Culture, Sex, and Group-Bias in Trait and State Empathy

Empathy is sharing and understanding others’ emotions. Recently, researchers identified a culture–sex interaction effect in empathy. This phenomenon has been largely ignored by previous researchers. In this study, the culture–sex interaction effect was explored with a cohort of 129 participants (61 Australian Caucasians and 68 Chinese Hans) using both self-report questionnaires (i.e., Empathy Quotient and Interpersonal Reactivity Index) and computer-based empathy tasks. In line with the previous findings, the culture–sex interaction effect was observed for both trait empathy (i.e., the generalized characteristics of empathy, as examined by the self-report questionnaires) and state empathy (i.e., the on-spot reaction of empathy for a specific stimulus, as evaluated by the computer-based tasks). Moreover, in terms of state empathy, the culture–sex interaction effect further interacted with stimulus traits (i.e., stimulus ethnicity, stimulus sex, or stimulus emotion) and resulted in three- and four-way interactions. Follow-up analyses of these higher-order interactions suggested that the phenomena of ethnic group bias and sex group favor in empathy varied among the four culture–sex participant groups (i.e., Australian female, Australian male, Chinese female, and Chinese male). The current findings highlighted the dynamic nature of empathy (i.e., its sensitivity toward both participant traits and stimulus features). Furthermore, the newly identified interaction effects in empathy deserve more investigation and need to be verified with other Western and Asian populations.


INTRODUCTION
Empathy is the sharing and understanding of others' emotions (Cohen and Strayer, 1996;Eslinger, 1998;Zhao et al., 2019) and is an essential social skill (Baron- Cohen and Wheelwright, 2004). This skill can be evaluated in terms of trait and state empathy (Van der Graaff et al., 2016). Trait empathy reflects one's general characteristics of empathy (Song et al., 2019), commonly gauged by self-report questionnaires (e.g., Zhao et al., 2019). State empathy represents one's context-specific empathy toward a given stimulus (Song et al., 2019), usually evaluated through computer-based tasks (e.g., Neumann et al., 2013). Recently, researchers identified a significant culture-sex interaction effect on trait empathy, suggesting a combined impact of cultural background and biological sex on one's empathy . However, to date, the interaction effect has only been explored in terms of the trait (e.g., Melchers et al., 2015;Zhao et al., 2019) but not state empathy. The current study aims to examine the culture-sex interaction effect on both trait and state empathy with a cohort of Australian and Mainland Chinese participants.
In the research field of the cross-cultural comparison of empathy, culture is defined as a multi-faceted concept, covering a plethora of factors, including the local customs, country of origin, ethnicity, and native language . Although culture is deemed to have an important impact on empathy, the current conclusions regarding the cultural impact on trait and state empathy are inconsistent among publications (see Tables 1-3). This inconsistency may be decoded from three aspects. First, empathy consists of two main components: emotional empathy and cognitive empathy (Cohen and Strayer, 1996;Shamay-Tsoory et al., 2009;Shamay-Tsoory, 2011). These two components work in unison to process others' emotions but with different emphases. Emotional empathy is the more rudimentary form of empathy, representing the automatic sharing of others' feelings (Shamay-Tsoory et al., 2009). Cognitive empathy is the more advanced form of empathy, representing the understanding of others' emotions (Shamay-Tsoory et al., 2009). As summarized in Tables 1, 2, the cultural impact on each evaluation of empathy (e.g., emotional, cognitive, and overall empathy) could be inconsistent.
Second, among the previous publications, researchers have primarily made comparisons of empathy between Western and Asian cultures (Tables 1-3). These two cultures present a sharp contrast of self-construal (Singelis, 1994;Zhao et al., 2019). Self-construal is the psychological distance between the self and others and is the pillar of an individual's psychological reactions (Singelis, 1994). For example, Zhao et al. (2019) identified that Australian and Chinese participants represented independent and interdependent self-construal, respectively. Moreover, with the Australian and Chinese participants, Zhao et al. (2019) found that self-construal was a mediator of the prediction from culture to trait empathy; however, the mediating effect varied between female and male participant groups 1 , indicating the culture-sex interaction effect in empathy in a broader sense (also see Schmitt, 2015). Nevertheless, the complicated issue is that the definitions of the Western and Asian cultures varied among previous studies (Tables 1-3). Some researchers referred to "culture" as the cultural background (e.g., Australians and Chinese locals recruited from their home country; Zhao et al., 2019), some researchers referred to it as ethnicity (e.g., Caucasian and Chinese participants; Xu et al., 2009), and some others referred to it as country of origin (e.g., Western and Asian born students, studying in Canada; Cassels et al., 2010). Furthermore, some researchers recruited participants with mixed cultural 1 Based on Australian and Chinese female participants, Zhao et al. (2019) found that both independent and interdependent self-construals could be mediators of the prediction from culture to the trait empathy scores (i.e., maintaining a balanced self-other distance could be essential for females to obtain a higher score on trait empathy). In contrast, for male participants, only interdependent self-construal was a mediator (i.e., seeking a closer self-other relationship might be indispensable for males to reach a higher score on trait empathy)  backgrounds (e.g., Cassels et al., 2010), mixed ethnicities (e.g., Kaelber and Schwartz, 2014), or mixed countries of origin (e.g., Xu et al., 2009).
Third, the impact of sex has been largely ignored by previous researchers (Tables 1-3). However, Zhao et al. (2019) proposed that one's cultural background and biological sex might interact to determine the participant's empathy. Specifically, they found that cultural differences in trait empathy (i.e., Australian > Chinese participant) were only significant with the female but not male participants . Meanwhile, they found that the sex difference (i.e., female > male participant) in trait empathy was only significant for the Australian but not Chinese participants . Moreover, they reported that the above culture-sex interaction effect was significant on all forms of trait empathy (i.e., emotional, cognitive, and overall empathy) . Among the previous researchers (Tables 1-3), only Melchers et al. (2015) examined the culture-sex interaction effect in trait empathy. Interestingly, they identified a similar 2 effect with German and Chinese participants (Melchers et al., 2015). In contrast, most of the other researchers examined the cultural impact on empathy per se but neglected to evaluate the possible culture-sex interaction effect. Based on the findings by Zhao et al. (2019), the sex ratio of participant samples could confound the results of cultural differences in empathy. In other words, among the previous studies (Tables 1-3), the larger Western-Asian cultural difference in empathy was likely to be reported by these studies with more female participants (i.e., larger female%) (see a discussion by Zhao et al., 2019).
The culture-sex interaction effect in empathy could be due to the inconsistent social expectations for sex roles across cultures . In Western cultures, females are expected to be warm-hearted and openly affectionate, while males are encouraged to be strong-willed and emotionally invulnerable (Jaggar, 1989;Merten, 2005). In contrast, the above social expectation for sex differentiation could be smaller in Asian than in Western cultures; for example, according to the traditional Chinese culture, both Chinese males and Chinese females are expected to seek a balance between Yin (e.g., negative and femininity) and Yang (e.g., positive and masculinity) (Huang, 2006;Zhao et al., 2019). Consistently, researchers found that the effect size of the sex difference on trait empathy is smaller for Asians (see a review by Zhao et al., 2019) than Westerners (see a review by Groen et al., 2015).
To date, the culture-sex interaction effect has been verified in trait (Melchers et al., 2015;Zhao et al., 2019) but not state empathy. This may reflect that the culture-sex interaction effect on state empathy is harder to detect relative to trait empathy. On the one hand, social expectations could have less impact on computer-based assessments of state empathy relative to selfreport evaluations of trait empathy (Baez et al., 2017). If the culture-sex interaction effect in trait empathy is largely due   Atkins et al. (2016) used video clips in which the main character orally related their negative social experience; in contrast, the other researchers used static pictures with a relevant emotional background (e.g., a natural disaster or a social activity). *For the empathy task, OE, the overall empathy (i.e., the integrated concept of emotional and cognitive empathy); EE, the emotional empathy; CE, the cognitive empathy; PT, the perspective-taking.
† For the Western-Asian comparisons, W, Western participants; A, Asian participants; , showed; ×, did not show. ab For the participant, a multi-nationalities; b multi-ethnicities. For the stimulus type, Cheon et al. (2011) presented emotional pictures with a relevant emotional background (e.g., a natural disaster or a social activity), while de Greck et al. (2012) presented photos of facial expressions with a plain back-drop. *For the empathy task, OE, the overall empathy (i.e., the integrated concept of emotional and cognitive empathy).
† For the Western-Asian comparisons, W, Western participants; A, Asian participants; b/l/m/rBA, the bilateral, left, middle, and right side/s of a Brodmann Area, correspondingly.
Frontiers in Psychology | www.frontiersin.org to social expectations , the interaction effect on state empathy will be smaller in magnitude. On the other hand, state empathy is sensitive not only to participant traits (i.e., participant culture and participant sex) but also to stimulus features (e.g., stimulus ethnicity, stimulus sex, and stimulus emotion). Consequently, the culture-sex interaction effect on state empathy could be statistically swamped by other phenomena of empathy (e.g., ethnic group biases and sex group favors), or even combined with the other effects to formulate higher-order interactions (e.g., three-to four-way interactions). Ethnic group bias refers to the phenomenon that people have different inclinations in empathy toward individuals with the same or different ethnicities (Neumann et al., 2013). Commonly, there is an ethnic in-group bias in empathy (see Table 2); that is, people tend to show more empathy for the ethnic in-group individuals (i.e., those with the same ethnicity) than out-group ones (i.e., those with a different ethnicity) (Cheon et al., 2011;Neumann et al., 2013). However, an ethnic out-group bias in empathy (i.e., the converse trend as introduced) can also be observed (e.g., Neumann et al., 2013). According to findings reported by Neumann et al. (2013), Chinese participants showed an ethnic out-group bias in state empathy for positive emotions (e.g., happiness) and an ethnic in-group bias for negative emotions (e.g., sadness). In contrast, Australian participants had ethnic in-group biases for both positive and negative emotions (Neumann et al., 2013). The results of Neumann et al. (2013) revealed a three-way interaction between participant culture, stimulus ethnicity, and stimulus emotion. To date, higherorder interaction effects in empathy (e.g., three-or four-way interaction) that extend beyond the fundamental culture-sex interaction effect have not been examined (see Tables 1-3).
The origin of the ethnic group bias in empathy is intricate, and three theories have been proposed. The first is "in-group familiarity" (Cao et al., 2015). Researchers have found that the mirror neuron system (i.e., a key brain region for empathy) (Shamay-Tsoory et al., 2009) tends to be activated more by familiar than unfamiliar stimuli (Platek et al., 2006). Meanwhile, Cao et al. (2015) observed that increasing the out-group contact (i.e., reducing the ethnic out-group unfamiliarity) might help participants to reduce ethnic in-group bias in empathy. Specifically, they found that Chinese overseas students with greater daily-life contact with Caucasians reacted to Caucasian characters in pain with higher brain activation in the anterior cingulate cortex (i.e., a brain area related to empathy for pain) (Cao et al., 2015).
The second theory is "out-group hate" (Avenanti et al., 2010). Avenanti et al. (2010) presented American participants with pictures of three ethnic hands being pricked by needles. They found that both Caucasian and African American participants exhibited strong pain-related empathic reactivity for the hand of their own ethnicity but not for the hand of the other ethnicity (Avenanti et al., 2010). Intriguingly, both groups of participants showed moderate empathic reactivity for the hand that belonged to the most unfamiliar and prejudice-free "ethnicity" (i.e., it is an artificial ethnicity with its skin pigmented into violet) (Avenanti et al., 2010). It should be noted that although the theories of both in-group familiarity (Cao et al., 2015) and out-group hate (Avenanti et al., 2010) could explain the ethnic in-group bias in empathy, neither could explain the ethnic out-group bias in empathy (e.g., Neumann et al., 2013).
Instead, both ethnic in-group and out-group biases in empathy might be explained by a third theory called "reciprocal altruism" (Trivers, 1971;Mathur et al., 2010). According to Mathur et al. (2010), for preserving resources within a reciprocal group, people tend to express altruism toward reciprocal individuals (e.g., ethnic in-group) over non-reciprocal ones (e.g., ethnic out-group). It should be emphasized that empathy is an essential motivator of altruism (Mathur et al., 2010). Meanwhile, either people showing in-group or out-group bias in empathy might be tipped by the valence of emotions (Neumann et al., 2013). First, negative emotions (e.g., sadness, fear, and anger) reveal one's powerlessness (Merten, 2005) or dissatisfaction (Wubben et al., 2011), which may elicit help and cooperative behaviors from others (Hackenbracht and Tamir, 2010;Wubben et al., 2011). Second, positive emotions (e.g., happiness) commonly represent feelings of joyfulness and contentment; however, a smile can also be applied to conceal negative inner thoughts (e.g., sadness and anger) (Ekman et al., 1988;Svetieva et al., 2019). Moreover, Tan and Forgas (2010) found that participants in a happy mood, relative to those in a sad mood, were greedier during resource distribution among themselves and others. It should be noted that if an ethnic outgroup person in a happy mood tends to be greedy, the resource distribution will be non-reciprocal. Hence, it may be the case that showing more concerns for in-group suffering (i.e., a proclivity to share resources reciprocally) and being more concerned by out-group happiness (i.e., a sensitivity to a potential resource loss) complementarily reflect the theory of reciprocal altruism (Mathur et al., 2010).
The current study aimed to explore the culture-sex interaction effect in both trait and state empathy with Australian and Chinese participants. Trait empathy was evaluated by two self-report questionnaires: the Empathy Quotient (EQ; Baron- Cohen and Wheelwright, 2004) and Interpersonal Reactivity Index (IRI; Davis, 1980). State empathy was assessed by computer-based tasks adapted from Neumann et al. (2013). Referencing Zhao et al. (2019), the effect of culture-sex interaction in empathy was expected to be verified in both trait and state empathy. Referencing Neumann et al. (2013), both ethnic in-and outgroup biases in empathy were anticipated, which might further interact with the culture-sex interaction effect on state empathy. Finally, apart from the ethnic group biases, we considered that participants might adjust their empathy according to stimulus sex (i.e., the sex group favor). Since the sex group favor effect in empathy was not examined in previous studies (Tables 1-3), no specific hypotheses were made on this point.

Participants
Participants in this study (i.e., Australian Caucasians and Chinese Hans) were recruited from their home countries (i.e., Australia and China, correspondingly). Both Caucasian and Han are the main ethnicity of the respective countries. In this study, Participant Culture is defined as an integrated concept, covering the mainstream culture (i.e., the Western culture and the Chinese culture), ethnicity (i.e., Caucasian and Chinese Han), country of origin (i.e., Australia and China), as well as the official language (i.e., English and Chinese). In China, the official spoken language is Mandarin, and the official written language is the simplified Chinese characters. Both the official spoken and written Chinese were used in the current study.
The inclusion criteria of the current study were presented for participants at the beginning of each experiment (criteria for Australians/Chinese, respectively): (1) nationality was Australian/Chinese; (2) ethnicity was Caucasian/Han; (3) place of birth was Australia/Chinese Mainland; (4) the main place of growing up was Australia/Chinese Mainland; (5) the main place of residence was Australia/Chinese Mainland (6) 18 years or older; (7) first-or second-year undergraduate student; (8) normal or corrected normal vision; and (9) without any history of mental or neurological illness. Moreover, it was explained to all participants that the current investigation was restricted to the ones satisfying all of the above inclusion criteria.
An English-Mandarin bilingual researcher administered the current investigation. Both Australian and Chinese participants received instructions and completed the tasks in their native languages. Participants were required to provide informed consent online at the beginning of each investigation. After completing the tasks, each participant was reimbursed according to the ethical protocols (i.e., US$6 for a Chinese participant, US$11 for an Australian participant, or two study credits for a participant of either cultural group). The ethical protocols for conducting the current study were provided by an ethics committee in both Australia and China.
In total, 139 people were recruited 3 . Ten of them were excluded for the following reasons: one withdrew from the experiment; one did not finish the task due to a technical issue; two were identified as not taking the tasks seriously (e.g., repeatedly selecting the same option for items of trait empathy); and six did not satisfy the inclusion criteria (identified afterward according to the information they provided on the online survey, such as having a history of coma or depression). Therefore, the final sample size was 129, with 61 Australian Caucasian university students (32 females, mean age = 19.37 years, SD = 1.39; 29 males, mean age = 19.69 years, SD = 1.67) and 68 Chinese Han university students (36 females, mean age = 19.47 years, SD = 0.97; 32 males, mean age = 19.69 years, SD = 0.69).

Demographic Sheet
Participants' responses for self-report questionnaires were collected online 4 . On the demographic sheet, participants' age, sex, and other traits covered by the inclusion criteria were enquired about.

Empathy Quotient
Empathy Quotient is a self-report questionnaire that measures trait empathy as a single concept (i.e., an evaluation of the overall trait empathy) (Baron- Cohen and Wheelwright, 2004). EQ consists of 40 empathy items and 20 filler items; according to the official rule, these filler items are not counted for the total score of the EQ (Baron- Cohen and Wheelwright, 2004). These filler items were designed by the original authors to keep participants from continually answering empathy-related questions (Baron- Cohen and Wheelwright, 2004). The empathy items should be scored according to the official rule and summed up to a total score ranging from 0 to 80, with a higher score reflecting a higher overall trait empathy (Baron- Cohen and Wheelwright, 2004). The original English version (Baron- Cohen and Wheelwright, 2004) and the simplified Chinese-translated version (Zhao et al., 2018) of the EQ were administered in this study for Australian and Chinese participants, respectively. Based on the current participants, Cronbach's αs for the English and the simplified Chinese-translated versions of the EQ were 0.92 and 0.90, correspondingly. The current results of the internal consistency were similar to previous reports for the EQ based on both Western (e.g., Groen et al., 2015) and Asian populations (e.g., Zhao et al., 2018).

Interpersonal Reactivity Index
Interpersonal Reactivity Index is a self-report questionnaire that assesses four aspects of trait empathy (Davis, 1980), namely, empathic concern (i.e., IRI-EC, seven items), perspective-taking (i.e., IRI-PT, seven items), personal distress (i.e., IRI-PD, seven items), and fantasy (i.e., IRI-FS, seven items). According to Davis (1980), IRI-EC and IRI-PT tap into emotional and cognitive trait empathy, respectively. IRI-PD evaluates one's empathic personal distress (i.e., the aversive feelings when witnessing others' suffering), while IRI-FS assesses one's empathy for imaginary characters portrayed in photos, books, or movies (Davis, 1980). The total score of each subscale of IRI ranges from 0 to 28, with a higher score reflecting a stronger trait in the corresponding category. The original English (Davis, 1980) and the simplified Chinese-translated version of the IRI (Chan, 1986;Zhao et al., 2018) were administered in this study for Australian and Chinese participants, respectively. Cronbach's αs of the four IRI subscales (viz., IRI-EC, IRI-PT, IRI-PD, and IRI-FS) of the English version were 0.84, 0.79, 0.64, and 0.83, in sequence, and of the simplified Chinese version were 0.75, 0.61, 0.73, and 0.86, correspondingly. These values of Cronbach's αs were similar to previous findings based on both Western (e.g., Davis, 1980) and Asian participants (e.g., Zhao et al., 2019).

Computer-Based Tasks
Instruction Two computer-based tasks (i.e., tasks I and II) were administered in this study. The paradigms of both tasks were adapted from Neumann et al. (2013). Participants were asked to look at two cohorts of emotional stimuli and answer a set of empathyrelated questions for each stimulus. Meanwhile, participants were instructed to be concerned with the feelings of the main character in the given stimuli (i.e., to test participants' explicit empathy). Additionally, participants were told to respond to each question according to their own feelings, rather than focus on judging whether an answer was right or wrong (i.e., to minimize the impact of social expectations).

Parameters and Equipment
For each trial, a cross was presented as the central fixation (onscreen for 3,000 ms). Afterward, a stimulus was given (onscreen for 6,000 ms), followed by a series of task questions presented one by one (onscreen until the participants provided a valid answer). Stimuli were presented in a pseudo-random pattern and task questions were presented in a fixed pattern. E-prime 2.0 (Schneider et al., 2002) was used to present the tasks and collect participants' responses. A 14-in. (35.56 cm) laptop was used in this study with both Australian and Chinese participants.

Stimuli of task I
In total, 24 stimuli of single-character portrait photos (2 stimulus ethnicity × 2 stimulus sex × 6 stimulus emotion) were extracted from the NimStim database 5 (Tottenham et al., 2009). In terms of stimulus ethnicity, there were two ethnic groups (i.e., typical Caucasian and typical East Asian characters) presented with an equal presenting ratio (i.e., 50% Caucasian and 50% Asian characters). In terms of stimulus sex, female and male stimuli were given with an equal presenting ratio (i.e., 50% female and 50% male stimuli). In terms of stimulus emotion, six emotional types (i.e., happiness, anger, sadness, surprise, fear, and neutralpeacefulness) were presented with an equal presenting ratio (i.e., 16.7% per emotion). It should be noted that empathy for most emotions (i.e., happiness, anger, sadness, surprise, and fear) was straightforward. Specifically, in light of the neutralpeacefulness stimuli, participants were instructed to empathize with the peacefulness of the main character.

Stimuli of task II
In sum, 24 documentary photos (2 stimulus ethnicity × 2 stimulus sex × 6 stimulus emotion) were presented in task II. The stimulus sets of tasks I and II differed in two essential aspects. First, the stimuli of documentary photos (i.e., task II) captured the naturally expressed emotions; in contrast, the facial expressions of the NimStim models (i.e., task I) were posed under the instruction of Tottenham et al. (2009). Second, the stimuli of task II informed participants of the relevant emotional background (e.g., a wedding party or a work strike; see more details of the background information in Supplementary Document 1) 6 ; however, the background of NimStim stimuli was a plain backdrop (Tottenham et al., 2009).
Since no published picture databases can provide the full stimulus set required by task II of this study, three resources were used. Four pictures 7 were selected from the International Affective Picture System (IAPS; Lang et al., 1999), two pictures of happiness were obtained from Neumann et al. (2013), and the remaining were sourced from the Internet 8 . The final stimuli set of task II consisted of 10 single-character pictures and 14 multi-characters pictures. To help participants identify the main character, an arrow pointing to the main character was added to all task II stimuli. Except for the aforementioned aspects, the stimuli selection of task II mirrored the procedure of task I.

Task Questions Validation questions
Four validation questions were presented after each stimulus. In sequence, the questions were (Q1) "How strongly was the main character feeling the emotion? 1 = not at all to 9 = very strongly"; (Q2) "How negative or positive did the picture seem? 1 = very negative to 9 = very positive"; (Q3) "Viewing the picture, I felt _____. 1 = very calm/relaxed to 9 = very aroused/jittery"; (Q4) "Viewing the picture, I felt _____. 1 = very at ease/comfortable to 9 = very distressed." These four questions were used to record participants' judgments on emotional intensity, emotional valence, emotional arousal, and emotional distress, respectively.

State empathy questions
Four questions were used to tap into state empathy, and they were (Q5) "What was the main emotion that the main character was feeling? 1 = Happiness, 2 = Surprise, 3 = Neutral, 4 = Fear, 5 = Anger, and 6 = Sadness"; (Q6) "I felt _____ the feeling of the main character. 1 = not at all to 9 = very strongly"; (Q7) "I understood _____ the situation of the main character. 1 = not at all to 9 = very fully"; and (Q8) "I can _____ imagine myself in the situation of the main character. 1 = not at all to 9 = very easily." These four questions assessed participants' empathic accuracy, emotional empathy, cognitive empathy, and perspective-taking, correspondingly. In light of task I, three questions of state empathy (i.e., Q5 to Q7) were asked, but without Q8, since the NimStim stimuli did not give a valid emotional background for participants to commit the perspective-taking. In light of task II, all four questions (i.e., Q5 to Q8) were administered.

Pilot Testing
Pilot testing (n = 7) was conducted to optimize and finalize the study procedure (see Supplementary Document 2).

Trait Empathy
A group of 2 (participant culture) × 2 (participant sex) ANOVAs (sum of squares type II) was conducted to verify the impacts of participant culture, participant sex, and the culture-sex bride at her wedding party as well as a stimulus of an Asian bride at her wedding party (see more information in Supplementary Document 1). 7 The four pictures were IAPS #2377, #2102, #6313, and #4598. 8 The Internet pictures can be acquired from the current corresponding author. interaction effect on trait empathy scores (i.e., the EQ and IRI scores). Furthermore, pairwise comparisons with Bonferroni adjustment (accounting for the inflated type I error) were carried out to identify the source of any significant interactions suggested by the ANOVAs.

Empathic Accuracies for Tasks I and II
An omnibus test of 2 (participant culture) × 2 (participant sex) × 2 (stimulus ethnicity) × 2 (stimulus sex) × 6 (stimulus emotion) generalized linear mixed models (GLMM) (Logistic regression) was conducted based on participants' empathic accuracy (i.e., emotion recognition ACC; dummy coded, 0 = inaccurate and 1 = accurate) of stimuli tasks I and II, separately 9 . In this study, each participant had a corresponding ACC for each stimulus (i.e., in total 48 variables of ACC per participant). Therefore, the current authors firstly converted the data from the original wide version into a long version, and then conducted the above omnibus test based on the long version data. Moreover, since the empathic accuracy (i.e., ACC) might have an impact on participants' responses of empathy (e.g., emotional and cognitive empathy), the current authors conducted further analyses of state empathy based on the long version data with the variable of ACC controlled as a covariate. Accordingly, the degree of freedom increased for the current analyses of state empathy (based on the long version data) relative to the analyses of trait empathy (based on the wide version data).

State Empathy for Task I
With ACC controlled as a covariate, an omnibus test 10 of 2 (empathy task) × 2 (participant culture) × 2 (participant sex) × 2 (stimulus ethnicity) × 2 (stimulus sex) × 6 (stimulus emotion) linear mixed model (sum of squares type III) was carried out based on all state empathy responses collected in task I. If the above omnibus test indicated any significant interaction effects, pairwise comparisons (with Bonferroni adjustment) were conducted to identify the source of the interactions. Moreover, if the omnibus test indicated any significant interaction effects relevant to the variable of the 'empathy task' , follow-up analyses 10 were carried out. For the follow-up analyses, 2 (participant culture) × 2 (participant sex) × 2 (stimulus ethnicity) × 2 (stimulus sex) × 6 (stimulus emotion) linear mixed model (sum of squares type III) were conducted for the two categories of the 'empathy task' of task I (viz., emotional and cognitive empathy), separately (adjusted significance level α = 0.050/2; ACC was controlled as a covariate). Finally, pairwise comparisons (with Bonferroni adjustment) were carried out to identify the source of any significant interaction effects identified in the followup analyses.

State Empathy for Task II
The overall analysis procedure was similar to that of task I, but with two modifications. First, the 'empathy task' of task II 9 A group of post hoc and follow-up analyses was planned for the omnibus tests of ACC, but the current results revealed no significant interaction effect on the ACC (see the section "Results"). 10 The omnibus tests of state empathy (i.e., both tasks I and II) mirrored the ANOVA of the EQ (i.e., the overall empathy). In contrast, the follow-up analyses (i.e., both tasks I and II) mirrored the ANOVAs of the IRI subscales (i.e., the individual components of empathy, separately). consisted of three categories (viz., emotional empathy, cognitive empathy, and perspective-taking). Accordingly, the omnibus test of task II was 3 (empathy task) × 2 (participant culture) × 2 (participant sex) × 2 (stimulus ethnicity) × 2 (stimulus sex) × 6 (stimulus emotion) linear mixed model (sum of squares type III). Second, the adjusted significance level of the three followup analyses of task II was α = 0.05/3. The correlation coefficients between trait empathy and state empathy are presented in Supplementary Document 3. All analyses were conducted using SPSS version 27 (IBM Corp).

Stimulus Validity
A summary of the stimulus validity of the current stimuli (i.e., empathic accuracy, emotional intensity, emotional valence, emotional arousal, and emotional distress) is presented in Supplementary Document 1.
In light of the significant culture-sex interactions on the trait empathy scores (i.e., EQ, IRI-PT, and IRI-EC), pairwise comparisons revealed two main findings. First, the participant culture effect was only significant with female participants (i.e., Australian female > Chinese female, p < 0.

Empathic Accuracy
Omnibus tests revealed neither main nor interaction effects on the ACC of either the NimStim stimuli (i.e., task I, all ps ≥ 0.884) or the documentary stimuli (i.e., task II, all ps ≥ 0.968). Nevertheless, the ACC was still controlled in the following analyses of state empathy as a covariate.

State Empathy for the NimStim Stimuli
A full introduction of the state empathy results is presented in Table 5 and summarized in Supplementary Document 4. Due to the word limit of the main text,  EQ, Empathy Quotient; EQ-40, total score for the 40-item EQ; IRI, Interpersonal Reactivity Index; IRI-PT, total score for the IRI perspective-taking items; IRI-EC, total score for the IRI empathic concern items; IRI-PD, total score for the IRI personal distress items; IRI-FS, total score for the IRI fantasy items. only significant interactions with participant traits, not being qualified by higher-order interactions, are presented in the following.

Two-way interaction
Three significant two-way interactions were related to participant traits and were not qualified by the above higher-order interaction effects.  The omnibus test (α = 0.050) Two follow-up tests (α' = 0.025)

The Documentary stimuli (task II)
The omnibus test (α = 0.050) Three follow-up tests (α' = 0.017) For the empathy task, c, cognitive empathy; e, emotional empathy; p, perspective-taking. † For the participant sex and the stimulus sex, f, female; m, male. ∵ For the stimulus ethnicity, a, typical East Asians; c, typical Western Caucasians. nFor the stimulus emotion, its pairwise comparison results were summarized in Supplementary Document 5.
For the interaction effects, A, empathy task (i.e., emotional empathy and cognitive empathy for task I; emotional empathy, cognitive empathy, and perspective-taking for task II); B, participant culture; C, participant sex; D, stimulus ethnicity; E, stimulus sex; F, stimulus emotion. Specifically, the participant traits are in bold, and the "B × C" represents the culture-sex interaction effect in empathy proposed by Zhao et al. (2019). In this table, only details of significant results were presented; √ , significant; ×, non-significant; /, not relevant.

Follow-Up Analyses
Since the above omnibus test revealed significant interaction effects relating to the 'empathy task' , follow-up analyses were conducted for two types of the empathy task of task I (viz., emotional empathy and cognitive empathy), separately (adjusted significance level α = 0.025).

Emotional empathy
Two-way interaction. Two significant two-way interactions were related to participant traits.

Three-way interaction
Two significant three-way interactions were related to participant traits, and were not qualified by the above four-way interaction.

Follow-Up Analyses
Since the above omnibus test revealed significant interaction effects on the 'empathy task' , follow-up analyses were conducted for three types of the empathy task of task II (viz., emotional empathy, cognitive empathy, and perspective-taking), separately (adjusted significance level α = 0.017).

Emotional empathy
Four-way interaction. One four-way interaction was marginally significant, which was participant culture × participant sex × stimulus sex × stimulus emotion 11 [F(5, 2,874

Cognitive empathy
Three-way interaction. One significant three-way interaction was participant culture × stimulus sex × stimulus emotion 11 [F(5, 2,874.03) = 4.14, p < 0.001]. The following effects were significant according to the pairwise comparisons. First, the participant culture effect was significant on both male anger (

Perspective Taking
Three-way interaction. Two significant interactions were found.

DISCUSSION
In this study, the culture-sex interaction effect in empathy was studied with Australian and Chinese participants. Moreover, this interaction effect was identified on both trait and state empathy. For trait empathy, the current observation was consistent with previous findings (Melchers et al., 2015;Zhao et al., 2019). For state empathy, the culture-sex interaction effect further interacted with stimulus traits (e.g., stimulus ethnicity, stimulus sex, and stimulus emotion), resulting in three-or four-way interactions (see Table 5). Follow-up analyses of the higherorder interactions revealed that the impacts of stimulus traits varied among the culture-sex participant groups (i.e., Australian female, Australian male, Chinese female, and Chinese male). To conclude, the current results support the theory of culture-sex interaction effect in empathy . Furthermore, the current results highlight that beyond the fundamental culturesex interaction effect in empathy, there could be more intriguing interactions across participant traits and stimulus features.

Trait Empathy
The culture-sex interaction effect emerged as a clear trend in terms of trait empathy (see Table 4). This finding is in line with that of Zhao et al. (2019), who evaluated trait empathy with Australian Caucasian (n = 196) and Chinese Han (n = 211) university students. Specifically, in both the current and the previous study , the cultural differences in trait empathy were significant in female participants (i.e., Australian female > Chinese female participant) but not in male participants. Furthermore, sex differences in trait empathy were only significant with Australian participants (i.e., Australian female > Australian male participant) but not with Chinese participants. Zhao et al. (2019) proposed that the culturesex interaction in trait empathy might be germane to social expectations for emotional expressions. Generally, Western cultures encourage females to externalize their emotions more than males (i.e., the so-called emotional female and rational male; Merten, 2005). In contrast, honoring Confucius' Golden Mean philosophy, both Chinese males and Chinese females are supposed to avoid being either extremely emotional or extremely restrained (Huang, 2006;Zhao et al., 2019), resulting in a diminished sex difference in empathy (also see Zhao et al., 2020). Nevertheless, it should be noted that the above relationship between empathy and social expectations is only a theoretical proposal by Zhao et al. (2019), and future empirical studies are necessary to verify this proposal.

State Empathy
The current state empathy results were more complex, spanning significant two-, three-, and four-way interactions (see Table 5). For example, there were four-way interactions on overall and cognitive empathy for NimStim stimuli (i.e., participant culture × participant sex × stimulus ethnicity × stimulus sex), four-way interactions on overall and emotional empathy for the documentary stimuli (i.e., participant culture × participant sex × stimulus sex × stimulus emotion), as well as one three-way interaction on perspective-taking of the documentary stimuli (i.e., participant culture × participant sex × stimulus emotion).

The Culture-Sex Interaction Effect
Within each of the aforementioned three-and four-way interactions, there is a culture-sex interaction effect. Moreover, these three-and four-way interactions covered all forms of state empathy examined in this study (i.e., overall empathy, emotional empathy, cognitive empathy, and perspective-taking). On the one hand, the current findings suggest that culturesex interaction effects in empathy are not restricted to trait empathy (e.g., Zhao et al., 2019) but can expand to state empathy. On the other hand, the current results are similar to the findings of Zhao et al. (2019), suggesting that the culture-sex interaction is significant for inclusive components of empathy (see Melchers et al., 2015 andLachmann et al. (2018), both of them found the interaction was not significant on cognitive trait empathy). It is worth mentioning that Schmitt (2015) had a theory of "culturally variable sex difference"; as per Schmitt (2015), the culture-sex interaction effect could be a non-negligible phenomenon in a broad range of social and psychological subjects in addition to empathy. Therefore, the culture-sex interaction effect deserves attention from future cross-cultural researchers of sociology and psychology.
However, the culture-sex interaction effect has been ignored by most of the previous investigators of the Western-Asian cultural difference in trait and state empathy (see Tables 1-3). As noted by Zhao et al. (2019), the culture-sex interaction effect could be an explanation for the inconsistent results among the publications (see Tables 1-3). Moreover, Zhao et al. (2019) proposed that the magnitude of the Western-Asian cross-cultural differences in trait empathy could be enlarged along with the female ratio of a sample (i.e., a positive correlation with the female%). Both the current study and Zhao et al. (2019) presented supporting evidence for the above notion since the effect size of the cultural difference in trait empathy tends to be larger for female participants relative to male participants.

Participant Culture Effect
Referencing the results of culture-sex interaction in trait empathy , the Australian females should be the most empathic among the four culture-sex participant cohorts. Nevertheless, the current findings for state empathy revealed a different trend; that is, the advantages and disadvantages of state empathy are relatively counterbalanced for the participant groups. First, in light of the NimStim stimuli (i.e., task I), Australian participants expressed more cognitive empathy for positive and neutral stimuli (i.e., happiness and neutralpeacefulness). In contrast, Chinese participants reported more emotional empathy for negative emotions (i.e., anger and fear). Second, in light of the documentary stimuli (i.e., task II), the Chinese participants commonly expressed more empathy (i.e., overall empathy, emotional empathy, cognitive empathy, and perspective-taking) than Australian participants. However, Australian participants specifically reported more cognitive empathy for stimuli of male anger than Chinese participants did.
The inconsistency among the findings of trait empathy and state empathy (for NimStim and for documentary stimuli) is intriguing and can be explained by a range of factors. The first factor is social expectation. On the one hand, as per Zhao et al. (2019), Australian females' higher self-evaluated trait empathy could be largely due to the social expectation placed on them. However, the impact of social expectation on the computer-based evaluations (i.e., state empathy) could be weaker than that on selfreport evaluations (i.e., trait empathy) (Baez et al., 2017). More importantly, in the current study, participants were explicitly required to answer each state empathy question according to their inner feelings rather than social justice (see the section "Materials and Methods"). This instruction might have minimized the impact of social expectation on the state empathy tasks. On the other hand, Chinese traditional cultures (e.g., Confucianism and Taoism) honor humility and modesty in individuals (Lin et al., 2018). Hence, Chinese participants could downplay themselves while answering the trait empathy items (i.e., the items enquire "how good the participant is in empathy") 12 but might be more objective during responding to state empathy questions (i.e., the questions ask "how much the participant felt for a given stimulus") 13 . Therefore, Chinese participants may seem to be less empathic than Australian participants in light of trait empathy (i.e., the self-report scales assessed) but not state empathy (i.e., the computer-tasks evaluated).
The second factor is the background information of the stimuli. The current results suggest that when the emotional background information was withheld (i.e., the NimStim stimuli), Australian participants had higher cognitive empathy for neutral and happy stimuli, while Chinese participants showed more emotional empathy for negative emotions. This observation was in agreement with the distinct Asian and Western cultural requirements of emotional expression and suppression. Generally, in Asian societies, negative emotions are expected to be masked (e.g., by a neutral or smiling face) for maintaining interpersonal harmony (Wei et al., 2013). This social rule is different from Western societies, in which externalizing emotions is accepted as an honest way to express oneself (Gross and John, 2003;Murata et al., 2012). Consequently, since childhood, Chinese individuals have been trained to decode others' emotions according to contextual information, as well as trained to be alert to others' subtle emotional downturns (i.e., watch the "face colors") (Wang, 2001). Therefore, emotional understanding (i.e., cognitive empathy) for neutral and happy faces without emotional background information could be a challenge for Chinese participants (i.e., as per the Chinese culture, a neutral or happy face by itself could indicate neutral, happy, 12 Item examples of trait empathy (i.e., EQ and IRI) are "I am good at predicting how someone will feel." EQ#25 and "I would describe myself as a pretty softhearted person." IRI#22. 13 Question examples of state empathy (i.e., computer tasks I and II) are "I felt _____ the feeling of the main character. 1 = not at all to 9 = very strongly, " "I understood _____ the situation of the main character. 1 = not at all to 9 = very fully, " and "I can _____ imagine myself in the situation of the main character. 1 = not at all to 9 = very easily." or masked negative feelings). However, the empathic sensitivity (i.e., emotional empathy) for negative emotions might be more intense for the Chinese than Australian participants (i.e., due to the necessity of watching others' "face colors" in Chinese society) (e.g., Wang, 2001).
In contrast, when the background information was given (i.e., the documentary stimuli), empathy for most of the emotions was promoted for Chinese participants. One exception was the cognitive empathy for the stimuli of male anger. Anger is an intense emotion that disturbs the harmony of interpersonal relationships (de Greck et al., 2012). Influenced by the Confucian Golden Mean philosophy, the Chinese may value social harmony much more than Westerners (Drummond and Quah, 2001;de Greck et al., 2012;Liu, 2014). In light of Chinese culture, expressing anger could be labeled as lacking in self-control (Kornacki, 2001;Kong et al., 2020). In contrast, for Westerners, sincerely expressing emotions could be deemed as a way to enhance interpersonal understanding (Gross and John, 2003;Murata et al., 2012). Moreover, de Greck et al. (2012) decoded the neurological basis of Western-Asian cultural differences in empathy for anger. They found that facing ethnic in-group anger, German participants had more brain activation in the cognitive empathy-related brain regions (i.e., the inferior temporal gyrus and middle insula). In contrast, Chinese participants showed more brain activation in the emotional regulation and personal distress-related brain region (i.e., the dorsolateral prefrontal cortex). de Greck et al. (2012) claimed that the Western participants might try to understand the anger; meanwhile, the Chinese participants might attempt to inhibit their aversive feelings stirred up by the anger. Noticeably, some previous researchers of cultural differences in empathy (see Tables 2, 3) adopted the concepts of "negative emotions" or "suffering" (i.e., mixed negative emotions) as emotional stimuli. However, the current results highlight that the participants' cultural differences in empathy can be qualified by the subtypes of negative emotions.

Ethnic Group Bias
In this study, the dominant trend of ethnic group bias in state empathy was the ethnic in-group bias for negative emotions together with the ethnic out-group bias for positive emotions. These findings were in line with our hypothesis (see the section "Introduction") as well as the previous observation by Neumann et al. (2013). Specifically, the current Chinese participants exhibited ethnic in-group biases on overall empathy (i.e., the holistic concept of emotional and cognitive empathy) for fear (NimStim stimuli) and sadness (documentary stimuli).
In contrast, the current Australian participants expressed an ethnic out-group bias on overall empathy for happiness (documentary stimuli). These findings cannot be fully explained by either the theory of in-group familiarity (Cao et al., 2015) or the one of out-group hate (Avenanti et al., 2010). Instead, as discussed in the Introduction section, being concerned about ingroups in need (i.e., the in-group bias for negative emotions) and out-groups in a triumphant mood (i.e., the out-group bias for happiness) could be two facets of the "reciprocal altruism" (Trivers, 1971;Mathur et al., 2010).
Nevertheless, two exceptions of ethnic group bias in state empathy were identified with the current Chinese participants (both for NimStim stimuli). First, the Chinese participants showed an ethnic out-group bias for the NimStim sadness. Sadness may be perceived as a symbol of powerlessness and low self-esteem (Merten, 2005); an exposure of one's weakness in front of others without a good reason could be interpreted by Chinese people as "losing face" (i.e., a Chinese word, describing the feeling of embarrassment and shame for oneself as a consequence of unsuitable conduct; Ho, 1976;Zhang et al., 2011). Trommsdorff et al. (2007) coined the term "non-acting" to explain the same situation. They stated that in cultures that discourage emotional externalization, individuals might purposely inhibit their reactions to an emotional person so as to "save that person's face" (Trommsdorff et al., 2007). Therefore, the current Chinese participants might refrain from empathy toward the Asian characters expressing sadness without a good reason (i.e., NimStim stimuli), leading to the out-group bias. However, as long as an emotional background was given for the sadness of the documentary stimuli (e.g., an earthquake or bushfire ruin), the ethnic group bias of the current Chinese participants turned into an ethnic in-group bias.
Second, there was a four-way interaction (i.e., participant culture × participant sex × stimulus ethnicity × stimulus sex) on the overall empathy for NimStim stimuli. Further examination of the four-way interaction showed an ethnic out-group bias with the Chinese female participants on NimStim male stimuli. The reasons for the ethnic out-group bias could be still due to the non-acting strategy (Trommsdorff et al., 2007). Relatively, Western cultures provide more freedom for individuals to express their emotions, while Asian cultures value emotion regulation more (Davis et al., 2012;Wei et al., 2013). Moreover, with Chinese and American participants, Davis et al. (2012) found that Chinese male participants expressed the highest emotion regulation, which was in line with their concern that the social pressure on moderating emotions was stronger for Chinese males than the other culture-sex participant groups (i.e., a culture-sex interaction effect in emotion regulation). Hence, the current Chinese female participants might adopt the nonacting strategy to specifically "save the face" of the NimStim Asian male over the NimStim Caucasian male (Trommsdorff et al., 2007). This turned out to be the Chinese female participants' ethnic out-group bias in empathy. Nonetheless, when the emotional background was illustrated with the emotion (i.e., the documentary stimuli), the ethnic out-group bias for male stimuli was absent from the Chinese female participants. The above results stress that the ethnic group bias may vary among the culture-sex participant groups, which can be moderated by the availability of the background information; however, these possibilities were overlooked by previous researchers (Tables 2, 3).

Sex Group Favor
Sex group favor in empathy was not examined in previous studies summarized in Tables 1-3. The current results revealed that the main sex group favor was biased to female (i.e., female > male stimuli, see Table 5). This main favor is consistent with a common social consensus, namely, females are more vulnerable and should be treated with extra consideration (i.e., the "ladies first" ideology) (Tuleja, 2012). Nevertheless, some minor variations on the sex favor effect could still be identified among the four culture-sex participant groups. First, in light of the NimStim stimuli, the sex group favor (i.e., female > male stimuli) was only significant with Australian male participants (i.e., the overall empathy for both Caucasian and Asian stimuli, as well as cognitive empathy for Asian stimuli), but not with the other three culture-sex participant groups. Second, in light of the documentary stimuli, the main sex group favor (i.e., female > male stimuli) was identifiable with all culture-sex participant groups. However, this 'ladies first' favor in empathy for the documentary stimuli tended to be stronger for the Australian than Chinese participants; this result also supported the notion that sex differentiation is more pronounced in Western than in Asian cultures .
Third, the opposite sex group favor (i.e., the "alpha male" ideology) was also presented in the current results, particularly with the Australian participants. On the one hand, Australian male participants expressed more overall empathy for male happiness of the documentary stimuli (i.e., a male runner in the marathon) than the female ones (i.e., a bride in the wedding ceremony). Intriguingly, toward the same stimuli, the Australian female participants' sex favor on the overall empathy was biased to female (i.e., the bride's happiness > the male runner's happiness). In contrast, Chinese female and Chinese male participants showed non-significant sex favor on the overall empathy for happiness (i.e., the bride's happiness = the male runner's happiness). Besides further stressing that sex differentiation can be more polarized in Western than in Asian cultures, the above results are in line with the stereotype of Australian males (i.e., the 'Sporting Manhood in Australia'; Adair et al., 1997Adair et al., , 1998. On the other hand, Australian female participants' sex group favors on overall empathy for the documentary stimuli of anger and sadness were biased to male (i.e., male > female stimuli; see Supplementary Document 1 for the stimuli's background information). Teague (2014) evaluated empathic accuracies with three ethnic groups of Americans (viz., Caucasian, African, and Chinese). Teague found that relative to the male participants, the female participants of all three ethnic groups tended to be more sensitive to negative emotions (e.g., anger and sadness) expressed by Caucasian characters (i.e., the main ethnicity of the country) (Teague, 2014, see pp. 107-108). Moreover, relative to African male participants, the African female participants were hypersensitive to in-group anger and sadness (i.e., expressed by African characters). In contrast, Chinese female and Chinese male participants' reactions toward in-group anger and sadness (i.e., expressed by Chinese characters) were relatively similar. Results of Teague (2014) and the current study imply that sex group difference and sex group favor in empathy for negative emotions may be relevant to social vulnerability, and the female vulnerability may be more obvious in Western than Asian societies. Nevertheless, since Teague (2014) did not split the stimuli according to stimulus sex, whether females in Western societies were specifically sensitive to male negative emotions was not definitive. Nevertheless, the current results indicate that Western females may be more empathic toward male anger and sadness than female ones. The sex group favor in empathy, especially the sex favor against common consensus (i.e., the alpha male ideology), is worthy of further investigation.

Limitations and Further Studies
The current study has several limitations. First, the sample size was small. Conclusions regarding the interaction effects in state empathy need to be replicated based on a larger sample size. Second, only university students were recruited in this study, and hence, the current findings might not be extended to the general populations of Australia and China. Third, in this study, the ethnic group bias and sex group favor were only explored in terms of state empathy but not trait empathy (i.e., the EQ and IRI items do not examine these phenomena). Further researchers might consider investigating these phenomena in trait empathy using self-report questionnaires. However, it should be noted that participants can interpret questions regarding ethnic group bias and sex group favor as tapping into racism and sexism. Consequently, participants may respond to these questions according to social desirability (i.e., without racism and sexism). Fourth, it should be noted that the empathic accuracies of some emotions (e.g., fear, surprise, and neutral-peacefulness) were low in the current study (see Supplementary Document 1). Result interpretations for these emotions with a low empathic accuracy should be done with care. Fifth, questions of state empathy presented in the current computer-based tasks could still be categorized as subjective (e.g., "I felt _____the feeling of the main character") although they were comparatively more objective than the self-report items of trait empathy (i.e., the EQ and IRI items). The culture-sex interaction, ethnic group bias, and sex group favor effects ought to be verified by more objective techniques, such as brain imaging or physiological measurements (see Neumann and Westbury, 2011;Neumann et al., 2015). Sixth, to date, the culture-sex interaction effect in empathy with adult participants has been identified by Melchers et al. (2015) (i.e., Germans vs. Chinese), Zhao et al. (2019) (i.e., Australians vs. Chinese), as well as the current study (i.e., Australians vs. Chinese). It is noteworthy that the Asian participants of these three studies were all Chinese. Thereby, it is essential to verify in further investigations whether the culture-sex interaction in empathy can be generalized to other Asian cultures; in other words, whether the culture-sex interaction effect is a common phenomenon of the Western-Asian contrast or is a specific term to the Western-Chinese contrast 14 .
In addition, some limitations of the current computerbased tasks of state empathy should be elaborated. Firstly, the current participants' attitudes toward the other ethnicity (e.g., whether they had out-group hate) were not collected. The current authors deemed that out-group hate might not be a serious issue in the current case since both Australian and Chinese participants expressed the ethnic out-group bias in state empathy. Nevertheless, it is highly recommended for 14 Nevertheless, Trommsdorff et al. (2007) examined the culture-sex interaction effect in empathy with children recruited from kindergartens of four countries (i.e., Germany, Israel, Indonesia, and Malaysia). They found that the sex difference (i.e., female > male participant) in state empathy (i.e., they used the term "sympathy") was significant with the German group but not with the other three cultural groups. further investigators to record participants' attitudes toward other ethnicities to elaborate on this topic. Secondly, each component of state empathy was evaluated by a single item (e.g., "I felt _____ the feeling of the main character. 1 = not at all to 9 = very strongly", for emotional state empathy). The single-item design (i.e., also used by all previous investigations, see Tables 2, 3) could be criticized as not sufficiently reliable to capture the relatively stable psychological traits of empathy. A multi-item evaluation of state empathy ought to be considered in future investigations. Thirdly, participants' state empathy could be confounded by stimulus traits (e.g., age, clothing, and attractiveness of the character), which were not controlled in the current examinations. Fourthly, we did not directly compare the results of state empathy for NimStim stimuli with that for the documentary stimuli (i.e., tasks I and II, respectively) to evaluate the impact of background information on empathy. It should be noted that the stimuli of tasks I and II were different in several important aspects, including the availability of background information, the facial expressivity of the main characters, and more importantly, whether the characters expressed an emotion naturally. To evaluate the impact of background information on empathy, a future investigation with better-manipulated stimuli is necessary (i.e., an identical facial expression with different background information). Fifthly, regarding the stimuli of task II, we chose documentary photos of naturally expressed emotions with matched background information across Western and Asian stimuli (see details in Supplementary Document 1). Alternatively, researchers can do a computer manipulation on the facial expressions of those main characters to get a standard facial expression across Western and Asian stimuli. However, we are concerned that computer-modified facial expressions may change the social meaning and the biological validity of the stimuli. It is because emotional expressivity naturally differs between cultures (Rychlowska et al., 2015). Under the same situation, Westerners' facial expressions could be more exaggerated than Asians' (e.g., laughing or smiling at their wedding party). Hence, a standard happy face deemed so by Westerners could seem ecstatic to Asians. Therefore, we recommend documentary photos (i.e., naturally expressed emotions) rather than computermanipulated ones. Finally, it should be stressed that due to the small sample size, the current investigation may not provide enough statistical power to reveal all subtle interaction effects on state empathy. Further investigation with a larger sample size is highly recommended.

CONCLUSION
The current study explored the culture-sex interaction effect for both trait and state empathy with Australian and Chinese participants. In line with previous findings, the two-way interaction emerged as a reliable effect for trait empathy. Moreover, this two-way interaction effect was also presented on state empathy through higher-order interaction effects (i.e., three-and four-way interactions). On the one hand, the culturesex interaction effect on trait empathy revealed that the sex difference in empathy was larger for Australian than Chinese participants, while cultural difference in empathy was larger for female than male participants. On the other hand, threeand four-way interactions on state empathy highlighted the dynamic nature of empathy (i.e., its sensitivity to traits of both participant and stimulus). It is noteworthy that due to a neglect of the culture-sex and higher-order interaction effects, previous conclusions regarding the cultural impacts on empathy could be biased (as summarized in Tables 1-3). These important interaction effects in empathy, especially the culturesex interaction effect, are worthy of further investigations.

DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request to the corresponding author.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Griffith University Human Research Ethics Committee and East China Normal University Committee on Human Research Protection.

AUTHOR CONTRIBUTIONS
QZ collected the data, conducted the data analysis, and wrote the manuscript. DN designed the study and wrote the manuscript. CY designed the study and participated in the data collection and manuscript writing. SD participated in the manuscript writing. DS designed the study and wrote the manuscript. All authors contributed to the article and approved the submitted version.