Why Adolescents Conform to High-Status Peers: Associations Among Conformity, Identity Alignment, and Self-Esteem

Abstract

This study examined whether conformity to high- but not low-status e-confederates was associated with increases in identification with popular peers, and subsequent increases in self-esteem. A sample of 250 adolescents (55.1% male; M_age = 12.70 years; 40.3% White, 28.2% Black, 23.4% Hispanic/Latino, 7.7% Multiracial/Other) participated in a well-established experimental chat room paradigm where they were exposed to norms communicated by high- and low-status e-confederates. Results revealed that for boys in the high-status condition only, but not girls, the positive relation between conformity and self-esteem was mediated by greater response alignment with popular peers. These findings bolster prior research by suggesting that conformity to popular peers may be partly motivated by drives for self-esteem and alignment with a valued reference group.

One of the most consistent phenomena reported in the social science literature pertains to the similarity between behaviors exhibited by adolescents and their peers (see meta-analysis: Giletta et al., 2021). Research over several decades has revealed that these similarities are due to two homophilic processes, including the tendency for adolescents to affiliate with peers who already possess attitudes and behavioral proclivities similar to their own (i.e., selection effects), and also adolescents’ conformity to their peers’ behavior (i.e., socialization effects). Socialization processes are relevant to a variety of traditionally adaptive and maladaptive behaviors (Laursen, 2018). Recent work suggests that peer influence effects are present for prosocial behavior and attitudes (Carlo & Padilla-Walker, 2020; Choukas-Bradley, 2015; Crone et al., 2022), the formation of political opinions (Oosterhoff et al., 2022), and improved academic performance (Duxbury & Haynie, 2020), and these effects may be particularly strong during adolescence (Foulkes et al., 2018). The majority of work to date, however, has examined the socialization of dangerous, illegal, and health risk behaviors, including aggression, substance use, sexual risk, and self-injurious behaviors (Brechwald & Prinstein, 2011; Montgomery et al., 2020). Thus, the study of peer socialization effects has major implications for clinical, developmental, social, and health psychology as well as related social science fields that similarly have great investment in understanding how norms, ideas, and behaviors may be contagious across social groups (e.g., sociology, public health, political science, economics, and marketing; see Prinstein & Giletta, 2016, 2020 for reviews).

Unfortunately, peer socialization processes are poorly understood. Although substantial data have amassed to suggest that both selection and socialization effects occur for a wide range of behaviors, and among broadly diverse groups of children and adolescents (Giletta et al., 2021), relatively little is known regarding mechanisms that may explain why individuals conform to their peers. This is a critical gap in knowledge; without data on the potentially malleable processes that underlie conformity, it is difficult to curtail adolescents’ socialization towards those maladaptive behaviors that are often perceived to be valued by peers.

Adolescents may be influenced by a wide range of their peers, and socialization within each peer context may involve discrete mechanisms. The majority of previous work has examined how adolescents are socialized by their best friends, primarily through the communication of descriptive norms (i.e., conferring opportunities for adolescents to perceive their peers’ behavior) or injunctive norms (i.e., adolescents’ perceptions of their peers’ attitudes towards a behavior) (Cialdini et al., 1991; Prentice & Miller, 1993; Veenstra & Lodder, 2022). Data suggest that adolescents’ exposure to each of these norms is associated with changes in adolescents’ own behavior across a variety of disparate circumstances (Brechwald & Prinstein, 2011). In addition to socialization among best friends, prior work has revealed that adolescents may be influenced by their romantic partners (Aikens et al., 2010), their “friends” on social media (Nesi & Prinstein, 2015; Choukas-Bradley & Nesi, 2020), peers portrayed within the media (Brown et al., 2006), and also by high-status peers within their own social milieu (Field & Prinstein, 2023; Helms et al., 2014). For instance, Helms et al. (2014) revealed that adolescents’ perceptions of their most popular high school classmates’ substance use (i.e., alcohol use, heavy episodic drinking, cigarette use, and marijuana use) predicted steeper increases in adolescents’ own substance use over a two-year period. Similarly, using an experimental paradigm, Cohen and Prinstein (2006) revealed that late adolescents’ exposure to injunctive norms ostensibly reported by popular peers led to greater endorsement of aggressive and deviant attitudes than when identical norms were reported by electronic confederates portrayed to be less popular (also see Rancourt et al., 2014). Similarly, greater susceptibility to norms communicated by high status peers, in conjunction with perceptions of popular peers’ number of sexual partners, also longitudinally predicts adolescents’ numbers of sexual partners (Choukas-Bradley et al., 2014). Collectively, these studies suggest that high status peers may exert greater influence than low status peers in late adolescence, consistent with work suggesting that adolescents’ awareness of social hierarchies and popularity may increase as they age (LaFontana & Cillessen, 2010). Further, research has suggested that adolescents have a heightened neural sensitivity to social evaluative feedback from peers relative to children and adults (Somerville, 2013), which cooccurs with parallel decreases in resistance to peer influence as youth progress throughout adolescence (Steinberg & Monahan, 2007).

The mechanisms underlying conformity to peers, and high status peers in particular, remain unclear, with emerging research suggesting numerous psychological processes that may promote or reinforce adolescents’ conformity. Early psychological work suggests that group processes may be especially relevant to the transmission of thoughts and behaviors. Social identity theory (Tajfel & Turner, 1986) suggests that individuals use group membership as a means of self-definition. The group one belongs to serves as a guide for one’s social identity, and the distinctiveness of that group relative to other groups leads to positive self-concept (Tajfel, 1972). These theories were later extended within self-categorization theory (Turner & Reynolds, 2011), suggesting that individuals derive positive self-worth from similarity with representative group members (Abrams & Hogg, 1990; Tajfel & Turner, 1986).

The influence compatibility model extends this framework to peer influence effects among adolescents, by positing that adolescents conform to their peers to increase similarity and compatibility among group members, both of which are beneficial for the relationship (Laursen & Veenstra, 2021). Providing support for this model, recent empirical work has suggested that differences among friends are associated with a higher likelihood of friendship dissolution (e.g., Hartl et al., 2015). Several additional theories such as the Theory of Reasoned Action (Fishbein & Azjen, 1977) and the Theory of Planned Behavior (e.g., Ajzen & Manstead, 2007) converge with the aforementioned theories to suggest that individuals’ intentions to engage in behavior are based in part on a desire to foster favorable self–other comparisons by using perceived group norms to guide behavioral decisions. Adherence to these perceived norms (e.g., perceived attitudes and behaviors) allows individuals to feel more similar to others (Kwon & Telzer, 2022), which may yield intrinsic rewards, such as positive self-regard (Gibbons, Stock, & Gerrard 2020).

Indeed, social comparison theories, such as the prototype/willingness model, extend this notion by suggesting that individuals’ conformity may be guided by a desire to conform not just to any social norms, but especially to those that represent a salient or desired reference group, such as high-status peers (Gibbons et al., 2020). Cognitive neuroscientists suggest that adolescent brain maturation involves the early proliferation of dopamine and oxytocin receptors in the limbic system, enhancing young people’s desire to elicit social rewards and positive regard among peers (Somerville, 2013; Kwon & Telzer, 2022). These mechanisms may be particularly powerful within adolescence, where peer feedback, approval, and acceptance from others are particularly salient, and may contribute to changes in identity development and self-concept (Harter, Stocker, & Robinson, 1996). Indeed, conforming to desirable prototypes (i.e., representation of high-status peers) may be intrinsically rewarding because it can improve one’s self-concept by increasing the perceived similarity between oneself and the prototype exhibiting the behavior (Abrams and Hogg, 1990). In other words, adolescents may be especially attuned to social norms offered by their most popular peers, and peer conformity may be associated with a desire to better align with high-status peers in order to experience more positive self-regard.

Emerging evidence provides preliminary support for some of the ideas offered by these theories, yet additional work sorely is needed. For instance, some prior work suggests that peer conformity may indeed be associated with social rewards, assessed by both behavioral and neural markers. For instance, best friend dyads most likely to foster deviant peer socialization are those in which utterances about risk or aggressive behavior are followed by friends’ displays of positive (i.e., behaviorally rewarding) affect (Dishion et al., 1996; Veenstra & Laninga-Wijnen, 2022). Neural indicators of social reward processing also were associated with peer conformity, revealing activation of the ventral striatum and anterior cingulate cortex when young adults were exposed to experimentally-manipulated feedback suggesting that peers endorsed responses similar to participants’ own (Nook & Zaki, 2015). While these findings offer important initial evidence regarding social rewards that are concomitant to peer socialization experiences, they unfortunately do not allow for conclusions regarding internal processes that may be predicted by conformity to popular peers in particular.

Preliminary evidence also is available for the notion that peer socialization may be related to self–other comparisons. For instance, many prior correlational and experimental studies in social psychology have revealed that individuals who perceive their engagement in a behavior as similar to that of a role model will be likely to perceive themselves as more similar to that prototype more generally (Gibbons and Gerrard, 2020; Todd & van Lettow, 2016), and this perceived “prototype similarity” predicts adolescents’ later risk behavior (van Lettow, de Vries, Burdorf, & van Empelen, 2016). Recent research in developmental neuroscience similarly suggests the relevance of self–other comparisons, revealing that adolescents’ exposure to peer feedback is associated with activation of mentalizing (i.e., theory of mind) neural circuitry, such as the medial prefrontal cortex and temporoparietal junction (Van Hoorn, Van Dijk, Güroğlu, and Crone, 2016; Welborn, Lieberman, Goldenberg, Fuligni, Galván, & Telzer, 2015). Extending this work, Reitz, Motti-Stefanidi and Asendorpf (2016) revealed that higher ratings of peer-perceived popularity were longitudinally associated with increases in self-esteem, via adolescents’ self-perceived popularity. These results suggest that adolescents are attuned to positive social approval from high-status peers, and this social approval leads to more favorable perceptions of adolescents’ own social standing among their peers, and eventual increases in self-esteem. Still, evidence is lacking regarding the processes by which peer conformity to desirable prototypes may be associated with greater perceived similarity to desirable high-status peers, and improvements in self-esteem.

The current study used a well-established experimental chat room paradigm (e.g., Cohen & Prinstein, 2006) to examine adolescents’ conformity to high- versus low-status peers in vivo. This study focused on youth in early adolescence, a critical developmental stage associated with greater conformity to peers (Steinberg & Monahan, 2007). Early adolescents were exposed to injunctive norms communicated by electronic confederates (e-confederates) endorsing high levels of aggressive and risk-taking behavior; subsequently, they were offered the chance to report their own aggressive and risk-taking intentions, ostensibly in the virtual presence of others. Immediately following their participation in the chat room, participants privately reported their identity alignment with popular peers and their self-esteem. To examine the proposed mechanisms of peer socialization, it was hypothesized that greater conformity to high-status peers would be associated with increases in identity alignment with popular peers, which in turn would be associated with increases in self-esteem. In contrast, conformity to low-status peers was expected not be associated with changes in identity alignment with popular peers or self-esteem. These hypotheses pertain to novel research questions, were not pre-registered, and thus should be considered exploratory. However, the use of an experimental paradigm and short-term longitudinal data enabled directional hypotheses heavily informed by previous research and theory.

Empirical data suggest differential socialization effects for boys and girls (Steinberg & Monahan, 2007; Stevens & Prinstein, 2005), and thus, the mechanisms of peer influence processes may differ by gender. Indeed, boys may be more susceptible to peer influence towards externalizing behavior (i.e., risk-taking behavior; Parsai et al., 2009; Steinberg & Monahan, 2007) than girls, while a reverse pattern may be evident for internalizing symptoms (Stevens & Prinstein, 2005). Additionally, research has revealed that males typically have higher baseline levels of self-esteem, and report higher levels of subjective well-being (Li et al., 2022), and further suggest that the peer correlates of self-esteem differ for boys and girls (e.g., relationship processes; Rose & Rudolph, 2006). Moreover, the experimental paradigm was designed somewhat differently for boys and girls in the current study (see Method section). Thus, the research questions were examined separately by gender, as in past work using the same experimental chat room paradigm (Choukas-Bradley et al., 2015).

Method

Participants

The participants were 250 adolescents (55.2% male; M_age = 12.70 years; SD = 0.56) at two rural, low-income middle schools in the southeastern United States. The sample was racially and ethnically heterogeneous (40.3% White, 28.2% Black, 23.4% Hispanic/Latino, 7.7% Multiracial/Other). According to school records, approximately 67% of students in the school district were eligible for free or reduced-price lunch. The participants were enrolled in a project examining peer influences on adolescents’ risk behaviors beginning in the spring of 2012 (see Choukas-Bradley et al., 2015). All seventh and eighth grade students from three schools in a single county were recruited for participation in the overarching project, with the exception of students in self-contained special education classes. Each student’s family received a letter of consent, with an option for parents to grant or deny consent. A variety of incentives were used to ensure the return of consent forms, at the level of individual adolescents, teachers, and schools (e.g., a $10 gift card was given to each student who returned the form, regardless of whether it provided or denied consent to participate). Because the experimental chat room paradigm used in this study was a time-intensive procedure that included deceptive elements, for feasibility, a subset of adolescents was selected for participation in the chat room – specifically, seventh graders from two of the three schools. Importantly, this rural school district randomly assigns students across these three centrally located middle schools, and analyses revealed there were no significant differences in students’ gender, ethnicity, or socioeconomic status among those eligible or ineligible for the experimental paradigm. The flowchart in Figure 1 illustrates how the final sample of 250 adolescents was selected.

Figure 1. Flowchart for Sample Determination.

Procedures

Youth provided assent to participate at baseline. The university human subjects committee approved all procedures, including the debriefing process. All data were collected in participants’ schools, with the use of privatizing dividers around laptop computers. The participants were compensated with a total of $30 in gift cards for participation in the assessments described in this study. In the late spring of seventh grade, adolescents completed sociometric assessments and self-report questionnaires, which were used as pre-experimental assessments. The experimental chat room paradigm was administered in early fall of eighth grade.

Measures

Demographic factors.

Adolescents were asked to report their age (in whole numbers) and race or ethnicity (African-American/Black, Asian, White/Caucasian [not Latino], Hispanic/Latino, Other). Data from school records also were utilized to code race and ethnicity; in cases of discrepancies between self-report data and school-records, self-report data were used.

Sociometric assessment.

A standard sociometric assessment was conducted with all 350 initial participants (both male and female) in seventh grade at the two selected schools, to measure adolescents’ peer status (popularity and likeability) and friendships. The participants were provided with five alphabetized rosters of all grademates, each counterbalanced (A-Z; Z-A). On two of these rosters, participants nominated an unlimited number of peers who were “most popular” and “least popular,” respectively. For each adolescent, the number of nominations received was computed and standardized within grade. Consistent with standard procedures, a difference score was computed between standardized “most popular” and “least popular” nominations and re-standardized to obtain a measure of peer-perceived popularity, with higher scores indicating greater popularity among peers (e.g., Prinstein & Cillessen, 2003). On two separate rosters, adolescents nominated an unlimited number of peers whom they “like the most” and “like the least”; the same process yielded a standardized difference score reflecting likeability (i.e., peer acceptance/rejection) among peers (Coie & Dodge, 1983). On a fifth roster, participants selected an unlimited number of students who were their “closest friends” and then specified a “very best friend” and two additional “best friends” from this selection. Sociometric nomination procedures are considered the most reliable and valid measures of peer status and friendship nominations (see Rubin, Bukowski, & Laursen, 2009). These sociometric peer status and friendship nominations were not used directly in the primary analyses for the current study. Rather, these data were used in the construction of the experimental paradigm. More specifically, as is described in detail below, these data were used to create electronic confederates (e-confederates) with different levels of implied peer status.

Self-esteem and identity alignment with popular peers.

At baseline, and again immediately following the completion of the hypothetical scenarios (and “logging out” from the chat room), participants completed measures of their self-esteem and identity alignment with popular peers. Self-esteem was assessed with a one-item measure (see Robins, Hendin, Trzeniewski, 2001, for a similar measure) commonly used to assess state self-esteem within experimental paradigms, based on Rosenberg (1979): “On the whole, I like myself right now.” They answered using a 9-point Likert scale in which they indicated the extent to which this statement was true for them, from 1=extremely untrue to 9=extremely true. Higher scores reflected higher self-esteem. Due to the original distribution of scores being negatively skewed, the self-esteem variable was first reflected, then log transformed.

Identity alignment with popular peers was assessed with three items, including: “I am very similar to the most popular kids in my school”; “I share the same interests and values as the most popular kids in my school”; “If most of the popular kids in my school really knew me, they would think I’m just like them.” As with the self-esteem measure, responses were made along 9-point Likert scales ranging from 1=extremely untrue to 9=extremely true. Responses were averaged (Cronbach’s αs = .79, .81 at pre-experiment, .76, .85 at post-experiment for boys and girls, respectively), with higher scores indicating higher levels of identity alignment with high status peers.

Hypothetical scenarios.

Seven hypothetical scenarios, adapted from previous work demonstrating the reliability and validity of such scenarios (e.g., Prinstein, Brechwald, & Cohen, 2011), were used to assess adolescents’ endorsement of aggressive, deviant, and substance use behaviors. These scenarios were developed in collaboration with focus groups of middle school students. The scenarios depicted hypothetical situations in which adolescents were asked how likely they would be to engage in a variety of behaviors related to deviance, aggression, or substance use. For example, one scenario read: “You are hanging out at a student’s house and he says it would be fun if you both drank some alcohol from his parents’ liquor cabinet. How likely would you be to drink some alcohol?” Following each scenario, the participant reported, using a 9-point Likert scale, how likely they would be to engage in the behavior (1=not at all likely to 9=definitely).

For the current study, as in prior work (e.g., Cohen & Prinstein, 2006), this instrument was used in three ways. First, it was administered to all adolescents selected to participate in the chat room paradigm. From this baseline assessment, normative (i.e., mean) responses to each scenario were identified by gender. For each scenario, this information was then used to determine the response that was “above average” (i.e., +1 SD) in its level of risky or aggressive endorsement, such that these risky responses could be attributed to e-confederates in the chat room paradigm (see Experimental Paradigm). Second, these scenarios were used again in the context of the chat room paradigm. Specifically, as is discussed in detail below (see Experimental Paradigm), for each scenario, the three “peers” in the chat room (i.e., e-confederates) responded with the above average risky or aggressive response determined based on the baseline assessment, and the participant’s subsequent response constituted the dependent variable. Third, the participant’s baseline (i.e., pre-experiment) scores on this instrument were included in analyses assessing the effects of the experimental manipulation.

For each participant’s baseline and chat room responses, responses to the seven hypothetical scenarios were combined to create a mean score, with good internal consistency (Cronbach’s αs = .84, .82 at pre-experiment, .80, .81 at chat room assessment for boys and girls, respectively).

Experimental paradigm.

The experimental paradigm simulated an Internet chat room. For a thorough description of the paradigm, see Cohen and Prinstein (2006). The description below focuses on aspects of the procedure necessary for an understanding of the current study.

The participants were informed that they would be participating in a study of how teens communicate through the Internet, and that they would be communicating via an Internet chat room with three same-gender students in their grade who were working on computers in other rooms of the school. In reality, however, the three “peers” in each chat room were preprogrammed e-confederates, computer-generated using Direct RT software (Jarvis, 2004). The participants were randomly assigned to a high-status or low-status peer condition; the peer status of each e-confederate was manipulated such that participants were led to believe they were interacting with high-status or low-status peers from their school and grade. Specifically, for each e-confederate, peer status was implied with two types of information that were provided on chat room screens and connected to the e-confederate’s identity: 1) the names of two ostensible “friends” of the e-confederate who were of either high or low peer status (presented in the format of the friend’s first name and last initial, and determined based on the sociometric data described above), and 2) two hobbies associated with either high or low peer status (determined based on focus groups). Notably, sociometric friendship nominations were used to ensure that adolescents were not placed in chat rooms with e-confederates representing members of their friend groups, given the potential for in-group bias to alter the magnitude of peer influence effects (Stallen, Smidts, & Sanfey, 2013).

A variety of methods increased the verisimilitude of the chat room. For example, graphic designers created screens that appeared to depict a live internet chat room space; participants were shown a screen that mimicked the researcher’s university homepage; and screens were used that appeared to depict “logging in” and downloading processes. Additionally, in order to account for why e-confederates’ friends and hobbies appeared on the screen, participants were asked to provide information about their own friends and hobbies, which appeared on the screen throughout the chat room procedure and ostensibly helped the “peers” in the chat room to get to know one another.

Assessment of peer conformity within the chat room.

After participants had “logged in” and received an orientation to the chat room, they were exposed to the same set of hypothetical scenarios to which they had previously responded at baseline. During the chat room, the scenarios appeared one by one on the screen. Participants were told that they had been randomly assigned to respond to all items last. When a scenario appeared, e-confederate #1 provided their response, followed by e-confederate #2, then e-confederate #3, and finally the participant. Thus, for each scenario, the participant was exposed to the e-confederates’ high-risk or aggressive responses prior to providing their own response. For verisimilitude, the predetermined e-confederates’ responses varied slightly across e-confederates and across items, but were consistently above average in their likelihood of engaging in the behavior. For each scenario, after the participant viewed the e-confederates’ responses, they reported the likelihood that they would engage in the behavior (1–9 scale discussed above); the participant’s response appeared on the screen, such that the participant believed their “peers” (i.e., the e-confederates) could see their responses. As noted previously, participants’ responses to the seven hypothetical scenarios were averaged, with these mean scores used in analyses.

Gender differences in chat room construction.

Two gender differences in how the chat room was conducted should be noted, and necessitated the hypotheses discussed herein to be examined separately by gender (see Choukas-Bradley et al., 2015). The first difference concerned the purported race and ethnicity of the e-confederates; the friend names that appeared in the girls’ chat room included Black, Latinx, and White adolescents. In the boys’ conditions, however, it was not possible to identify a sufficient portion of students from ethnic minority groups who consented to participate, were perceived to be of high peer status, and had identified other consented friends perceived to be of high peer-status. Thus, it was decided to include only White e-confederates in the boys’ chat room condition to minimize potentially confounding effects of race and ethnicity.

The second difference between the chat room conditions across gender concerned peer status; prior work suggests that group-based peer reputations differ by gender (Cillessen & Mayeux, 2004; Mayeux & Kiesner, 2020; Rose & Rudolph, 2006). For instance, examinations of social hierarchies reveal that the correlation between likability and popularity, two metrics of high peer status (Parkhurst & Hopmeier, 1998), significantly decreases over time for girls, but not for boys (Cillessen & Mayeux, 2004 Van den Berg et al., 2020). Consistent with this work, for boys in our sample, it was possible to identify potential e-confederates who were high in both popularity and likability as metrics of peer status (obtained via sociometric nominations); yet, in the current sample, this was not possible for girls, consistent with findings regarding the low correlation between popularity and likability among adolescent girls (Cillessen & Mayeux, 2004).

Due to these two differences, it was not possible to create identical chat room conditions for boys and girls. For boys, two chat room conditions were created, consistent with original hypotheses: a high popularity/high likability condition (high peer-status condition; N = 70), and a low popularity/low likability condition (low peer-status condition N = 68). In contrast, for girls, three conditions initially were created: high-popularity/low-likability (n = 58), low-popularity/high-likability (n = 54), and low-popularity/low-likability (n = 54). A manipulation check suggested that conditions varying on e-confederates’ likability were not reliably perceived as different among participants. However, a second manipulation check revealed that participants successfully differentiated between two conditions of e-confederates based on popularity only: high-status (high-popularity/low-likability; N = 58) and low-status (low-popularity/low-likability; N = 54). Thus, the girls’ low-popularity/high likability condition (n = 54) was dropped from the analysis. Notably, analysis of girls thus only examined conditions of conformity to e-confederates that were low in likability.

Manipulation check.

At the end of the chat room, participants completed two additional measures to examine the effectiveness of the experimental manipulations. Specifically, participants were asked to report on a 7-point Likert scale how popular (1=one of the least popular, 7=one of the most popular) and how liked (1=one of the least liked, 7=one of the most liked) they perceived each of the three “peers” (i.e., e-confederates) to be, relative to grademates. Subsequently, each participant’s responses for the three e-confederates were averaged to create measures of their perceptions of the e-confederates’ popularity and likeability. Across both boys’ conditions, this measure was internally consistent for the assessment of perceived e-confederates’ likeability α = .85, and popularity α = .89), suggesting that boys saw all three e-confederates within their conditions as comparably likeable and comparably popular. Average scores of e-confederates’ likeability and e-confederates’ popularity thus were computed. T-tests examining the effectiveness of the manipulation between conditions revealed significant effects, suggesting that boys in the high status condition perceived e-confederates to be more likeable (N = 70; M = 4.93, SD = 1.16) than did boys in the low status condition (N = 68; M = 3.42, SD = 1.34; t (136) = 7.12, p < .001, Cohen’s d = 1.20). Similarly, boys in the high-status condition perceived e-confederates to be more popular (M = 4.72, SD = 1.19) than did boys in the low status condition (M = 3.29; SD = 1.25; t (136) = 6.89, p < .001, Cohen’s d = 1.17).

However, among girls, internal consistency was lower (α = .62 for likeability and .70 for popularity) than what was observed for boys. Scores of e-confederates’ likeability and popularity were computed for comparison within the high- and low-status conditions. T-tests examined the effectiveness of this manipulation and revealed that girls in the high-status condition perceived e-confederates to be more popular (N = 58; M = 4.60, SD = 0.96) than girls in the low-status condition did (N = 54; M = 3.31, SD =0.80; t (110) = 7.71, p < .001, Cohen’s d = 1.46). Similarly, girls in the high-status condition perceived e-confederates to be more likeable (M = 4.53, SD = 0.87) than girls in the low status condition did (M = 3.42; SD = 1.03; t (110) = 6.13, p < .001, Cohen’s d = 1.16). This finding was surprising, given that e-confederates in the high- and low-status condition were selected based on near identical likability scores (i.e., social preference; Coie & Dodge, 1983); nevertheless, results suggested that differences in e-confederates’ popularity between conditions may have influenced adolescents’ perceptions of likability.

Debriefing procedures.

All participants were thoroughly debriefed after data collection had been completed for all adolescents (in order to minimize the chances of diffusion). Debriefing used a “funnel” procedure approved by the university human subjects committee. First, the participants were asked to report their general impressions of the study. Next, they were asked more specific questions about their “peers” in the chat room and the perceived purpose of the study. Finally, participants were provided with an explanation of the deceptive elements of the study, including that their chat room interactions had been with computer-generated e-confederates rather than actual peers, and that the e-confederates had endorsed responses that were substantially riskier than the average responses of their grademates.

Data Analyses

Descriptive statistics included computation of means and standard deviations separately for boys in the high- and low-status conditions, and were conducted in SPSS v28.0. The main study hypothesis posited that for adolescents in the high-status condition, higher levels of conformity would be associated with higher levels of change in self-esteem, as mediated by higher levels of change in identity alignment with popular peers. To examine this hypothesis, a series of latent change score models (McArdle, 2009) were conducted within a structural equation modeling framework using M-Plus 8.4 (Muthen & Muthen, 2017). First, three unconditional latent change score models were estimated to examine mean changes in conformity, identity alignment with popular peers, and self-esteem across conditions from pre- to post-experiment. Each latent change score was created by regressing the post-experiment variable onto the pre-experiment variable with factor loadings fixed to 1, constraining the intercept of the post-experiment variable to 0, defining a latent change score by the post-experiment variable with a factor loading fixed to 1, and lastly by covarying the change score and the pre-score (as recommended by Kievit et al., 2018). For each unconditional model, latent means and variances were estimated to determine whether the average pre-scores significantly differed from the average post-scores, and whether there was individual variation in the degree to which the pre- and post-scores differed. Wald tests were then conducted to examine whether these latent means and variances significantly differed across the high-status and low-status conditions.

To examine the potential for a mediation model in both conditions, first the self-esteem latent change score was regressed on the conformity latent change score. Subsequently, a conditional latent change score model was then estimated per condition in which the three latent change scores were included into one model to examine whether the association between the conformity change score and the self-esteem change score was mediated by the identity alignment with popular peers change score. Note that, in this model, each latent change score (i.e., pre-experiment aggressive and health risk responses, identity alignment, self-esteem at pre-experiment) was regressed on the corresponding pre-experiment score to account for individual baseline differences. Next, the indirect effect of conformity on changes in self-esteem via changes in identity alignment with popular peers was estimated in both conditions by examining whether the bias-free 95% confidence interval (generated via 5000 bootstrap resamples) excluded zero (see Figure 2, for the theoretical model). Finally, a Wald Test evaluated whether the associations among the change scores and the indirect effects significantly differed across conditions.

Figure 2. Final Conditional Model.

Note. β represents standardized regression coefficients; λ represents factor loadings; The β between conformity and changes in self-esteem is the direct effect without considering the mediator, β’ is the direct effect when the mediator is included in the model. The betas and indirect effects associated with this model for boys are reported in the results, and for girls, reported in the supporting information.

Results

Descriptive Analyses

Table 1 presents the means and standard deviations of the main study variables separately by experimental condition (exposure to injunctive norms communicated by high-status vs. low-status e-confederates) and by gender. No significant differences were observed at the pre-experiment assessment on any study variables. In addition, Table 2 presents the correlations among study variables separately by gender.

Table 1.

Means and Standard Deviations of Study Variables by Experimental Condition

	High-status peer condition (N = 701; 582)		Low-status peer condition (681; 542)
	M	SD	M	SD	t	df	p	95% CI	Cohen’s d
Boys: Risk-behaviors
Boys: Pre-Experiment	2.56	1.33	2.63	1.64	−.29	136	.775	[−.58, .43]	.04
Boys: “Public” Chat Room	4.20	1.45	4.11	1.59	.36	136	.721	[−.42, .61]	.06
Boys: Identification with popular peers
Boys: Pre-Experiment	5.44	2.16	5.42	2.12	.04	134	.970	[−.71, .74]	.01
Boys: Post-Chat Room	6.05	1.86	5.92	1.91	.41	136	.684	[−.50, .76]	.07
Boys: Self-esteema
Boys: Pre-Experiment	7.17/−.31	2.33/.34	7.86/−.22	1.78/.29	−1.76	133	.081	[−.20, .01]	.28
Boys: Post-Chat Room	7.71/−.24	2.04/.30	8.25/−.16	1.29/.24	−1.67	136	.098	[−.17, .01]	.29
Girls: Risk-behaviors
Girls: Pre-Experiment	2.65	1.39	2.92	1.56	−.95	110	.343	[−.82, .29]	.18
Girls: “Public” Chat Room	4.34	1.44	4.29	1.55	.200	110	.841	[−.50, .62]	.04
Girls: Identification with popular peers
Girls: Pre-Experiment	5.29	2.18	5.46	2.02	−.44	110	.659	[−.97, .61]	.08
Girls: Post-Chat Room	5.54	1.96	5.67	2.29	−.33	110	.742	[−.93, .66]	.06
Girls: Self-esteema
Girls: Pre-Experiment	7.05/−.36	2.23/.30	7.26/−.30	2.40/.34	−1.03	110	.306	[−.18, .06]	.20
Girls: Post-Chat Room	7.43/−.32	1.91/.29	7.15/−.34	2.50/.34	0.45	110	.964	[−.11, .12]	.10

Note. 95% CI = 95% confidence intervals for mean differences

The sample sizes in italics represent the girls’ conditions.

Raw scores of self-esteem were skewed for both boys (pre-experiment: −1.39; post-experiment: −2.26) and girls (pre-experiment: −1.57; post-experiment: −1.40).

^aBoth raw and transformed (i.e., reflected and log-transformed) mean scores of self-esteem are presented in the table (i.e., raw/transformed). Comparisons between the two experimental conditions and the resulting t-tests, 95% CI and Cohen’s ds were calculated using transformed scores.

^bEqual variances were not assumed.

Cohen’s d effect sizes ranged from 0.04 – 0.29 and thus should all be considered to be small.

Table 2.

Correlation Matrix of Study Variables by Gender

	1	2	3	4	5	6
Pre Aggressive and Risk Taking Behavior	-	.59 **	.17	.22 *	−.09	−.14
Post Aggressive and Risk Taking Behavior	.49**	-	.05	.19 *	−.15	−.12
Pre Alignment with Popular Peers	.29**	.25**	-	.63 **	.31 **	.21 *
Post Alignment with Popular Peers	.22**	.40**	.45**	-	.16	.15
Pre Self-Esteem	−.09	−.15	.17*	.03	-	.37 **
Post Self-Esteem	.02	.09	−.04	.23**	.37**	-

Note. Girls’ correlations are above the diagonal, while boys’ correlations fall below the diagonal.

^**Correlation is significant at the 0.01 level.

^*Correlation is significant at the 0.05 level (2-tailed)._*

“Pre” indicates that the variable was measured prior to the chat room, and “post” indicates it was measured after the chat room.

Unconditional Latent Change Score Models: Changes From Pre- to Post-Experiment

To examine whether there were mean changes and individual variation in the magnitude of the change from pre- to post-experiment in aggressive and risk-taking behavior, identity alignment with popular peers, or self-esteem, three unconditional latent change score models were estimated among boys in the sample. A first unconditional model examining changes in aggressive and risk-taking behavior (i.e., conformity) revealed a significant positive mean for the latent change score in both the low- and high-status condition. These results indicate that on average, boys reported significantly higher endorsements of aggressive and risk-taking behavior following exposure to injunctive norms ostensibly communicated by both high- as well as low-status peers. Significant variance also was observed in both groups, indicating individual differences in the extent to which boys conformed to their peers’ injunctive norms. A nonsignificant Wald Test suggested that, contrary to prior work with late adolescents, adolescents conformed similarly to both high-status and low-status peers. An identical pattern of results was revealed among girls (Table 3).

Table 3.

Latent Means and Variances of the Unconditional Latent Change Score Models

	High-status peer condition (N = 70; 58)		Low-status peer condition (N = 68; 54)
	Latent Mean (SE)	Latent Variance (SE)	Latent Mean (SE)	Latent Variance (SE)	Wald Test (df)
Boys: Conformitya	1.64** (.17)	2.0** (.34)	1.48** (.19)	2.52** (.43)	1.33 (2)
Boys: Identity Alignment with Popular Peersa	.61** (.26)	4.83** (.82)	.53* (.24)	3.90** (.68)	.81 (2)
Boys: Self-Esteema	.08 (.04)	.11** (.02)	.05 (.04)	.11** (.02)	.234 (2)
Girls: Conformitya	1.68** (.18)	1.75** (.36)	1.37** (.18)	1.76** (.34)	1.55 (2)
Girls: Identity Alignment with Popular Peersa	0.15 (.23)	2.88** (.54)	.21 (.29)	4.39** (.84)	2.28 (2)
Girls: Self-Esteema	.07 (.05)	.11** (.02)	−.02 (.05)	.15** (.03)	2.78 (2)

Note. 95% CI = 95% confidence intervals for mean differences

^aLatent Change Scores per variable.

^**p <.01.

^*p < .05.

The sample sizes in italics represent the girls’ conditions.

Similarly, the unconditional model examining changes in identity alignment with popular peers from pre- to post-experiment revealed significant means and variances of the latent change scores in both the low- and high-status condition. In other words, identity alignment increased from pre- to post-chat room for all adolescents, and there was significant variability around this mean, but no differences between high and low status conditions (Table 3). Unlike for boys, for girls there was a nonsignificant mean in both the low- and high- status condition, indicating that on average, there were no changes in female adolescents’ identity alignment with popular peers from pre- to post-experiment. However, significant variance in both conditions revealed that there was significant variation in within-person changes in identity alignment from pre- to post-experiment for female participants in both conditions.

Finally, the unconditional model examining changes in self-esteem revealed a nonsignificant latent mean in both the low- and high-status condition, indicating that, on average, there were no changes in self-esteem from pre- to post-experiment. However, significant variance was observed in both conditions, indicating significant variability in changes in self-esteem from pre- to post-experiment for participants in both conditions; however, this did not vary across conditions. An identical pattern of effects was revealed for female participants (see Table 3).

Conditional Latent Change Score Models

Prior to generating the conditional latent change score model, the association between the latent change scores for conformity and self-esteem was examined to assess the potential for mediation in both the high- and low-status conditions. There was a positive total effect between greater conformity and the latent change score for self-esteem in the high-status condition (standardized β = .20, SE = .09, p = .034) but not the low-status condition (β = −.02, SE = .09, p = .808). These results indicate that greater conformity to popular, but not unpopular, e-confederates was significantly associated with greater increases in self-esteem from pre- to post-experiment. However, a Wald-test revealed that these groups only marginally differed, and therefore group-level differences in this direct effect should be interpreted with caution (χ2(1) = 3.23, p = .07). For girls, however, this examination revealed that there was neither a significant association between greater conformity and the latent change score for self-esteem in the high-status condition (standardized β = .09, SE = .11, p = .427) nor the low-status condition (β = .06, SE = .11, p = .59). A Wald test indicated that these null associations did not differ by condition (χ2(1) = 0.02, p = .899).

Model building and model fit.

Next, to examine the primary study hypothesis among boys and girls, a conditional latent change score model was estimated to examine whether the association between higher levels of conformity in the high-status condition, but not the low-status condition, and the corresponding greater increases in self-esteem in the former than the latter condition, was mediated by greater increases in identity alignment to popular peers from pre- to post-experiment. The conditional model examined associations among latent change (i.e., pre- to post-chat room) scores for conformity (i.e., using pre- and post-chat room responses to aggression and health risk behavior scenarios), identity alignment, and self-esteem. Means and variances at pre-experiment were all fixed to be equal across conditions, as were the associations among the pre-experiment scores. These model constraints did not negatively impact model fit (boys: χ2(9) = 12.33, p = .196; girls: χ2(9) = 6.88, p = .650). The fit of this final model for girls was acceptable; however, among boys, the model fit of this initial conditional model was not satisfactory (see Table 4). Therefore, based on the modification index, the latent change score of identity alignment with popular peers was regressed on the pre-experiment aggressive and risk-taking behavior. This result suggested that in the high-status condition, but not in the low status condition (χ2(1) = 4.39, p = .036), adolescents’ higher endorsement of aggressive and risk-taking behavior pre-experiment was associated with greater increases in identity alignment with popular peers from pre- to post-chat room (high-status condition: (β = .37, SE = .09, p < .001); low-status condition: (β = .09, SE = .11, p > .05). While the inclusion of this path to the model significantly improved the fit of the model, the pattern of results was identical to the ones without the inclusion of this path. Note that while the RMSEA exceeds traditionally acceptable cut-off points, it has been suggested that well-fitting models with few degrees of freedom and limited samples sizes may have RMSEA values exceeding .05, particularly if the chi-square test statistic and relative indices of fit suggest good fit (Curran et al., 2003). The chi-square test statistic was nonsignificant, and the relative fit indices are within the recommended range; the fit of this final model was therefore deemed acceptable (Table 4).

Table 4.

Fit Statistics for the Conditional Latent Change Scores

Models	Chi-Square	df	CFI	TLI	RMSEA
1 - Boys	29.23*	12	.867	0.734	.144
2a - Boys	28.80	19	.924	0.905	.086
3 - Girls	9.11	12	1.00	1.00	0.00

Note. NFI = Normed Fit Index; CFI = Comparative Fit Index; TLI = Tucker-Lewis Index;

RMSEA = Root Mean Square Error of Approximation.

^aBased on the modification index from Model 1, the path of pre-experiment aggressive and risk-taking behavior on the latent change score of identity alignment with popular peers was added to the model to improve model fit, but the inclusion of this path did not change the results in either model.

^*p <.05.

Final conditional model.

In the final model among boys (Figure 2), as hypothesized, a significant positive association was found between conformity and the latent change score for identity alignment with popular peers in the popular condition (β = .50, SE = .08, p < .001), but not the unpopular condition (β = .05, SE = .11, p = .649). This association indicated that boys who conformed more to the high-status e-confederates also showed greater increases in their identity alignment with popular peers. A Wald test indicated that this association was significantly different across conditions (χ2(1) = 13.31, p <.001). Moreover, a significant positive association was found between the latent change scores for identity alignment with popular peers and self-esteem in both the popular (β = .28, SE = .19, p = .004) and unpopular (β = .20, SE = .08, p = .015) conditions. This association suggested that boys who reported larger changes in identity alignment with popular peers from pre- to post-experiment also reported stronger increases in their self-esteem from pre- to post-experiment. A Wald test indicated that this association did not significantly differ across conditions (χ2(1) = .07, p = .785). Finally, when changes in identity alignment with popular peers entered the relation between conformity and changes in self-esteem, the direct effect was no longer significant in the popular (β = .09, SE = .1, p = .366) or unpopular (β = −.02, SE = .08, p = .796) condition. Again, a Wald test indicated that this association did not significantly differ across conditions (χ2(1) = .78, p = .38).

In the final model for girls, there was not a significant association between conformity and the latent change score for identification with popular peers in the popular condition (β = −0.07, SE = .12, p = .57), or the unpopular condition (β = 0.09, SE = .13, p = .46). This association indicated that, contrary to hypotheses, girls who conformed more to the high- or low-status e-confederates did not show greater increases in their identification with popular peers. A Wald test indicated that this association was not significantly different across conditions (χ2(1) = 0.89, p = .34). However, there was a significant negative association between the latent change scores for identification with popular peers and self-esteem for girls in the high-status condition (β = −.33, SE = .11, p = .002), but not the low-status (β = .08, SE = .12, p = .473) condition. A Wald test indicated that this association did significantly differ across conditions (χ2(1) = 7.26, p < .01). Thus, contrary to hypotheses, greater identification with popular peers was associated with decreases in self-esteem following the chatroom for girls in the high-status condition, but not the low-status condition. Finally, when changes in identification with popular peers were included in the model examining associations between conformity and changes in self-esteem, the direct effect was not significant in the popular (β = .09, SE = .11, p = .431) or unpopular (β = .05, SE = .11, p = .659) condition. Again, a Wald test indicated that this association did not significantly differ across conditions (χ2(1) = .03, p = 862).

For a significant mediational effect to be considered, the indirect effect of conformity on changes in self-esteem via changes in identity alignment with popular peers needs to produce a bias corrected 95% confidence interval from 5000 bootstrap resamples which excludes zero. For boys, the bias corrected standardized estimate for the indirect effect in the popular condition was 0.41 with a 95% confidence interval ranging from 0.10 – 0.82 which excluded zero. In the low-status condition, the bias corrected estimate for the indirect effect was 0.03 with the 95% confidence interval ranging from −0.11 – 0.20 which includes zero. A final Wald-test comparing indirect effects across conditions was significant (χ2(1) = 4.59, p = .032). Thus, as hypothesized, changes in identity alignment with popular peers fully mediated the relation between greater conformity and larger changes in self-esteem for boys in the popular condition, but not for boys in the unpopular condition. For girls, the bias corrected estimate for the indirect effect in the popular condition was −0.01, with a 95% confidence interval ranging from −0.05 – 0.11, which includes zero. In the low-status condition, the bias corrected estimate for the indirect effect was 0.01 with the 95% confidence interval ranging from −0.04 – 0.09 which includes zero. A final Wald-test comparing indirect effects across conditions was insignificant (χ2(1) = 0.20, p = .653). Thus, for girls, changes in identification with popular peers was a not mediator for greater increases in conformity on changes in self-esteem for adolescents in either the high- or low-status condition.

Sensitivity Analyses

Lastly, the results presented above regarding possible temporal associations among the variables are limited because identification with popular peers and self-esteem were measured concurrently. Thus, sensitivity analyses were conducted to examine a plausible alternative model: whether an association between conformity and identification with popular peers was mediated by self-esteem. For boys, this model did not fit the data well (χ2(19) = 36.20 p <.05, RMSEA = .115 CFI = .867 TLI = .832), and modification indices revealed no paths which resulted in a substantial improvement in model fit (Δχ2(1) > 10). Further, the indirect effect included zero and was thus nonsignificant (bias corrected estimate: 0.112, 95% CI −0.040 – 0.370), suggesting that conformity did not lead to changes in alignment with popular peers via change self-esteem. An identical pattern was revealed among the girls’ data with a better fitting model (χ2(19) = 28.29 p <.05, RMSEA = .077 CFI = .956 TLI = .944), and the indirect effect was not significant (bias corrected estimate: −0.01, 95% CI −0.08 – 0.05). Poor fit to the data and nonsignificant indirect effects provide further support for the hypothesized model, which examined alignment with popular peers as a mediator rather than an outcome.

Discussion

Substantial research has revealed that peer influence is a remarkably powerful process (see Giletta et al., 2021 for a recent meta-analysis), but few empirical studies have offered evidence for why peer influence occurs. Theories and research from multiple psychological disciplines provide initial explanations, but more work is needed to fully clarify peer influence mechanisms. Seminal developmental work suggests that adolescents may be most at risk to conform when their deviant behavior is positively and behaviorally reinforced (Dishion et al., 1996). Theories from social psychology simultaneously suggest that individuals conform to others’ behavior to elicit favorable self–other comparisons (Abrams & Hogg, 1990; Ajzen & Manstead, 2007; Fishbein & Azjen, 1977), especially when the behaviors of high-status prototypes are emulated (Gibbons et al., 1998). However, this perspective rarely has been applied to developmental science including youth, among whom peer influence effects often are observed. Despite the ever-increasing methodological and analytical sophistication of longitudinal designs (e.g., separation of between and within-person effects, covariation between constructs across multiple timepoints), strong claims cannot be made due to the inherent correlational nature of these examinations. The use of an experimental paradigm to assess adolescent conformity improves upon these limitations, and provides exciting opportunities to examine why adolescents conform to others’ behaviors, and whether the mechanism differs based on who they are conforming to.

Study Findings

Results from this study on early adolescents begin to better elucidate poorly understood processes of peer influence, and synthesize prior research and theory. Consistent with the primary study hypothesis, boys who conformed more to high-, but not low-status e-confederates, showed greater increases in their identity alignment with popular peers and concomitant increases in self-esteem. Not only do adolescents conform to peers for rewarding feedback (Dishion et al., 1996; Kwon & Telzer, 2022; Somerville, 2013), and to improve their own self–other comparisons (Azjen & Manstead, 2007), but these data further suggested that adolescents’ conformity to high-status peers may confer improvements in their own self-esteem. It must be noted, however, that this association was present under relatively strict conditions. The association between conformity and increases in self-esteem occurred via identity alignment with popular peers, suggesting that conformity alone was not sufficient to elicit increases in self-esteem. These results are generally consistent with theories proposed by social psychologists (Fishbein & Azjen, 1977).

Interestingly, results suggested that early adolescent boys and girls in the low-status condition also conformed within the chat room paradigm, at levels comparable to the conformity observed among adolescents in the high-status condition. This finding contrasted with prior work utilizing similar methods to assess conformity within a later adolescent sample (Cohen & Prinstein, 2006). Collectively, these results suggest an important need to understand developmental differences in susceptibility to the influence of different sources. Perhaps due to the increases in general susceptibility to peer influence as adolescents age (Steinberg & Monahan, 2007), as well as an increased emphasis on peer status (LaFontana & Cillessen, 2010), it may be that adolescents may become more discerning about who they conform to as they mature. Most relevant to this study, conformity to low status peers (i.e., male or female e-confederates) was not associated with greater alignment to popular peers, nor with changes in self-esteem. These results suggest that conformity may occur for multiple reasons, with different mechanisms at play based on whom adolescents are conforming to (see Giletta et al., 2021). It may be that adolescents’ conformity to low status peers occurs in public settings to avoid the perception of non-conformity, but perhaps these results would not sustain over time. This is an important direction for future research.

Also, interesting – and unexpected – results revealed that boys’ own tendencies towards aggressive or risky behavior, as indicated by their pre-experiment responses to hypothetical scenarios, was a predictor of their identification with popular peers. Findings are consistent with research suggesting a strong association between aggressive behavior, risk behavior, and high levels of popularity (Prinstein & Cillessen 2003; Prinstein et al., 2011). However, in the context of findings from this study, this result further underscores the remarkably important and potentially recursive links between youths’ peer status and their engagement in aggressive and health risk behaviors. Those who engage in these behaviors are more likely to see themselves as popular in their peer group, those who are exposed to high status peers are likely to conform to aggressive and health risk behaviors, and following conformity, adolescents regard themselves as even more fully aligned with popular peers (which in turn leads them to feel even better about themselves). At least within cultures wherein aggression is associated with high peer status (Choukas-Bradley et al., 2019), the very notion of popularity appears to be intertwined with motivations to engage in, or conform to these maladaptive and potentially dangerous behaviors.

Gender Differences

Unlike for boys, findings among girls from the current study were not aligned with study hypotheses or identity-based theories of peer influence. The results suggest that among girls, conformity to aggressive and high-risk behaviors communicated by either high- or low-status e-confederates was unrelated to self-esteem, whether examined directly or indirectly via identification with popular peers. This gender difference may exist for several reasons; for instance, the current study utilized conformity to aggressive and risk-taking behavior to operationalize peer influence, a domain which prior research has suggested girls are less susceptible to than boys (Parsai et al., 2009; Steinberg & Monahan, 2007). Girls’ self-esteem may not hinge on conformity to these types of behaviors, but rather to behaviors which are more prevalent among adolescent girls, such as prosocial behavior (Choukas-Bradley et al., 2016; Crone et al., 2022). The hypotheses examined herein should be assessed among conformity to more gender-appropriate behaviors, and such work represents an area of future research.

Surprisingly, results further revealed that for girls in the high-status condition only, identity alignment with popular peers was associated with decreases in self-esteem, but unrelated to conformity. Beyond those discussed above, there are several additional gender differences which may account for these disparate findings. First, physical attractiveness is a much more salient feature of popularity for girls than it is for boys (Mayeux & Kleiser, 2020) and girls’ digital media interactions tend to involve highly visual social media with a focus on physical appearance (Choukas-Bradley, Roberts, Maheux, & Nesi, 2022). The digital interaction simulated in the current study was entirely text-based, without images, and thus may not have been able to optimally capture peer influence processes towards high-status female prototypes. Note that when physical representations of body shapes are depicted in this chat room paradigm, girls are indeed influenced towards or away from weight-related behaviors by peers (Rancourt et al., 2014) Additionally, as noted above, the high-status condition for girls consisted of e-confederates who were selected for being high in popularity (i.e., social reputation) but low in likability (i.e., social preference; Coie & Dodge, 1983); it is not entirely surprising that girls did not report boosts in self-esteem following conformity to unlikable peers.

Future Directions and Limitations

Using an experimental paradigm, the current study provides an important advance in understanding why peer influence occurs for adolescent boys. Still, future research would benefit by addressing the limitations of this study. For instance, the findings may not generalize to real-world contexts; adolescents conformed to high-status peers insofar as their own attitudes towards aggressive and risky behavior grew more positive from pre- to post-experiment, but future studies should examine whether these attitudinal changes manifest in behavioral changes as well. For example, future studies should examine whether exposure to injunctive norms from high-status peers or e-confederates leads to increased externalizing behavior and thus changes in self-esteem, although such an examination may be difficult to ethically implement. Second, regarding the measurement of identity alignment with popular peers, the applicability of some of the items may have differed based on adolescents’ own peer status. Average or less popular peers may have interpreted some of these items differently than those already popular.

Third, the boys’ condition only included White e-confederates in a racially and ethnically heterogeneous sample. For youth of color, peer status may differ based on the nomination context – i.e., when examined within the whole peer context versus only among youth of the same race or ethnicity. Further, it may be that youth differentially conform to high-status youth of the same race vs. other racial and ethnic identities, and such conformity may have varying implications for youths’ self-esteem. These areas represent an understudied, yet sorely needed avenue for future work.

While the current study was able to employ a short-term longitudinal design to examine a mediation model, several methodological limitations should be noted. For instance, the self-report measures of identity alignment with popular peers and self-esteem were measured concurrently, albeit after adolescents were exposed to the chatroom condition. Mediation analyses typically require at least three timepoints to examine adequately the directionality of effects (Baron & Kenny, 1986; Agler & De Boeck, 2017). Within the current study, because the mediator and outcome were examined simultaneously, a statistically plausible alternative model may exist, such that self-esteem could mediate the relation between conformity and identification with popular peers. Sensitivity analyses were conducted to examine this model, and revealed not only poor fit to the data, but nonsignificant direct and indirect relations among the variables. Nevertheless, statements regarding causality should be made with caution. In addition, prior work examining friendship networks and status hierarchies has utilized a range of quantitative and qualitative and ethnographic approaches that are worth considering in future research.

Finally, youth’s susceptibility to peer influence and engagement in risk-taking behavior, while strongest in adolescence, varies throughout developmental stages and across behaviors (Steinberg & Monahan, 2007). Thus, future work should not only attempt to replicate these analyses within children and young adults, but should examine these associations longitudinally, to see how they covary with cognitive, social, and biological development. Indeed, research reveals that increases in self-esteem have resulted in positive changes to behavioral and social development (Hanley & Durlak, 1998), but little is known about the role of conformity in this process.

Overall, the current study provides a crucial next step in explicating peer influence mechanisms and paves the way for future directions in peer influence research. Results offer compelling evidence that boys who conform to their high-status peers experience high levels of alignment with popular peers, which in turn increases their self-esteem. Future research should examine how these mechanisms of peer influence further differ by gender and other demographic characteristics, and whether improvements in self-esteem may dampen the robust effects of peer influence.