Initially, 440 undergraduate students from a large, southwestern university were enrolled in the study. As we were interested in emerging adults' Instagram experiences, 23 participants who identified their age as 26 or older and 12 participants who did not identify their age were removed. The final analytic sample included 405 emerging adults (84% female, M_age = 20.05, SD_age = 1.62, Range_age = 18–25), recruited over two semesters in the 2017–2018 academic year. The sample was ethnically diverse (44% Asian/Asian-American, 29% White/Caucasian, 15% Hispanic/Latinx, 6% Black/African-American, 6% Mixed/Other). The study was advertised online to students in undergraduate psychology courses as a study examining how Instagram use relates to one's personality. Students received course research credit for their participation. Participation was limited to those with an active Instagram account (defined to students in the study advertisement as logging into their account at least once a week); there were no other requirements for participating. Sensitivity power analyses in G^*Power suggest that, with a sample of 405, an alpha of 0.05, power set to 0.80, and an average correlation among repeated measures of 0.67 (the average correlation observed among the pre- and post-assessments of the variables), power was adequate to detect small effect sizes (f = 0.06) for within factor effects.

On arriving at the laboratory, participants placed their belongings, including their cell phone, on a chair across the room to reduce distractions. After obtaining informed consent, participants provided demographic information and completed baseline measures of state affect and self-perceptions using a laboratory desktop computer. To maximize ecological validity, participants were told they would be browsing Instagram as they normally would, and therefore were not provided with a cover story. Participants were only provided a vague description of the study (i.e., how social media relates to personality), and were not provided information on the different conditions.

Participants then logged into their Instagram profile using the laboratory computer and were randomly assigned to one of three conditions. In the self condition, participants browsed only the posts that they themselves had previously uploaded. In the acquaintance condition, participants selected an acquaintance and browsed only the posts uploaded by this individual. An acquaintance was defined as an individual who is approximately the same age and gender, who posts relatively frequently, and someone who is not and has never been close friends with the participant (as browsing the content of close friends can have differential effects than browsing content from acquaintances or strangers; Lin and Utz, 2015; Liu et al., 2016). More specifically, participants were encouraged to choose a target that they knew from high school, but were not friends with then or friends with now. In the Instagram influencer condition, participants browsed only the posts uploaded by an influencer previously chosen by the researchers, with the influencers framed to the participants as strangers. The influencers were gender matched (one male, one female) to participants and were young adults. The male and female influencers were matched in terms of content uploaded (e.g., photos of themselves in exotic destinations). The influencers were chosen after discussions with undergraduates to determine profiles that were an accurate representation of an Instagram influencer, and a pilot study (n = 23) assessing the feasibility of the procedure. It is important to note that the profiles were not chosen to deliberately represent a specific type of self-presentation (e.g., high levels of the beauty ideal); however, analyses suggest that participants perceived that the influencers engaged in highly positive self-presentation (see Supplemental Materials), likely due to the inherent nature of these profiles.

All participants browsed the assigned profile for 10 min, with instructions to browse the Instagram profile as they normally would, without interacting directly with the profile (e.g., leaving likes or comments). For all three conditions, participants were instructed to browse the content their browsing target uploaded to their main page (i.e., participants did not view content in Instagram stories). An experimenter who was blind to the study hypotheses monitored time. Participants then logged out of their account and returned to the online questionnaire to complete post-browsing assessments of state affect and self-perceptions, perceptions of the target profile they browsed, number of and types of posts viewed, and surveys assessing personality traits. The experimenter checked and cleared the Instagram search history after each participant to remove any digital traces and to ensure that participants did not view other websites (or visit other Instagram pages or features) during browsing. No instances of visiting other websites or other Instagram pages or features occurred, and a timer embedded in the survey revealed that no participants returned to complete survey measures when they were supposed to be browsing. Because of this, along with how participants placed their phones across the room, we can be reasonably confident that participants engaged in the assigned task.

To examine how post-browsing reports of affect and self-perceptions differed between conditions, a series of ANCOVAs were run (Figure 1). All analyses were run controlling for the corresponding pre-browsing assessment of each variable; marginal means are reported in-text and in Figure 1. Post-browsing self-esteem significantly differed by condition, F_(2,399) = 22.60, p < 0.001, MSE = 0.85, ηp2 = 0.10. Participants in the self condition reported higher self-esteem (M = 5.88, SE = 0.08) than those in the acquaintance (M = 5.30, SE = 0.08) and influencer (M = 5.17, SE = 0.08) conditions; the acquaintance and influencer conditions did not differ from each other (p = 0.277). Post-browsing positive affect significantly differed by condition, F_(2,400) = 8.19, p < 0.001, MSE = 0.90, ηp2 = 0.04. Participants in the self condition reported higher positive affect (M = 4.49, SE = 0.08) than those in the acquaintance (M = 4.17, SE = 0.08) and influencer (M = 4.04, SE = 0.08) conditions; the acquaintance and influencer conditions did not differ from each other (p = 0.263). Post-browsing negative affect did not significantly differ by condition, F_(2,400) = 0.53, p = 0.588, MSE = 0.38, ηp2 = 0.00. Post-browsing interpersonal negativity significantly differed by condition, F_(2,320) = 28.65, p < 0.001, MSE = 1.33, ηp2 = 0.15. Participants in the self condition (M = 1.32, SE = 0.11) reported lower interpersonal negativity than those in the acquaintance (M = 2.14, SE = 0.11) and influencer (M = 2.46, SE = 0.11) conditions; the acquaintance and influencer conditions also significantly differed from each other (p = 0.045). Post-browsing positive self-perceptions significantly differed by condition, F_(2,400) = 23.42, p < 0.001, MSE = 0.56, ηp2 = 0.11. Participants in the self condition reported greater positive self-perceptions (M = 4.72, SE = 0.06) than those in the acquaintance (M = 4.20, SE = 0.06) and influencer (M = 4.15, SE = 0.07) conditions; the acquaintance and influencer conditions did not significantly differ from each other (p = 0.616).

To examine within-person changes in affect and self-perceptions, a series of multilevel analyses were conducted. Due to the nested nature of the data (two time points nested within each individual), multilevel modeling (MLM) was used with Restricted Maximum Likelihood Estimation. A random intercept was included to account for non-independence. Analyses were run examining changes in affect and self-perceptions within each condition, with the conditions dummy coded and the condition of interest the reference group. Time was effects coded (pre-browsing = −0.5, post-browsing = 0.5). Prior to the analyses reported below, analyses examined whether changes in affect and self-perceptions differed significantly across conditions, by testing whether the interaction between time and condition was significant. All changes within conditions significantly differed from each other (p's <0.001), except for negative affect (p = 0.762); thus, negative affect was not examined further.

Results are shown in Figure 2. Participants in the self condition reported significant increases in self-esteem (b = 0.37, 95% CI [0.21, 0.53], SE = 0.08, p < 0.001, d = 0.41), and positive self-perceptions (b = 0.38, 95% CI [0.25, 0.50], SE = 0.06, p < 0.001, d = 0.44), as well as marginal increases in positive affect (b = 0.17, 95% CI [0.00, 0.33], SE = 0.08, p = 0.051, d = 0.16). There were no changes in interpersonal negativity, b = 0.05, 95% CI [−0.17, 0.26], SE = 0.11, p = 0.654, d = 0.08.

Participants in the acquaintance condition reported significant decreases in self-esteem (b = −0.21, 95% CI [−0.38, −0.05], SE = 0.08, p = 0.010, d = 0.20) and positive self-perceptions (b = −0.14, 95% CI [−0.27, −0.01], SE = 0.06, p = 0.030, d = 0.20), and significant increases in interpersonal negativity (b = 0.87, 95% CI [0.65, 1.08], SE = 0.11, p < 0.001, d = 0.72). There were no changes in positive affect (b = −0.14, 95% CI [−0.31, 0.02], SE = 0.08, p = 0.094, d = 0.14).

Participants in the influencer condition reported significant decreases in self-esteem (b = −0.41, 95% CI [−0.57, −0.24], SE = 0.08, p < 0.001, d = 0.43), positive affect (b = −0.32, 95% CI [−0.49, −0.15], SE = 0.09, p < 0.001, d = 0.32), and positive self-perceptions (b = −0.19, 95% CI [−0.32, −0.06], SE = 0.07, p = 0.003, d = 0.26), as well as significant increases in interpersonal negativity (b = 1.19, 95% CI [0.97, 1.42], SE = 0.11, p < 0.001, d = 0.79).

MLM analyses were run examining feedback seeking and FoMO as moderators for within-person effects. The moderators were continuous and were grand-mean centered. Significant interactions were probed using simple slope analyses, in order to determine effects at one standard deviation above and one standard deviation below the mean. Due to the large number of tests, the p-value was set to 0.01. Prior to running the analyses reported below, analyses were run to examine if moderation significantly differed across conditions, by testing whether the three-way interaction between time, condition, and each moderator was significant. There were no significant differences between conditions for positive affect or negative affect (p's > 0.032), and therefore follow-up analyses were not run for either outcome. Additionally, there were no differences between condition for interpersonal negativity when feedback seeking was a moderator (p = 0.014).

For the self condition, the effects of browsing on positive self-perceptions were contingent on FoMO (b = 0.20, 95% CI [0.06, 0.33], SE = 0.07, p = 0.004) and feedback seeking (b = 0.22, 95% CI [0.09, 0.34], SE = 0.07, p = 0.001). Self-perceptions did not change for those lower in FoMO and feedback seeking (p's > 0.062), but increased for those higher in FoMO (b = 0.52, 95% CI [0.35, 0.68], SE = 0.08, p < 0.001) and feedback seeking (b = 0.59, 95% CI [0.42, 0.77], SE = 0.09, p < 0.001). There was no significant moderation for self-esteem (p's>0.061) or interpersonal negativity (p = 0.686).

For the acquaintance condition, the effects of browsing on self-esteem were contingent on FoMO (b = −0.38, 95% CI [−0.56, −0.19], SE = 0.09, p < 0.001) and feedback seeking (b = −0.23, 95% CI [−0.39, −0.08], SE = 0.08, p = 0.004). Self–esteem did not change for those lower in FoMO or feedback seeking (p's>0.187), but decreased for those higher in FoMO (b = −0.51, 95% CI [−0.74, −0.29], SE = 0.11, p < 0.001) and feedback seeking (b = −0.43, 95% CI [−0.64, −0.21], SE = 0.11, p < 0.001). The effects on interpersonal negativity were contingent on FoMO (b = 0.47, 95% CI [0.22, 0.71], SE = 0.12, p < 0.001). Interpersonal negativity did not change for those lower in FoMO (p = 0.018) but increased for those higher in FoMO (b = 1.21, 95% CI [0.92, 1.51], SE = 0.15, p < 0.001). Effects on positive self-perceptions were contingent on FoMO (b = −0.19, 95% CI [−0.33, −0.04], SE = 0.07, p = 0.010), and feedback seeking (b = −0.17, 95% CI [−0.29, −0.05], SE = 0.06, p = 0.005). Self-perceptions did not change for those lower in FoMO or feedback seeking (p's>0.632), but decreased for those higher in FoMO (b = −0.29, 95% CI [−0.47, −0.12], SE = 0.09, p = 0.001) and feedback seeking (b = −0.30, 95% CI [−0.46, −0.13], SE = 0.08, p < 0.001).

For the influencer condition, there was no moderation for self-esteem (p's>0.081). Effects for interpersonal negativity were contingent on FoMO (b = 0.33, 95% CI [0.08, 0.57], SE = 0.12, p = 0.009). Interpersonal negativity increased for both those lower (b = 0.93, 95% CI [0.64, 1.22], SE = 0.15, p < 0.001) and higher in FoMO (b = 1.51, 95% CI [1.19, 1.84], SE = 0.16, p < 0.001), with the effect stronger for those higher in FoMO. The effect on positive self-perceptions was contingent on FoMO (b = −0.28, 95% CI [−0.42,−0.14], SE = 0.07, p < 0.001) and feedback seeking (b = −0.20, 95% CI [−0.32, −0.07], SE = 0.06, p = 0.002). For those lower in FoMO and feedback seeking, positive self-perceptions did not change (p's > 0.756). Browsing decreased positive self-perceptions for those higher in FoMO (b = −0.47, 95% CI [−0.66, −0.28], SE = 0.10, p < 0.001) and feedback seeking (b = −0.40, 95% CI [−0.58, −0.22], SE = 0.09, p < 0.001).

Additional analyses are presented in the Supplemental Materials. First, analyses were run controlling for gender, race/ethnicity, number of posts viewed, whether scenery pictures were viewed, and whether comments were viewed. This was due to college women preferring Instagram more than men (Shane-Simpson et al., 2018), and differences across conditions in race/ethnicity, number of posts viewed, and whether scenery pictures were viewed. Specifically, those in the influencer condition viewed more posts than those in the self and acquaintance conditions, and those in the influencer condition viewed more scenery pictures than those in the acquaintance condition. Viewing comments did not differ across conditions (81% of participants in the self condition, 74% in the acquaintance condition, 79% in the influencer condition); however, we controlled for this because our interpretation of the results is influenced by the presence of and participants' attention to the comments. With these covariates, the effect of participants in the influencer condition reporting greater interpersonal negativity compared to those in the acquaintance condition was now marginally significant (p = 0.078). The moderation effect of FoMO on positive self-perceptions in the acquaintance condition was marginally significant with the more stringent p-value (p = 0.011). Otherwise, results remained the same.

Exploratory analyses assessed how the participant's Instagram posting and browsing frequency may moderate effects on well-being. No significant findings emerged. We also assessed how interpersonally-oriented perceptions (popularity, acceptance) of the browsing target may influence effects on well-being. This was of particular interest for the acquaintance condition, as targets higher in these perceived traits may have a qualitatively different type of profile, such as through accruing more likes and comments. Perceived popularity and acceptance were significant moderators for the self condition on positive self-perceptions. Increases in self-perceptions were strengthened for those who perceived themselves to be higher in popularity and acceptance. Analyses for the acquaintance condition were not significant with the more stringent p-value. Popularity was a significant moderator for the influencer condition on interpersonal negativity. Increases in interpersonal negativity were only observed for those who perceived the influencer as higher in popularity.

We examined if FoMO and feedback seeking interacted with each other in predicting changes in well-being, and if both moderators interacted with key demographic variables (age, gender, race/ethnicity) in predicting changes in well-being. None of these analyses were significant. Finally, due to how FoMO and feedback seeking emerged as significant moderators in the self condition, we examined if those higher in these traits may differ in the type of content they post and attend to when viewing their own profiles, in terms of groups of people, scenery pictures, captions, and comments. Those higher in FoMO had greater odds of reporting viewing groups of people and marginally higher odds of reporting viewing comments, and those higher in feedback seeking had lower odds of viewing scenery pictures; analyses were otherwise not significant.

Compared to pre-browsing, emerging adults who browsed their own profile reported small-to-moderate increases in self-esteem and positive self-perceptions. These results replicate previous research that has found that viewing and editing one's own Facebook page boosts self-esteem (Gonzales and Hancock, 2011; Gentile et al., 2012; Toma, 2013). People may experience self-affirmation after viewing the presumably positive events of their lives that they showcased online (Toma, 2013; Toma and Hancock, 2013). People may also experience positive reminiscing looking back at these positive events (Good et al., 2013), and may enjoy re-viewing positive feedback left by others (e.g., Burrow and Rainone, 2017). There were no significant findings observed for interpersonal negativity, perhaps because emerging adults are unlikely to experience fluctuations in envy and jealousy by looking at their own content.

The effects on positive self-perceptions were only observed for emerging adults higher in FoMO and feedback seeking. Perhaps those who are engrossed with online feedback seeking experience boosts in their positive self-perceptions when looking back at, presumably, the positive feedback they previously received. Moreover, perhaps those who fear that others are having more rewarding experiences than the self are appeased by looking, presumably, at their own positive life experiences. This highlights the transformative nature of SNSs such as Instagram, particularly in respect to visuality and quantifiability. Emerging adults who may be sensitive to others' evaluations of the self, or anxious about how their own social experiences compared to others, may find (temporary) relief looking back at their own content. The lack of significant findings for those lower in these traits could be because these individuals feel secure in their sense of self, and browsing their own profile does not necessarily affirm their positive self-image, at least in regards to positive self-perceptions.

Supplemental analyses examining emerging adults who viewed their own profile indicated that those higher in FoMO reported greater odds of viewing groups of people and marginally greater odds of viewing comments, whereas those higher in feedback seeking reported lower odds of viewing scenery pictures. It is possible that those higher in FoMO and feedback seeking have qualitatively different profiles (in terms of content uploaded and comments received) than those lower in these traits, or perhaps attend to different types of content and features when viewing their profile (e.g., those higher in FoMO may be more likely to look at comments compared to those lower in FoMO). Future research should systematically assess different types of Instagram content to see how those higher on these traits may differ in the types of content they upload. Future research can also adapt eye-tracking technology to examine if those higher in these traits visually attend to their own content differently than those lower in these traits.

Compared to pre-browsing, emerging adults who browsed the profile of an acquaintance reported small decreases in self-esteem and positive self-perceptions, and large increases in interpersonal negativity. Perhaps self-esteem and self-perceptions decreased (whereas positive and negative affect were unaffected) because they are especially tied to one's overall sense of self. By viewing the content of a known acquaintance, emerging adults may be especially likely to evaluate their own sense of self in relation to this individual, and therefore detectable immediate changes may be limited to self-oriented domains rather than overall mood. Interpersonal negativity may have increased because viewing the positive self-presentation of others may trigger feelings that others are leading better lives (e.g., Chou and Edge, 2012).

Several moderation effects were observed. Specifically, decreases in state self-esteem and positive self-perceptions, and increases in interpersonal negativity, were only observed for those higher in FoMO (although notably, moderation for self-perceptions was at the threshold of significance when using a more stringent p-value). This is in line with previous research suggesting that negative effects of SNS browsing are generally limited to those who have a tendency to compare (Vogel et al., 2015; de Vries et al., 2018), and expands on these findings by focusing on a type of comparison tendency that is specifically socially-oriented. Perhaps those who have a tendency to fear that others are having more rewarding experiences feel particularly worse when viewing content that may confirm these fears, thus potentially engaging in harmful, upward social comparisons. Additionally, decreases in state self-esteem and positive self-perceptions were only observed for those higher in feedback seeking. Those who value receiving feedback on SNSs may be especially negatively affected when viewing the presumably positive feedback that others receive. Together, these findings suggest how Instagram can transform emerging adults' social experiences, particularly for those higher in their tendency to seek feedback and experience FoMO. For example, those higher in their tendency to seek feedback may be more sensitive to the quantifiable nature of Instagram, and compare their own received feedback to the feedback received by an acquaintance. Those higher in FoMO may be particularly affected by how the visual affordances of Instagram can easily portray others' positive social experiences (e.g., Underwood and Ehrenreich, 2017).

Compared to pre-browsing, emerging adults who browsed the profile of an Instagram influencer reported small-to-moderate decreases in self-esteem, positive affect, and positive self-perceptions, and large increases in interpersonal negativity. These results generally mirrored the findings observed for the acquaintance condition, with the notable additional finding concerning positive affect. The Instagram influencers were especially perceived positively by participants, suggesting that these individuals engaged in particularly high positive self-presentation. Therefore, immediate effects may generalize to domains beyond those most relevant to one's sense of self. With that said, however, between-subject analyses suggested no robust post-browsing differences between the acquaintance and influencer conditions.

Additionally, decreases in positive self-perceptions only occurred for those higher in FoMO and feedback seeking. Again, this suggests that negative effects of browsing in certain domains may be limited to those who a) tend to perceive that others have more rewarding experiences than the self (and may engage in more upward social comparisons when browsing), and/or b) who have a tendency to elicit feedback from others on SNSs. For interpersonal negativity, it was observed that although interpersonal negativity increased for those both lower and higher in FoMO, these changes were especially pronounced for those higher in FoMO. Perhaps no browser is fully immune to experiencing feelings of envy and jealousy when browsing the highly positive accounts of Instagram influencers, with those who have a tendency to perceive that others have more rewarding social experiences especially affected. The tenets of the transformation approach may operate similarly but amplified when browsing the profile of an influencer compared to browsing the profile of an acquaintance, namely in terms of visuality (with influencers perhaps engaging in particularly highly positive self-presentation) and quantifiability (with influencers potentially accruing more feedback). The amplification of these components may be aided by publicness, as a fundamental goal for influencers is to obtain a larger following, which may be done by engaging in especially positive self-presentation, and may result in more feedback. Given that results were generally similar for acquaintances and influencers, future research is needed that better tests these tenets of the transformation approach, to determine if and when browsing influencers' profiles may have differential effects on emerging adults' psychological well-being compared to others' profiles.

These findings should be considered in the context of several limitations. First, we omitted implementing a cover story to maximize ecological validity. We instructed participants to browse the Instagram profiles as they normally would, and a cover story (e.g., using a bogus memory task; Fardouly et al., 2015) may have diverted their full attention from browsing. Thus, some of our findings may have been influenced by demand characteristics. Moreover, although we aimed to maximize ecological validity, the study nonetheless took place in a laboratory setting and also involved browsing using a desktop computer, which may provide a different browsing experience compared to browsing Instagram using one's smartphone. Future research could supplement experimental designs with daily diary designs that can capture more naturalistic browsing experiences. Second, we opted for a repeated measures design to maximize power, and therefore we did not use a control group. Given that different effects were observed across conditions, we can be reasonably confident that our findings are a result of the manipulation and not due to the passing of time. Third, our Instagram influencer targets, although matched for content and age, were chosen somewhat arbitrarily. Before the study, we conducted discussions with undergraduates to find typical influencer profiles and further tested the chosen profiles in a pilot study. However, we did not quantitatively assess the degree to which the targets engaged in overall positive self-presentation or specific types of positive self-presentation (e.g., exemplifying the beauty ideal), which would have provided a more systematic approach to choosing profiles.

Additionally, future research could examine participants' personal connections with influencers as a potential moderator. The chosen influencers were posed as strangers to participants, and it is plausible that many would not normally follow these types of profiles, which may reduce ecological validity. To date, experimental examinations of browsing influencers' profiles are lacking, and we aimed to provide a foundation for future research by including this popular type of profile in our design. To further enhance ecological validity, future studies could have participants personally select an influencer to browse, which can increase personal relevance. Studies could also assess characteristics such as participants' perceived similarity to and likelihood to form a parasocial relationship with a given influencer, and test for moderation. It is possible that under some circumstances, the effects of browsing are positive. Specifically, for emerging adults who perceive a browsing target to be more similar to the self, they may be more likely to engage in assimilative comparisons (Collins, 1996), which elicits positive emotions including inspiration, admiration, and optimism (Smith, 2000). Perceived similarity with a given influencer is strongly correlated with forming a parasocial relationship with that influencer (Lou and Kim, 2019), suggesting that browsing the Instagram content of an influencer with whom one has a parasocial relationship may have positive effects.

In addition, the chosen acquaintances differed across participants, introducing variability in the participants' browsing experiences. It is plausible that characteristics of the browsing target (e.g., number of followers, number of likes and comments) further influence how well-being is affected from browsing, which future research should explore further. There was also variability within the self condition. In some circumstances, viewing one's own Instagram profile may have negative effects. For example, receiving few likes may negatively influence well-being (Hayes et al., 2016), and therefore viewing content that did not receive this positive feedback may be harmful. Browsing old content may reveal preserved negative content (e.g., cyberbullying; Underwood and Ehrenreich, 2017), and some types of content that may have once been positive may now instead trigger negative memories (e.g., the demise of a close friendship; a romantic relationship breakup; Lukacs and Quan-Haase, 2015). Future research should examine if and when viewing one's own content may have negative effects on well-being.

More research is needed that assesses how closeness to the browsing target may influence effects on well-being. There is some evidence that browsing the SNS content of close friends can boost well-being (Lin and Utz, 2015; Liu et al., 2016). We did not include close friends as a condition in the current study due to our focus on acquaintances, whose content is more common on SNSs (Manago et al., 2012; De Meo et al., 2014), and to maximize power. Future research should systematically test how the effects of browsing content from close friends comparatively differs from browsing content from acquaintances and strangers.

We were also limited in our ability to examine gender and ethnic differences in the effects of browsing Instagram. The sample was largely female, and despite random assignment, Hispanic/Latinx participants and White/Caucasian participants were disproportionately represented across conditions. Supplemental analyses assessed gender and race/ethnicity as potential moderators for effects, with no significant findings observed. However, future research is needed that carefully recruits equal group sizes to appropriately test gender and race/ethnicity as moderators, in order to fully assess if moderation effects exist. For example, it is possible that significant moderation may emerge depending on if the race/ethnicity of the participant matches the race/ethnicity of the browsing target.

Our measures also had some limitations. We used brief assessments of positive and negative affect to reduce participant burden, although our items were adapted from previous research and a factor analysis supported our composites. Perhaps we observed few significant differences for these variables because our items may not have fully captured positive and negative affective states. Likewise, we utilized a single item to measure state self-esteem; although this has been previously done (Gross, 2009), the item may not have fully captured state self-esteem. Finally, although we assessed what types of content participants were exposed to in each condition (e.g., pictures with groups of people) and controlled for when there were differences in conditions (i.e., scenery pictures), more could be done to capture the diverse set of content that people post on Instagram, which can further enhance generalizability. These items were limited as they were designed to a) examine broad differences in viewing content across conditions, and b) serve as an extra check to ensure that participants engaged in browsing. Additionally, as the participants themselves reported on these categories, responses were subject to interpretation (e.g., whether scenery pictures include people). Future research should carefully assess different types of image content to examine how exposure to various content (e.g., viewing pictures of groups of people versus selfies) may moderate the observed effects.

Although we interpret our findings in the context of positive self-presentation due to the wealth of previous literature suggesting that images and feedback on SNSs are positive, we did not strictly quantify the extent to which this occurred, especially in terms of feedback (although perceptions of the browsing targets suggested that the targets did engage in positive self-presentation). Future research could explore the extent to which positive images and comments moderate effects on browsers' well-being, such as by implementing observational coding techniques. Importantly, Instagram is currently experimenting with hiding the number of likes when browsing others' content (Mosseri, 2019), which influences the extent to which feedback is displayed to browsers. Although this change was made after data collection for this study was complete and therefore does not influence our results, this has important implications for understanding how emerging adults may be influenced by browsing in the future, as effects on well-being may change. Finally, future research could also explore how viewing others' profiles may elicit positive effects. For example, there may be circumstances where positive effects on well-being occur through the elicitation of downward social comparisons, in which browsing targets are perceived to be worse off than the self. Alternatively, there could be browsing experiences that evoke feelings of optimism and inspiration (i.e., assimilative emotions; Park and Baek, 2018).