's impact in different scenarios. Below we will offer a few options of future research that should assist in the formulation of such a model.

# QUESTIONS FOR FUTURE RESEARCH

Based on the framework described above, it is evident that research has found both negative and positive impacts of multiple identity GGs in the context of intergroup relations. However, it is mostly unclear when one might expect GGs to function as bridges or barriers that improve or worsen intergroup relations. The explanatory mechanisms mentioned in the previous section provide initial explanations as to how GGs might affect intergroup relations, but do not provide a substantial way of predicting when the presence of a GG will be beneficial, and when it might backfire and deteriorate intergroup relations. For us, this question is one of the most important ones that future research should address.

So far, the literature offers mixed results, such as in the case of the projection model (Mummendey and Wenzel, 1999) and subgroups model (Hornsey and Hogg, 2000). Following these two lines of thought, when one of the identities at hand significantly differs from the superordinate identity, and both identities are made salient, the work done by Hornsey and Hogg would probably predict lower levels of prejudice, while the projection model (Mummendey and Wenzel, 1999) would predict the contrary. Similar contradictory findings appear in the research on the effects of dual identity's interaction with group size and status mentioned above. While González and Brown (2006), found their effect mainly among dual identifying minority groups, Dovidio et al. (2009) found similar effects mainly among high status dual identifying groups. This suggests that there might be a more significant predictor of the type of impact GG have on intergroup relations, or alternatively, that a combination of several variables add up to affect the impact of GGs either positively or negatively.

Accordingly, future research should explore the elements that have been found as influential in affecting the type of impact GGs have on intergroup relations. For instance, status and size of the GG has been found to interact with the GG's influence on intergroup bias reduction, but the manner in which these variables interact is still unclear (González and Brown, 2006; Dovidio et al., 2009). Therefore, future research should focus on GG status and size as possible moderators and examine what conditions are necessary in this regard for intergroup bias reduction. Additionally, based on the conceptual work described above it is likely that sense of threat from the GG might inhibit the GG's ability to facilitate conflict resolution. Future research dealing with multiple identities and GGs should test for threat as a possible mediator, and check if being threatened by the GG diminishes its positive impact on intergroup relations. Research examining variation in such variables should enable the construction of a detailed model describing the conditions for the different possible outcomes, and would greatly improve the ability to predict the optimal way to manage such identities in the context of intergroup dynamics. Moreover, such a model would also enable the development of practical implementations of the GG potential<sup>6</sup> .

Besides the specific variables described above there are also broader questions related to the GG prospect that call for additional research. As we have proposed, at this point in time it seems best to approach the issue of GGs while maintaining the distinction between how such groups identify themselves, and how these groups are perceived by others (Kang and Bodenhausen, 2015). While the work done in the realm of how these multiple identities affect the GG themselves is more substantial, research on how GG that hold multiple identities affect the other groups involved has only just began scratching the surface. Moreover, as a result of the distinction between how individuals in GGs experience their identity, and how these groups are perceived by others, there is the possibility that these two perspectives might not always be aligned. It is likely that when both the GG itself and its counterparts all perceive the GG as holding multiple identities, then the presence of the GG will impact intergroup relations as described above. However, it is not clear what would happen in cases in which the GG is perceived as holding multiple identities, but the GG itself only identifies with one group. Accordingly, future research should examine the outcomes of such dissonance, both on the GGs and on their counterparts as well.

Due to the lack of empirical research on the possible negative outcomes in the context of GGs being perceived as having multiple identities by others, future research should start studying this gap in the literature. As described above, the expectation of GG members to side solely with the ingroup, might perceive the GG's interaction with the outgroup as an act of betrayal. Moreover, affiliation of the GG with the outgroup coupled with mistrust are likely to have a negative intergroup impact as well. Accordingly, future research should examine expectation of GG loyalty, as well as elements of trust and threat, in order to account for possible GG backlash effects.

An additional aspect that may play a significant part in the integration of multiple identities into the intergroup relations framework is the GG motivation to take on the responsibility of facilitating the intergroup relations among its different counterparts. It is likely to assume that GGs might have varying levels of motivation to take on, or shy away from, the role of intergroup facilitation. The underpinnings of these motivations

<sup>6</sup>For example, Extended contact might be further developed as a conflict resolution tool based on the findings of this research. The extended contact hypothesis contends that knowledge about cross-group friendships (i.e., knowing that an ingroup member has a positive relationship with an outgroup member) can improve outgroup attitudes and thus advocates the advantages of intergroup contact by proxy (Wright et al., 1997; Christ et al., 2014). Usually, in order for extended contact to have a positive effect, the process has to entail an ingroup exemplar having contact with an outgroup exemplar (Wright et al., 1997). Nonetheless, although dual identifiers are not classic ingroup exemplars, the frequency of their unmediated interaction with both counterparts holds the potential for effects similar to the ones observed in the extended contact literature.

should be explored and mapped out in order to enhance the ability to predict GG action.

Another noteworthy research topic that has yet to be addressed is the development of novel identities among groups holding multiple identities. So far, the literature in the field has addressed groups of people with shared multiple identities as maintaining different existing identities coupled together. However, it is possible that for individuals and groups that maintain a state of multiple identities over time, the multiple identities morph into a new single identity that is distinct and independent from its identities of origin (Rockquemore, 1998; Simon and Ruhs, 2008). This notion may account for some of the variation in the existing literature regarding multiple identities. Indeed, in some instances individuals who hold multiple identities, may in reality be holding a new fused single identity.. Moreover, the presence of multiple identities (e.g., as reflected in the GG empirical studies) might play a role in affecting intergroup relations simply as a result of exposure to complex identity structures. This would mean that exposure to any group with multiple identities would have the same effect on intergroup relations regardless of its connections to the conflicting groups at hand. Both these options contend with the notion presented in this paper by either eliminating the uniqueness of GGs, or by extending that uniqueness to any multiple identity group regardless of its affiliation, and require further research.

Finally, while almost all the research described in this paper is from the realm of the social psychology of intergroup relations, there are several disciplines that have a vested interest in multiple identity GGs as well. For example, researchers in the field of social cognition have developed an approach in which social categories are perceived in a continuous manner (Maddox, 2004; Eberhardt et al., 2006). Based on the racial GG studies presented above, it seems likely that the perception of racial categories as continuous may have the potential to alleviate racist behavior, and that the presence of racial GGs have the potential of emphasizing this continuum. Future research should examine both these assumptions that may lead to new and innovative ways to combat racism. Another possible interdisciplinary connection is with the work on social networks, from the fields of sociology and communications. This work has great potential to enrich our understanding of the role GGs have to play in intergroup relations. Research in these fields has found that strategically situated groups such as the GG can be expected to facilitate the dissemination of information between conflicting social groups, to create interpersonal connections across social fault lines, and to induce efficient communication between groups in conflict (Granovetter, 1973; Long et al., 2013; Centola, 2015; Repke and Benet-Martínez, 2017). Accordingly, an interdisciplinary approach might prove very useful in this regard.

# CONCLUSION

When it comes to social-psychological theories regarding groups of people with shared multiple identities in intergroup relations, it seems that different theories only address specific facets of the potential role these groups can play in intergroup dynamics. Our GG framework builds on several of the existing theories and extends them by outlining what functions multiple identity GGs may serve (as bridges or barriers), and how their presence and/or experience may improve or deteriorate intergroup relations. As such, it also provides a research agenda along those lines, as well as additional important questions such as when GGs can be expected to serve as bridges or barriers.

Importantly, our review of relevant published and unpublished findings suggests that multiple identity GGs hold a unique potential when it comes to improving intergroup relations. GGs which can potentially be situated interchangeably in regard to a given social border, can act as catalysts for the attempts to shift or redefine the borders between social categories, and the mere presence of a GG in situations of intergroup conflict can be expected help partially dismantle social categories that otherwise facilitate intergroup strife. GGs multiple affiliation might signal to their respective groups that a superordinate identity, incorporating both groups, is possible. Additionally, the manner in which GGs cross social categories can help confront stereotypical and heuristic modes of thinking, and raise tolerance for outgroups in general. These positive effects multiple identity GGs may have on intergroup relations originate both from such groups perceiving themselves as having multiple identities, and from such groups being perceived as having multiple identities by their social counterparts. However, the fleshing out of these multiple identities and social affiliations of GGs can also be detrimental under certain circumstances. Hopefully, future research will shed more light on the doubleedged sword of multiple identities, and tease apart the factors that induce negative intergroup outcomes from those that promote positive intergroup dynamics and facilitate intergroup conflict resolution.

# AUTHOR CONTRIBUTIONS

All authors substantially contributed to the conception of the work; and drafted the work; and approved the version to be published; and agreed to be accountable for all aspects of the work.

# FUNDING

This work was supported by the Israeli Science Fund Grant (Grant Number 1772/14) awarded to TS, and by the European Research Council Grant awarded to EH (Grant Number 335607).

# REFERENCES

fpsyg-08-01097 June 27, 2017 Time: 12:17 # 11


reduction of intergroup bias. Eur. Rev. Soc. Psychol. 4, 1–26. doi: 10.1080/ 14792779343000004



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Levy, Saguy, Halperin and van Zomeren. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.