Psychology assignment 2

Peer Influence on Gender Identity Development in Adolescence

Olga Kornienko, Carlos E. Santos, Carol Lynn Martin, and Kristen L. Granger Arizona State University

During adolescence, gender identity (GI) develops through a dialectic process of personal reflection and with input from the social environment. Peers play an important role in the socialization of gendered behavior, but no studies to-date have assessed peer influences on GI. Thus, the goal of the present study was to examine peer influences on four aspects of adolescents’ GI in racially and ethnically diverse 7th- and 8th-grade students (N � 670; 49.5% boys, M age � 12.64) using a longitudinal social network modeling approach. We hypothesized stronger peer influence effects on between-gender dimensions of GI (intergroup bias and felt pressure for gender conformity) than on within-gender dimensions of GI (typicality and contentedness). Consistent with expectations, we found significant peer influence on between-gender components of GI–intergroup bias among 7th and 8th graders as well as felt pressure for gender conformity among 8th graders. In contrast, within-gender components of GI showed no evidence of peer influence. Importantly, these peer socialization effects were evident even when controlling for tendencies to select friends who were similar on gender, gender typicality, and contentedness (8th graders only). Employing longitudinal social network analyses provides insights into and clarity about the roles of peers in gender development.

Keywords: gender identity, peer influence, social network analysis, stochastic actor-based modeling, adolescence

Supplemental materials: http://dx.doi.org/10.1037/dev0000200.supp

Although gender development has been a focus of adolescent research for many years, much of this research has been concen- trated on the development of gender stereotypes or attitudes (see Ruble, Martin, & Berenbaum, 2006, for a review) rather than changes in gender-related self-concepts. However, in 2001, with the introduction of a multidimensional view and measure of gender identity (GI) by Egan and Perry (2001), research on GI flourished. This model posits that young children develop a single component of GI—membership knowledge—and by middle childhood, they develop four additional self-concepts, all of which comprise the construct of GI: gender typicality, gender contentedness, inter- group bias, and felt pressure to conform to gender norms. Once reaching adolescence, normative changes occur in these four di- mensions of GI, and individual differences in these aspects of GI

are related to social and psychological adjustment outcomes (e.g., Carver, Yunger, & Perry, 2003; Smith & Leaper, 2006).

As adolescents consolidate their identities, questions about GI become relevant to their emerging sense of self and peer experi- ences may influence how youth think about and experience aspects of GI. Individual differences in GI self-concepts emerge through an increased capacity for making social comparisons (Ruble & Dweck, 1995) and due to pressure from parents or peers to con- form to gender norms (Yunger, Carver, & Perry, 2004). Although some attention has been paid to the role of peers in gender typing (Jewell & Brown, 2014; Smith & Leaper, 2006) and sexism (Leaper & Brown, 2008), and theoretical arguments have been made about the importance of peers in influencing gender devel- opment (Leaper & Friedman, 2007; Maccoby, 1998; Harris, 1995), to our knowledge, no studies have empirically tested peer influ- ence on dimensions of GI. This gap is surprising given a large number of studies documenting the potent role of peer influence on adolescent socioemotional and behavioral development (for re- views, see Brechwald & Prinstein, 2011; Dishion & Tipsord, 2011; Veenstra, Dijkstra, Steglich, & Van Zalk, 2013).

We address this gap by examining peer influence on GI in adolescent friendship networks. Because GI is, by definition, an evolving aspect of the self, its development needs to be considered using longitudinal designs (Egan & Perry, 2001). When examining the role that friendship networks play in contributing to changes in GI self-concepts, it is particularly important to control for how adolescents initially select their friends (to prevent inflated esti- mates of peer influence; Snijders, van den Bundt, & Steglich, 2010). In other words, do friends become similar to one another on dimensions of GI over the course of their friendships (influence), or does this similarity between adolescents and friends result from

This article was published Online First September 1, 2016. Olga Kornienko, Department of Psychology, Arizona State University;

Carlos E. Santos, Counseling and Counseling Psychology, Arizona State University; Carol Lynn Martin and Kristen L. Granger, T. Denny Sanford School of Social and Family Dynamics, Arizona State University.

This research was supported in part by funds provided by the T. Denny Sanford School of Social Dynamics at Arizona State University as part of the Lives of Girls and Boys Research Enterprise (http://lives.clas.asu.edu/). We thank the graduate and undergraduate students who contributed to this project and the students, teachers, principal, staff, and parents for their participation. We are grateful for the helpful comments regarding analyses provided by David R. Schaefer.

Correspondence concerning this article should be addressed to Olga Kornienko, Department of Psychology, Arizona State University, P.O. Box 871104, Tempe, AZ 85287-1104. E-mail: [email protected]

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

Developmental Psychology © 2016 American Psychological Association 2016, Vol. 52, No. 10, 1578–1592 0012-1649/16/$12.00 http://dx.doi.org/10.1037/dev0000200

1578

adolescents preferring to befriend others with similar levels of GI self-concepts (selection)? Answering the former question requires also considering the latter, and involves employing longitudinal designs to study the changes in GI dimensions and friendship networks. Because we are interested in examining contributions of a broader network of social relationships (i.e., multiple friends of the focal individual, friends of friends, etc.; for a review, see Veenstra et al., 2013), we do not restrict our examination to best friendships, which are commonly a focus of developmental re- search (for reviews, see Gifford-Smith & Brownell, 2003; Bagwell & Schmidt, 2011). Accordingly, we use longitudinal social net- work analysis (SNA) methods that have been developed to disen- tangle peer influence from confounding processes of network selection (Snijders et al., 2010). Thus, the main objective of this study is to examine the role of peer influence on changes in dimensions of adolescent GI. We also describe the role of gender and GI for network selection (who adolescents select as friends) and explore gender differences in the magnitude of influence on GI self-concepts.

Development of Multiple Dimensions of GI

The Egan and Perry (2001) model of GI consists of five com- ponents, including: self-identification (typically as ‘female’ or ‘male,’ though it is plausible for a young person to identify outside of the gender binary, and less is known about how such identifi- cation might alter GI development), gender typicality (the degree to which one feels as a typical member of one’s gender group), gender contentedness (the degree to which one is happy with one’s gender group), felt pressure for gender conformity (the degree to which one feels pressure from parents, peers, and self for confor- mity to gender stereotypes), and intergroup bias (the extent to which one believes one’s own gender is superior to the other; Carver et al., 2003). After achieving self-identification early in life, the four remaining dimensions involve adolescents making global, summary judgments about themselves and others, which are constructed through a dialectical process involving personal reflection and input from the social environment (Yunger et al., 2004). Research has shown that GI dimensions are relatively stable but undergo changes (Priess, Lindberg, & Hyde, 2009; Yunger et al., 2004). It is also useful to distinguish between-gender and within-gender facets of GI (Pauletti, Cooper, & Perry, 2014), such that intergroup bias and felt pressure focus more on how an individual relates to the other gender, typicality and contentedness concern how an individual relates to his or her own gender. These different facets find some support in the correlations among di- mensions. For both genders, typicality and contentedness are pos- itively related, but only for girls are intergroup bias and felt pressure related (also related to contentedness; Egan & Perry, 2001; Carver et al., 2003).

Considering within-gender GI self-concepts, gender typicality is the dimension of GI that has received the largest share of empirical attention. To appraise oneself as being gender typical involves considering both the specific and more abstract features of one’s gender to derive a global view of oneself (e.g., “Overall, I am/am not a good fit for my gender category”; Spence, 1985; Egan & Perry, 2001). Younger children may appraise their typicality based on their activities and playmate preferences, whereas older chil- dren may expand their perspectives and take into account person-

ality traits (Egan & Perry, 2001). Additionally, as children grow older, peers take on even greater importance in their social lives (Harris, 1995), and peer group comparisons may begin to contrib- ute to feelings of typicality. Another, albeit less studied, aspect of GI is gender contentedness, which describes the degree to which one feels positive toward one’s gender group. Gender contented- ness is more likely to occur when children are happy with their gender (Carver, Yunger, & Perry, 2003; Egan & Perry, 2001). These within-gender aspects of GI appear to be positively related with psychological adjustment, social skills, and peer acceptance (Carver et al., 2003; Smith & Leaper, 2006; Yunger et al., 2004).

Regarding between-gender dimensions of GI, felt pressure for gender conformity captures the extent to which children feel pres- sure from parents, peers or themselves to conform to gender norms (Egan & Perry, 2001). For instance, this dimension addresses pressures not to behave in cross-gender ways (e.g., for a girl, “Other kids, who are girls like me, would get upset if someone who is a girl acts like a boy”). Finally, intergroup bias describes the extent to which youth tend to associate more positive and fewer negative traits to their own gender group compared with the other gender (Powlishta, 1995; Yee & Brown, 1992; Zosuls, Miller, Ruble, Martin, & Fabes, 2011). These between-gender GI dimen- sions operate through magnification of gender differences and expanded stereotype use (Powlishta, 1995), and can translate into challenges with peer relations and can potentially result poor psychological adjustment (Carver et al., 2003; Smith & Leaper, 2006; Yunger et al., 2004).

The Role of Peer Context and Peer Influence for Development of GI Dimensions

Although peers are theorized to have a major impact on chil- dren’s gender development (Bussey & Bandura, 1999; Maccoby, 1998), few studies have tested their influence, and the few that have, focused on gendered behaviors such as interests and activi- ties among preschool children (see Martin & Fabes, 2001; Martin et al., 2013). No research has examined peer influence on GI development in adolescents. However, existing research has illus- trated links between GI and peer relationships. For instance, stud- ies have revealed how aspects of GI are associated with various dimensions of peer relations including peer popularity, acceptance, teasing, and victimization. Specifically, youth who describe them- selves as being gender typical and being content with their gender tend to be more accepted by, and popular with their peers, whereas adolescents who score low in gender typicality are more likely to be teased by peers (Egan & Perry, 2001; Jewell & Brown, 2014; Young & Sweeting, 2004). Adolescents who report feeling a strong pressure to conform to gender norms tend to become less accepted by their peers over time (Yunger et al., 2004). Felt pressure for gender conformity has been found to mediate the relation between gender atypicality and peer victimization, under- scoring the importance of peer dynamics surrounding GI develop- ment for adolescent psychosocial adjustment (Drury, Bukowski, Velásquez, & Stella-Lopez, 2013). Finally, little is known about peer relations correlates or antecedents of gender intergroup bias. However, evidence regarding gender-related biases (e.g., ho- mophobic biases; Poteat, 2007) has documented peer contributions in influencing such attitudes.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1579PEER INFLUENCE ON GENDER IDENTITY

The notion that the peer context may shape emerging self- concepts of GI is consistent with a large body of evidence docu- menting that the peer group is a potent source of influence for adolescents (Brechwald & Prinstein, 2011; Dishion & Tipsord, 2011). Adolescents spend increasingly higher amount of time in the company of their peers (Larson & Richards, 1991), whose motivational and affective salience is heightened (Parker, Rubin, Erath, Wojslawowicz, & Buskirk, 2006), making peer experiences relevant for shaping GI. Youth may be susceptible to peer influ- ence due to an increased capacity for making social comparisons (Ruble & Dweck, 1995), or due to increased peer pressure to conform to social norms (Brechwald & Prinstein, 2011; Yunger, Carver, & Perry, 2004). Specifically, social comparisons to own- gender peers allow youth opportunities to weigh how they com- pare to others in their gender group and change their behaviors to become more similar to perceived peer group norms. Becoming more similar to peer group norms may foster a positive self- concept (Gibbons, Gerrard, & Lane, 2003) and increase a sense of belongingness (Walton, Cohen, Cwir, & Spencer, 2012). Overall, emerging mastery of social comparison (Ruble & Dweck, 1995) combined with an increasingly rich peer environment, may lead to changes in adolescent’s social and personal identities, including GI.

Another source of peer influence on GI involves peer rewards and punishments for gender-related behaviors. A number of stud- ies have illustrated that negative sanctions (or less acceptance) for certain behaviors may be directed toward adolescents who are gender atypical or who behave in atypical ways (Horn, 2008; Jewell & Brown, 2014; Leaper & Brown, 2008; Lee & Troop- Gordon, 2011). It is likely that adolescents are sensitive to mes- sages from peers about gender typicality expressed through gender teasing or bullying (Drury, Bukowski, Velásquez, & Stella-Lopez, 2013). Peers may also play a role in contributing to youth’s sense of felt pressure for gender conformity through peers’ acceptance or nonacceptance of youth’s behaviors, appearance, or interests. In- deed, adolescents, who felt strong pressure to conform to gender norms, became less accepted by their peers (Yunger et al., 2004). Presumably, peers may also influence each other to develop gender group-related biases because these comments are likely expressed overtly and can be easily overheard or observed. Although the evidence is limited, these studies suggest that peers express the norms of their group and are involved in the reinforcement of similar views within their group.

Contributions From SNA to the Study of Peer Influence on GI Development

Investigating peer influence on GI dimensions without control- ling for how adolescents come to have a particular network of friends (i.e., network selection) risks overestimating the impor- tance of peer influence in GI development. Fortunately, stochastic actor-based modeling (SABM) methods have been developed to estimate peer socialization effects while controlling for alternative processes implicated in peer network selection (Snijders et al., 2010). Consider the following processes linking gender and GI self-concepts to network selection. First, developmental and social networks literatures show that homophily, or preference for similar others, contributes to friendship formation such that friend choices are driven by preferences for friends who are of the same gender,

ethnic, socioeconomic background, and other characteristics (Aboud & Mendelson, 1996; Mehta & Strough, 2009; McPherson, Smith- Lovin, & Cook, 2001). Second, given that preference for similarity on gender-typed behaviors has been shown in children’s peer networks (Martin et al., 2013), preferences to affiliate with friends who have similar levels of GI self-concepts may also contribute to friendship network selection. The final contribution to network selection comes from structural processes describing how connec- tions between individuals depend on the nature of their ties with other members of a group (e.g., triad closure, or a tendency to form ties with friends of friends; Snijders et al., 2010). Triad closure, as well as other structural processes (i.e., reciprocity, popularity), will further amplify network selection on individual attributes, such as gender and GI (Wimmer & Lewis, 2010). The consequence of these multiple, intertwined processes generating networks is that if we attempt to study peer influence on GI within a network without statistically controlling for network selection processes, we are likely to obtain inflated estimates of peer socialization effects. For these reasons, SABM methods are gaining popularity in develop- mental science research focusing on peer socialization of various outcomes (e.g., Veenstra et al., 2013).

The Present Study

Decades of research underscore the role of peers in gender socialization, but no studies to-date have examined whether youth become more similar to their friends on dimensions of GI. Because GI is an evolving aspect of the self, its development needs to be considered over time; thus, our use of SABM permits estimating changes in GI as a part of the network-behavior co-evolution, which is an important methodological advantage. For each dimen- sion of GI, SABM estimates peer influence on GI while controlling for confounding processes including the effects of youth selecting peers who are similar to themselves on gender and GI dimensions, as well as network structural processes. We examined peer influ- ence on GI by using panel data on GI and friendship networks collected from seventh- and eighth-grade students from an ethni- cally diverse public middle school. We hypothesized that adoles- cents would change their GI self-concepts to become similar to their friends, and that these effects would occur even when net- work selection effects were controlled. We also explored the possibility of peer influence being more pronounced on certain dimensions of GI rather than others, informed by the distinction between between-gender and within-gender dimensions of GI (Pauletti et al., 2014). We hypothesized stronger peer influence effects on between-gender dimensions of GI (intergroup bias and felt pressure for gender conformity) than on within-gender dimen- sions of GI (typicality and contentedness). This pattern was ex- pected due to the differences in salience of peer feedback and social norms for the two dimensions. Specifically, we expected stronger socialization effects for between-gender dimensions be- cause this information may be made very salient: youth may engage in discussions about the other gender collective, similar to how they discuss stigmatized racial groups (Kiesner, Maass, Cadinu, & Vallese, 2003), which would contribute to reinforce- ment and magnification of gender differences and peer socializa- tion of between-gender identity self-concepts. Increased salience of social categories, such as gender could occur in these peer discussions, and it is one factor that has been proposed as contrib-

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1580 KORNIENKO, SANTOS, MARTIN, AND GRANGER

uting to social stereotyping and prejudice in developmental inter- group theory (Bigler & Liben, 2007). We anticipated weaker socialization effects for within-gender dimensions because these aspects of self-concept appear to be driven largely by private cognitive processes comparing how adolescents view themselves in relation to their own gender, although there may be some contribution by external social forces. Given the largely private nature of gender typicality and contentedness, we expected these aspects of GI to be less susceptible to peer influence than the other two aspects of GI.

A related goal of this study is to explore how peers select their friends (i.e., network selection). Consistent with large literature documenting gender segregation in peer groups (e.g., Maccoby, 1998; Mehta & Strough, 2009), we expected that adolescents would have a tendency to select peers based on being of the same gender. Because past research has documented that children prefer to affiliate with others who have similar levels of gender-typed activities (Martin et al., 2013), it is possible that youth may prefer to form friendships with those who have similar levels of GI dimen- sions (both between- and within-gender self-concepts). Given that past research has documented that intergroup bias and felt pressure for gender conformity can translate into externalizing problems and peer difficulties (Carver et al., 2003; Smith & Leaper, 2006; Yunger et al., 2004), we anticipated that these between-gender self-concepts would be deleterious for friendship selection dynam- ics and thus negatively associated with the number of friendship nominations sent out to and received from grade-mates. On the other hand, because the same research has shown that within- gender dimensions have positive effects on peer acceptance and social skills, we expected that gender typicality and contentedness would promote friendship network selection processes and be associated with higher number of friendship nominations sent and received. Past research has documented that relative to girls, boys endorse higher levels of some of the GI dimensions (e.g., felt pressure; Ruble et al., 2006); thus, as a final exploratory goal, we examine whether gender differences exist in network selection and in the magnitude of peer network influence.

Method

Participants

Participants of the present study were part of a large study examining longitudinal associations between identity develop- ment, psychological adjustment, friendship networks and educa- tional outcomes (Kornienko & Santos, 2014; Santos, Kornienko, & Rivas-Drake, in press; Santos & Updegraff, 2014). This study was conducted at a large Title I public middle school in a southwestern U.S. metropolitan city, where 89.7% students were racially and ethnically diverse, including 62.5% Latina/o students, and 85.5% of students were eligible for a free and reduced-price lunch (see Table 1 for a more detailed description of sociodemographic characteristics of this sample). All students (N � 1052) attending sixth, seventh, and eighth grades were invited to participate in the study. Teachers introduced the study to students and parents via an informational letter in either English or Spanish (to accommodate Spanish-speaking parents). In this letter, parents were given the option to opt out of their child’s participation in the study as the study used a passive consent procedure since the school served as

loca parentis of the study. All study recruitment and measurement procedures were approved by the school district and the universi- ty’s institutional review board. Rates of participation of the student population in our study were high in any given wave by grade, ranging from 89% to 96% across waves.

The present study includes all students who participated in the study from seventh and eighth grades between Fall 2011 and Spring 2012 (i.e., Wave 2 and 3 of the larger study). Wave 1 data of the larger study were rendered unusable for two reasons. First, the questionnaire at Wave 1 (Spring 2011) did not include the final version of the felt pressure subscale. Second, due to another school’s closure in the district, 142 new students were added to seventh grade (48% of total seventh-grade population at Wave 2) and 95 new students from another school were added to Grade 8 (30% of total eighth-grade population at Wave 2) during the fall of 2011. Given this addition of large number of new students due to the closing of a nearby school, the proportion of stable ties based on Jaccard index between Spring 2011 (Wave 1) and Fall 11 (Wave 2) fell below 13.5%, which violates SABM assumption that network ties represent stable states (as opposed to brief events, Snijders et al., 2010). Thus, we focused on Waves 2 and 3 of the larger study, and, unfortunately, also had to exclude 6 graders who participated in Waves 2 and 3 from the current analysis because only 16% of the friendships present at both time points were present at either one of the time points (based on Jaccard index, see Snijders et al., 2010). This indicates there was not enough stability in 6th graders’ friendships to support the SABM assumption that friendships represent stable states. Because 19% of seventh grad- ers’ friendships and 22% of eighth graders’ friendships remained stable from Wave 2 to 3 (based on Jaccard index), the analytical sample was restricted to students from seventh and eighth grades from Waves 2 and 3 of the larger study. For simplicity, they are herein referred to as Wave 1 and 2 of this longitudinal network- behavior study.

Table 1 Descriptive Statistics on Sociodemographic Characteristics of the Sample

Characteristic Grade 7 Grade 8

Boy 46% 53% Age, M (SD) 12.12 (0.53) 13.15 (0.52) Ethnic/racial background

European American 12% 11% African American 16% 19% Native American 8% 7% Latina/o 57% 56% Other race/ethnicity 7% 8%

Free/reduced lunch 79.9% 79.2% Generational status of parents

Both U.S. born 47.2% 43.5% One parent foreign born 11.7% 12.8% Both parents foreign born 40.5% 42.2%

Languages spoken at home English 37.8% 37.6% Spanish 3.0% 1.6% English and Spanish 54.8% 53.7% Other language 4.3% 6.7%

Family structure Two-parent family 63.6% 57.2% Single-parent family 29.8% 32.6%

Note. Each categorical variable was dummy-coded (1 � yes, 0 � no).

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1581PEER INFLUENCE ON GENDER IDENTITY

Our analytic sample (N � 670) consisted of 330 students from seventh grade and 340 students from eighth grade for whom peer nomination and survey data were obtained. In Wave 1, 95% of seventh graders and 96% of eighth graders completed our survey. In Wave 2, 92% of seventh graders and 94% of eighth graders completed our survey. Considering the percentage of students who participated in the study at both time-points, 84.9% of students from seventh grade and 87.2% of students from eighth grade completed assessments at both time points. Boys comprised 46% of seventh and 53% of eighth grades. The mean age for seventh graders was 12.12 (SD � .53) and for eighth graders, it was 13.15 (SD � .52). For each grade, we provided descriptive information (see Table 1) on the percentage of students receiving free/reduced lunch (school-district reported proxy for socioeconomical status), generational status of parents, languages spoken at home, and family structure.

Procedure

All participating students completed a survey packet in their classrooms. Nonparticipating students were asked to read a book or complete homework assignment during survey administration. Student assent was obtained on the day of data collection. The survey was completed in English as all students were fluent in English. Research staff read the survey aloud to the students to facilitate comprehension, and individual assistance was provided to students if needed. The survey took approximately two class periods (�90 min) total to complete, and students responded questions about their identity, psychological and educational out- comes on Day 1, and completed friendship nominations on Day 2. To thank participants for their time, all students received a small gift (i.e., a pen and a water bottle with the school’s logo).

Measures

GI. We developed a modification of Egan and Perry’s (2001) measure of GI to assess the five dimensions of GI: self- identification as a member in a gender category (assessed as a categorical question in the demographic section of our survey), and four continuous measures of GI: gender typicality (the degree to which one feels as a typical member of one’s gender group), felt pressure from peers for gender conformity (the degree to which one feels pressure from peers to conformity to gender stereotypes), and intergroup bias (the degree to which one holds negative views toward the other gender). In a study with a similar sample (i.e., multiethnic, 9 to 14-year-olds) conducted in the same region as the present study, the Harter response format (originally used by Egan and Perry) was shown to be confusing to children (Michaels, Barr, Roosa, & Knight, 2007), therefore, we used a Likert format to assess the four continuous measures of GI with response options to the items ranging from 1 (strongly disagree) to 5 (strongly agree). In order to use a Likert scale, we also had to make minor modi- fications in the wording of the questions. For gender typicality, questions included, for example (for girls), “I feel that I am just like all other girls” (see the appendix in the Online Supplemental materials for all items in these scales). The gender typicality adapted scale for the present study consisted of five items (original measure consists of six), and the adapted version of this scale was shown to be internally consistent and reliable (Cronbach’s alpha’s

were .86 across the grades at Wave 1 and .89 at Wave 2). A higher score indicates feelings that one is a typical member of their gender category. We also adapted Egan and Perry’s gender peer felt pressure scale because the original scale includes 10 items assessing felt pressure from peers as well as family. Because these may represent distinct sources of pressure, and because we are particularly interested in felt pressure from peers, we kept only four adapted items that assessed felt pressure from peers (e.g., “Other kids, who are girls like me, would get upset if someone who is a girl acts like a boy”). The gender peer felt pressure scale was shown to be internally consistent and reliable (Cronbach’s alpha’s were .75 across the grades at Wave 1 and .92 at Wave 2). The gender intergroup bias scale adapted for the present study consisted of three items (original measure consists of eight), and the adapted version of this scale was shown to be internally consistent and reliable (Cronbach’s alpha’s were .85 across the grades at Wave 1 and .85 at Wave 2). A higher score indicates negative feelings toward the other gender group. In order to measure gender contentedness (the degree to which one feels content with one’s gender group), we adapted Sellers, Smith, Shelton, Rowley, and Chavous (1998) measure of private regard (3 items). We did so because the items in the Sellers et al. (1998) scale are already in Likert-format and due to similarities to the Egan and Perry measure of contentedness (for a boy, e.g., “I feel good about being a boy”). In keeping with our use of Egan and Perry’s framing, we refer hereafter to this adapted scale as gender contentedness. The gender contentedness scale was shown to be internally consistent and reliable (Cronbach’s alpha’s were .64 across the grades at Wave 1 and .76 at Wave 2). A higher score indicates feelings that one is content with one’s gender category. Because SABM requires discrete ordinal behavioral outcome vari- ables, we recoded each of the continuous GI variables into ordinal variables. Each was recoded to four levels, using increments of 1 SD of the continuous z-score (z � �1, �1 � z � 0, 0 � z � 1, z � 1; e.g., Delay, Ha, Van Ryzin, Winter, & Dishion, 2016).

Friendship networks. Participants received a roster listing all students from their grade and were asked to nominate up to 10 friends of either gender. These friendship nomination data were used to construct friendship network matrices of seventh and eighth grade students for Wave 1 and Wave 2 such that these matrices contained unilateral (i.e., A nominated B) friendship nominations that were coded in a binary fashion such that 1 denoted that a friendship tie existed between A and B, and 0 indicated no tie existed between A and B.

Demographic variables. Adolescents reported their gender (male � 1, female � 0) and race/ethnicity (dummy-coded: 1 � yes, 0 � no, for the following categories: English European American, African American, Native American, Latina/o, Other Race/Ethnicity), language spoken at home (dummy-coded: 1 � yes, 0 � no, for the following categories: English, Spanish, both English and Spanish, and other language), family structure (two parent � 1 vs. single parent � 0), and generational status (dummy-coded: 1 � yes, 0 � no, for the following categories: both U.S. born, one parent foreign born, both parents foreign born). The district provided data on each student’s free and reduced lunch status (a proxy for socioeconomic status, coded 1 � yes, 0 � no).

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1582 KORNIENKO, SANTOS, MARTIN, AND GRANGER

Overview of Stochastic Actor-Based Modeling

Employing SABM approach provides key advantages for the study of peer influence on GI dimensions because it allows for estimation of the degree of peer influence on changes in GI dimensions while controlling for a host of potentially confounding processes, including (a) initial selection into friendships based on gender and similar levels of GI aspects and (b) network structural processes (e.g., reciprocity and popularity).

Model overview. The SABM consists of two submodels that are simultaneously estimated (Snijders et al., 2010). The network submodel tests the likelihood of friendship ties between adoles- cents based on various network selection processes. The behavior submodel captures effects related to changes in GI over time. The model estimates changes between the observed networks using a continuous-time Markov process that allows for a sequence of a large number of unobserved microsteps to be taken between ob- servation points (one network tie or a behavior can be changed in one microstep). An evaluation function describes the “rules” that guide actors’ decisions, which are the model parameters for the hypothesized selection and influence effects. A rate function de- termines how many opportunities for change occurs between waves. Model estimation uses a method of moments procedure to estimate parameters. This procedure calculates summary statistics based on the effects included in the model. These statistics are counts that represent various network structures, such as the num- ber of gender homophilous dyads, observed at Time 2 (for details, see Snijders, Steglich, & Schweinberger, 2007). The goal during estimation is to identify parameter values that allow the model to produce networks whose summary statistics match those observed in the data (i.e., at Time 2). The estimation algorithm reaches convergence when t statistics representing deviations between the observed and model-implied networks are less than 0.1 for each model parameter and less than 0.25 across all of the model pa- rameters. Model parameters are tested for significance based on a t-ratio (estimate divided by the standard error).

Model effects. With respect to the SABM specification, for the network submodel, we considered three types of effects on network selection for each of the four dimensions of GI. The GI ego effect estimates the effect of GI (e.g., typicality, felt pressure) on an adolescent’s tendency to nominate others as friends. A positive effect would indicate that adolescents with greater levels of GI nominated more friends over time. The GI alter effect describes how GI affects adolescents’ likelihood of receiving nom- inations from peers. A positive effect would indicate that adoles- cents with higher levels of GI (e.g., typicality, felt pressure) were more likely to be nominated as friends by their peers. In our preliminary models, we also considered alter squared effect that accounts for nonlinearity in the associations between GI dimen- sions and incoming friendship ties; however, these effects were not significant and thus were omitted from the final models. The GI similarity effect estimates the tendency of adolescents to nominate friends who have similar levels of respective GI self-construct (measured by their absolute difference). A positive effect of GI similarity would mean that adolescents were more likely to form friendships among peers with similar levels of GI. Additionally, we estimated the effect of similarity on gender and ethnic/racial background on the likelihood of network selection. Finally, we included parameters for several network structural processes. Rec-

iprocity captured whether adolescents were more likely to nomi- nate peers who had nominated them. We used geometrically weighted edgewise shared partners (GWESP) effects and an inter- action of GWESP and reciprocity to model triadic closure pro- cesses by assessing whether having multiple friends in common increased the likelihood of friendship formation. The indegree popularity effect estimated whether students who previously re- ceived more nominations were more likely to receive additional nominations over time. The indegree activity effect estimated whether students who received more nominations were more likely to send out a greater number of nominations. Finally, the outde- gree activity effect estimated whether students who previously sent out a higher number of ties were more likely to subsequently send many ties. We used a square-root transformation of these activity and popularity effects to give greater weight to differences in popularity and activity at low versus high levels. The network function also included effects for outdegree, which controlled for the number of ties. Finally, network rate effect represented net- work change opportunities.

Turning to the behavior submodel, we tested two effects that represent feedback on the four dimensions of GI. The linear shape effect expresses the basic tendency toward higher or lower values of GI, whereas the quadratic shape effect allows for the self- reinforcement of GI that can result in a bimodal distribution of GI. We then estimated the peer influence effect on each of the four GI dimensions using total similarity effect. This effect predicts changes in GI based upon how similar an adolescent’s GI is to the total levels of GI across all of his or her friends. A positive effect indicates that changes in GI bring an adolescent closer to his or her friends’ level of GI. This effect is weighted by the total number of friends, and thus it considers the overall level of a particular aspect of GI in a friendship network. Lastly, for each of GI dimensions, the models included distinct rates representing behavior change opportunities.

Results

Descriptive Analyses

For each grade, we present descriptive statistics for the sample in Table 1. Zero-order correlations among GI dimensions are presented for each grade by gender in Table 2. Results indicated relative stability in GI dimensions from Time 1 to 2: significant and positive correlations were found between Waves 1 and 2 for gender typicality, intergroup bias, felt pressure, and contentedness, for both boys and girls from seventh and eighth grades. Correla- tions across GI dimensions are presented for each grade and by gender in Table 2. Results reveal relatively low but significant associations and somewhat similar patterns across the two grades.

To examine gender differences in GI dimensions, we used independent samples t tests and reported Cohen’s d values, which assessed the magnitude of mean-level differences in GI dimen- sions that is attributable to each gender (Cohen, 1988; Table 3). For seventh graders, results showed gender differences at Waves 1 and 2 in typicality, intergroup bias, and felt pressure. Specifically, at both waves, boys reported feeling more gender typical and reported more felt pressure for gender conformity from peers than girls. However, girls reported higher intergroup bias than boys at both waves. Interestingly, there were no significant gender differ-

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1583PEER INFLUENCE ON GENDER IDENTITY

ences in contentedness at either waves. Results were similar for eighth grade, with the exception of contentedness at Wave 1, in which girls reported feeling more gender contentedness than boys.

Considering network characteristics for the two grade-level net- works, our descriptive analyses revealed that at Wave 1, on aver- age, seventh graders nominated 6.33 friends and eighth graders reported 6.83 friendship ties (at Wave 2, the respective average outdegrees were 6.02 and 6.98). We observed that the grade-level networks had a fairly low density (i.e., proportion of existing friendship ties relative to the total possible ties) of .02, sug- gesting that most of the total possible ties in the network do not exist. In Wave 1, there were 2032 ties among seventh-grade and

2246 among eighth-grade students; at Wave 2, there were 1881 and 2306 friendships in these grade-level networks. To assess the degree of network change, we calculated the Jaccard index, which denotes the proportion of ties that are present at both waves. These indices were .19 for seventh and .22 for eighth grades suggesting a sufficient balance of stability and change in friendships over time, making the SABM approach suitable (Snijders et al., 2010).

Overview of SABM Analyses

We estimated two types of models for each of the two grades. The first model addressed our main goal of examining peer net-

Table 2 Correlations Among GI Dimensions Per Grade

Dimension 1 2 3 4 5 6 7 8

7 grade 1. GI typicality Wave 1 — .475�� .259�� .125 �.103 .080 .150 �.017 2. GI typicality Wave 2 .578�� — .134 .395�� �.023 �.057 .069 .001 3. GI contentedness Wave 1 .438�� .329�� — .317�� .100 .023 .106 �.013 4. GI contentedness Wave 2 .450�� .438�� .549�� — .064 .140 �.012 .120 5. GI intergroup bias Wave 1 .267�� .334�� .258�� .214�� — .250�� �.130 .071 6. GI intergroup bias Wave 2 .241�� .233�� .306�� .470�� .395�� — .056 .120 7. GI felt pressure Wave 1 �.068 �.11 �.183� �.226� .163� �.003 — .429��

8. GI felt pressure Wave 2 �.108 �.25�� .141 �.213� �.071 �.119 .412�� — 8 grade

1. GI typicality Wave 1 — .522�� .434�� .210� .042 �.044 .093 .020 2. GI typicality Wave 2 .685�� — .274�� .388�� .086 .086 .161 .213��

3. GI contentedness Wave 1 .402�� .495�� — .398�� .035 �.091 .226�� .035 4. GI contentedness Wave 2 .306�� .570�� .538�� — �.054 .055 .125 .254��

5. GI intergroup bias Wave 1 .075 .188� .180� .141 — .424�� �.025 �.156 6. GI intergroup bias Wave 2 .294�� .382�� .246�� .230� .316�� — �.141 �.025 7. GI felt pressure Wave 1 .018 �.057 �.011 �.028 �.033 �.022 — .274��

8. GI felt pressure Wave 2 .001 .091 .028 .045 .111 .165 .465�� —

Note. GI � gender identity. Boys are above the diagonal and girls below the diagonal. Ns range from 240 to 333. � p � .05. �� p � .01. ��� p � .001.

Table 3 Gender Differences in GI Dimension Per Grade

Dimension

7 Grade 8 Grade

M SD t df Cohen’s d M SD t df Cohen’s d

Girl GI typicality Wave 1 3.51 0.93 �4.46�� 295 �0.52 3.43 0.88 �3.80�� 311 �0.43 Boy GI typicality Wave 1 3.95 0.73 3.81 0.87 Girl GI typicality Wave 2 3.38 0.90 �4.60�� 271 �0.56 3.32 0.84 �4.81�� 288 �0.57 Boy GI typicality Wave 2 3.86 0.813 3.80 0.85 Girl GI intergroup bias Wave 1 3.52 1.03 8.20�� 295 0.95 3.41 0.91 7.54�� 310 0.86 Boy GI intergroup bias Wave 1 2.61 0.86 2.65 0.86 Girl GI intergroup bias Wave 2 3.50 0.96 6.85�� 269 0.83 3.26 0.93 6.21�� 287 0.73 Boy GI intergroup bias Wave 2 2.75 0.82 2.60 0.88 Girl GI felt pressure Wave 1 2.49 1.06 �11.05�� 294 �1.29 2.41 0.98 �13.68�� 309 �1.56 Boy GI felt pressure Wave 1 3.84 1.03 4.02 1.07 Girl GI felt pressure Wave 2 2.40 1.05 �8.77�� 268 �1.07 2.420 .95 �11.70�� 287 �1.38 Boy GI felt pressure Wave 2 3.65 1.27 3.85 1.11 Girl GI contentedness Wave 1 4.37 0.67 �0.02 295 0 4.33 0.62 1.98� 311 0.22 Boy GI contentedness Wave 1 4.37 0.56 4.20 0.58 Girl GI contentedness Wave 2 4.25 0.72 �0.72 271 �0.09 4.13 0.79 �1.55 287 �0.18 Boy GI contentedness Wave 2 4.32 0.67 4.26 0.65

Note. GI � gender identity. The effect size corresponding to Cohen’s d is determined using the following ranges (Cohen, 1988): small (.2–.3), medium (.31–.50), and large (�.8) effect sizes. � p � .05. �� p � .001.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1584 KORNIENKO, SANTOS, MARTIN, AND GRANGER

work influence on GI dimensions while controlling for contribu- tions of gender and GI to network selection as well as network structural effects. A second model addressed our exploratory goal of examining whether gender moderated peer network influence and network selection on the four dimensions of GI. We conducted these analyses using RSiena 4.0 (version 1.1–290; Ripley, Sni- jders, Boda, Voros, & Preciado, 2016) in R (version 3.3, R-Project; http://www.r-project.org). We followed the recommended forward- fitting model specification approach to avoid multicollinearity among network effects (Snijders et al., 2010). This iterative ap- proach uses a score-type test to compare the fit of statistics representing potential effects with the effect included versus ex- cluded from the model (Schweinberger, 2012). Effects that are significant, indicating that they are likely to improve the fit of the model, are sequentially included. Our preliminary score-type test analyses indicated that there were no significant gender differences in contributions of GI components to network selection, but there were potential gender differences in magnitude of peer influence and GI dimensions levels, which were examined in the second model. Additionally, because students were given an option to nominate up to 10 friends, we initially specified maximum outde- gree to be 10. When evaluating the goodness of fit, which com- pares characteristics of model-implied simulated networks to the characteristics of the observed network at Time 2, we discovered that our model implied simulated data had fewer networks with outdegrees of 9 and 10 compared with the observed data. This lower simulated count of outdegrees occurred because for those individuals with 10 outgoing ties, the model proceeded by drop- ping ties and adding new ties, not necessarily exchanging ties. Thus, in the final models we specified maximum outdegree to be 11 to approximate the number of higher outdegrees in simulated networks and improve the goodness of fit. We present the good- ness of fit for the first model for seventh and eighth grades (see Figures 1A and 1B in the online supplementary materials). Finally, in the presented models, convergence was achieved because t statistics representing deviations between the observed and model- implied networks were less than 0.1 for individual model param- eters and less than 0.25 across all of the model parameters.

Peer influence on GI dimensions. We begin by considering the results of Model 1, which addressed our goal of examining peer influence on GI while controlling for network selection dynamics (GI Dynamics, Model 1, Table 4). Whereas we expected to find network influence on GI, we considered the possibility that the four dimensions of GI may show differing levels of influence, with influence being greater for the between-gender aspects of GI. As anticipated, we found evidence for significant and positive peer influence on felt pressure for gender conformity for eighth graders (est. � .31, p � .05) suggesting that, over time, these students’ levels of gender felt peer pressure became similar to those of their friends compared with those with whom they were not friends. To assess the magnitude of this effect, we calculated odds ratio for change in these dimensions of GI (see Ripley et al., 2016 for details). Having one additional friend who felt higher level of pressure for gender conformity than the focal individual increased the odds for the focal youth to also increase their levels of felt pressure by a factor of exp(.31/3) � 1.11, compared with no change in this GI dimension. This means that eighth-grade students had an 11% increase in the odds of GI change when at least on their friends had a higher level of felt pressure for gender confor-

mity. We also observed significant and positive peer influence on intergroup bias among both seventh and eighth graders (est. � .80, p � .001; est. � .52, p � .001). Considering the magnitude of these effects, the results show that for seventh graders, having one additional friend who felt higher levels of intergroup bias than the focal individual increased the odds for the focal youth to also increase their levels of intergroup bias compared with no change by a factor of exp(.80/3) � 1.31, or a 31% increase in the odds of GI change. Among eighth graders, the same scenario of peer influence would result in an increase of intergroup bias change by a factor of exp(.52/3) � 1.19, or an 19% increase in the odds of GI change. Finally, we did not document significant peer influence effects for gender typicality and gender contentedness (i.e., the within-person dimensions of GI) in the networks of seventh or eighth graders.

Network selection on gender and GI dimensions. To deter- mine whether influence occurred and to obtain unbiased estimates of influence, we examined these processes in a model that also controlled for network selection on GI dimensions of interest (i.e., how GI affects initial selection of friends; Network Selection Dynamics, Model 1, Table 4). Thus, we assessed how the four GI dimensions were associated with network selection. Our results showed that eighth graders befriended others with similar levels of gender contentedness (est. � .42, p � .01) and gender typicality (est. � .31, p � .05). We found that both seventh and eighth graders who reported higher levels of gender contentedness were more like to send out a higher number of friendship ties (est. � .12, p � .05; est. � .12, p � .01). Our results showed that seventh graders with higher levels of intergroup bias were less likely to be nominated as friends by their peers (est. � �.08, p � .05). We found no evidence that adolescents selected friends based on similarity on levels of intergroup bias or on felt pressure for gender conformity.

Network selection on control variables and network struc- tural processes. Our modeling approach obtained estimates of network influence and selection on GI dimensions, while control- ling for several confounding network selection and network struc- tural processes. When studying associations between GI self- concepts and networks, the key process to control for is gender segregation in friendship networks. Our models included such controls and revealed a significant homophily on gender in both grades (est. � .43, p � .001, OR � 1.54 for seventh graders; est. � .25, p � .001, OR � 1.28 for eighth graders) suggesting that friendship ties were 54% more likely to be formed among seventh graders of the same gender and 28% more likely among eighth graders of the same gender. We also controlled for contribution of similarity on ethnic/racial background as affecting the likelihood of friendship ties. In both grades, we found significant and positive Latina/o similarity effect suggesting that friendships were more likely to form among Latina/o youth (est. � .16, p � .001; est. � .10, p � .001). Among eighth graders only, we also documented that students were more likely to select each other as friends if both of them were African American (est. � .19, p � .001). In both grades, being in the same Social Studies classroom increased the odds of friendship ties (est. � .44, p � .001; est. � .51, p � .001).

In addition to estimating these confounding network selection processes, we also included network structural effects (e.g., reci- procity, transitivity) in our models. We found that adolescents’ networks were formed through several common network structural

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1585PEER INFLUENCE ON GENDER IDENTITY

Table 4 SABM Results for Peer Influence on GI and Network Selection on GI

Effects

7th grade 8th grade

Model 1 Model 2 Model 1 Model 2

Est. St. Er p Est. St. Er p Est. St. Er p Est. St. Er p

GI dynamics

GI typicality linear shape �.02 .08 �.03 .10 .10 .09 .12 .10 GI typicality quadratic �.12 .19 �.30 .24 .01 .18 �.21 .24 Peer influence on GI typicality .10 .22 �.08 .26 .32 .22 .13 .25 GI typicality from Boy .56 .26 � .62 .21 ���

Peer influence on GI Typicality � Boy �.29 .21 .02 .21 GI contentedness linear shape .31 .09 ��� .30 .09 ��� .11 .07 .11 .07 GI contentedness quadratic shape �.18 .21 �.17 .18 .09 .14 .09 .15 Peer influence on GI contentedness �.23 .25 �.23 .23 .13 .17 .13 .18 GI contentedness from Boy .11 .16 .34 .15 �

Peer influence on GI Contentedness � Boy .03 .18 .06 .17 Intergroup bias linear shape .10 .10 .05 .11 .28 .09 ��� .27 .10 ���

Intergroup bias quadratic shape .26 .18 .01 .23 .13 .15 �.04 .18 Peer influence on intergroup bias .80 .29 ��� .61 .32 .52 .19 ��� .36 .21 Intergroup bias from Boy �.45 .29 �.33 .20 Peer influence on intergroup bias � Boy .59 .35 �.12 .21 Felt pressure linear shape .10 .07 .09 .07 .05 .08 .05 .08 Felt pressure quadratic shape .15 .11 �.05 .20 .15 .11 �.15 .19 Peer influence on felt pressure .24 .14 .02 .22 .31 .15 � .05 .20 Felt pressure from Boy .46 .31 .68 .30 �

Peer influence on felt pressure � Boy �.11 .18 �.23 .20

Network dynamics

GI effects on network selection GI typicality alter �.02 .03 �.02 .04 �.03 .03 �.03 .03 GI typicality ego �.01 .05 �.01 .04 �.10 .05 � �.11 .06 GI typicality similarity .34 .17 � .34 .19 .31 .15 � .31 .16 �

GI contentedness alter .06 .05 .06 .05 �.02 .04 �.02 .04 GI contentedness ego .12 .06 � .12 .06 � .12 .05 �� .11 .06 GI contentedness similarity �.08 .26 �.08 .24 .42 .17 �� .42 .16 ���

Intergroup bias alter �.08 .04 � �.08 .05 .04 .04 .05 .04 Intergroup bias ego �.02 .06 �.02 .07 .05 .06 .05 .07 Intergroup bias similarity �.39 .20 �.40 .26 �.07 .19 �.07 .23 Felt pressure alter .01 .04 .01 .05 .02 .03 .02 .04 Felt pressure ego �.01 .05 �.01 .06 .05 .05 .06 .06 Felt pressure similarity �.39 .18 � �.42 .21 � .03 .14 .03 .15

Individual attributes effects on network selection Boy alter �.11 .08 �.11 .09 �.02 .07 �.02 .07 Boy ego �.01 .09 �.01 .11 �.04 .08 �.06 .10 Boy similarity .43 .05 ��� .44 .06 ��� .25 .04 ��� .24 .05 ���

European American similarity .07 .07 .06 .08 .12 .07 .12 .07 African American similarity .08 .06 .08 .06 .19 .05 ��� .19 .05 ���

Native American similarity .06 .08 .07 .08 .03 .08 .04 .08 Latina/o similarity .16 .05 ��� .16 .05 ��� .10 .04 �� .10 .04 �

Other ethnicity similarity .09 .09 .09 .10 �.11 .07 �.11 .08 Same wave started .06 .04 .06 .04 .00 .04 .00 .04 Same classroom .44 .06 ��� .44 .06 ��� .51 .06 ��� .51 .06 ���

Network structural processes Network rate 22.93 1.11 ��� 22.95 1.41 ��� 22.99 .95 ��� 22.98 1.22 ���

Outdegree (density) �1.43 .31 ��� �1.40 .38 ��� �1.87 .33 ��� �1.87 .35 ���

Reciprocity 1.83 .14 ��� 1.82 .15 ��� 2.01 .11 ��� 2.01 .11 ���

Number of actors at distance 2 �.16 .04 ��� �.16 .05 ��� �.11 .04 ��� �.11 .03 ���

GWESP i ¡ k ¡ j .62 .20 ��� .62 .23 ��� .44 .18 �� .44 .17 ���

GWESP i ¢ k ¢ j �.04 .12 �.04 .12 �.19 .09 � �.19 .11 GWESP i ¡ k ¢ j .50 .23 � .50 .21 �� .84 .12 ��� .83 .20 ���

GWESP i ↔ k ↔ j �.60 .15 �� �.60 .17 �� �.62 .19 ��� �.62 .12 ���

Indegree—popularity (sqrt) .09 .04 � .09 .05 .07 .03 � .07 .03 �

Indegree—activity (sqrt) �.09 .11 �.10 .12 �.30 .10 ��� �.29 .12 ��

Outdegree—activity (sqrt) �.09 .07 �.10 .09 .14 .08 .14 .08

Note. Peer influence was parameterized as total similarity. Categorical variables (boy, ethnic/racial categories, same school wave started, same social studies classroom) were dummy-coded (1 � yes, 0 � no). SABM � stochastic actor-based modeling; GI � gender identity; GWESP � geometrically weighted edgewise shared partners; sqrt � square root. � p � .05. �� p � .01. ��� p � .001 (all two-tailed).

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1586 KORNIENKO, SANTOS, MARTIN, AND GRANGER

processes. Specifically, a high degree of reciprocation of friend- ship nominations was observed (est. � 2.51, p � .001; est. � 3.05, p � .001). To model transitivity, we examined several GWESP effects that represented relevant triadic configurations. First, a significant and positive GWESP i ¡ k ¡ j effect indicated that individual i was increasingly likely to form a connection to j when they had at least one mutual friend k such that i nominated k and k nominated j, whereas having additional mutual friends beyond k contributed to a smaller extent to this likelihood (est. � .62, p � .001; est. � .44, p � .001). Next, a significant and positive GWESP i ¡ k ¢ j effect indicated that individual i was increas- ingly likely to form a connection to j when they had at least one mutual friend k such that both i and j nominated k, whereas having additional mutual friends beyond k contributed to a smaller extent to this likelihood (est. � .50, p � .05; est. � .84, p � .001). Finally, a significant and negative GWESP i ↔ k ↔ j effect showed that i was less likely to form a connection to j when they had at least one mutual friend k to whom they were connected through reciprocated ties, whereas having additional mutual recip- rocally connected friends beyond k contributed to a smaller extent to this likelihood (est. � �.60, p � .001; est. � �.62, p � .001). The negative parameter for the number of actors at distance of 2 is another indicator of transitivity in the network, suggesting that individuals are not friends when they do not have several friends in common (est. � - 0.16, p � .001; est. � - 0.11, p � .001). In both grades, the positive indegree popularity effect (est. � .09, p � .05, est. � .07, p � .05) showed that students who received a high number of friendship nominations were more likely receive addi- tional friendship ties (i.e., network popularity reinforces itself). Among eighth-grade students, a significant and negative indegree activity effect indicated that students who received a high number of friendship nominations were less likely to send out friendship ties (est. � �.30, p � .001).

Gender differences in peer influence processes. Our final goal was to explore whether the strength and magnitude of peer influence processes were different for boys and girls (GI dynamics, Model 2, Table 4). Our results revealed no significant gender differences in the magnitude of peer influence on any dimensions of GI. Finally, several gender differences are noteworthy in the mean levels of GI constructs. Specifically, boys from eighth grade reported significantly greater levels of gender typicality, content- edness and felt pressure for gender conformity compared with girls. Boys from seventh grade reported significantly greater levels of gender typicality only.

Assessing goodness of fit. As the final step, we followed established procedures for evaluating goodness of fit for statistical network models (Hunter, Goodreau, & Handcock, 2008; Ripley et al., 2016). Conventional model fit indices (e.g., AIC and BIC) are not appropriate in network modeling because “observations are not independent and identically distributed sample” (Hunter et al., 2008, p. 257). Alternative methods have been developed to eval- uate goodness of fit of network models, which involve a compar- ison of model-implied simulated networks to the observed data in regards to various network properties. We assessed fit for the final Model 1 with respect to the following auxiliary statistics: distri- butions of outgoing friendship ties (outdegree), incoming friend- ship ties (i.e., indegree), geodesic distance distributions, triad census as well as behavior distribution for each of the four GI self-concepts (Lospinoso, Schweinberger, Snijders, & Ripley,

2011). Using the sienaGOF function, goodness of fit is assessed by comparing the observed values at the end of the period (i.e., Time 2) with simulated values from the model (Ripley et al., 2016). These differences are assessed by the Mahalanobis distance, which at p � .05 levels suggests that the predicted auxiliary statistic distribution does not significantly depart from the observed statis- tic, indicating adequate fit of the model to the data. Considering GOF for the seventh-grader network (see Figure 1A in the online supplemental materials), our results show that the distributions of indegrees (p � .34), triadic configurations (p � .70), GI content- edness (p � .41), and gender intergroup bias (p � .26) in the model-implied simulated networks were not different from the distribution of these configurations in the observed network at Time 2. Our GOF results demonstrated that for the eighth graders (see Figure 1B in the online supplemental materials) the distribu- tions of indegrees (p � .50), gender typicality (p � .22), gender intergroup bias (p � .30), and felt pressure for gender conformity (p � .08) in the model-implied simulated networks were not different from the distribution of these configurations in the ob- served network at Time 2. These analyses suggest that, overall, our model did a good job at reproducing these properties of the observed networks. The GOF for outdegree distributions for the seventh graders (p � .004) and eighth graders (p � .007) and geodesic distance distributions for both age groups (p’s � .00) showed a significant difference between the model-implied sim- ulated networks and the observed network, suggesting that the model is not perfectly reproducing the distribution of outdegree and geodesic distances. To convey why the outdegree distribu- tion was just shy of fitting, we present Figure 1 (see the Online Supplemental materials). In these figures, the red dots and values represent the observed network properties, and gray lines represent 90% confidence intervals for the model-implied sim- ulated networks. The placement of the red line within the 90% confidence interval boundaries indicates that our model pro- duced adequate to good goodness-of-fit statistics for the key auxiliary statistics. The lack of fit for outdegree for the seventh and eighth-grade network was because the model was not able to account for the “jump” in the cumulative outdegree distri- bution from 9 and 10 suggesting that there is a higher number of individuals who send out 9 and 10 friendship nominations, compared with what is expected under the model. The lack of fit for geodesic distance appears to be due to the model producing significantly higher levels of geodesic distance of 4 suggesting that model-implied networks are more connected than the ob- served network. The lack of fit in triad census statistic in eighth grade (p � .027) appears to be driven by abundance of a particular triadic configuration in our data (030T, or A ¡ B ¢ C, A ¡ C). Finally, our model shows deviations between model-produced distributions of the following three behavioral outcomes: GI typicality (p � .03) and felt pressure for gender conformity (p � .02) for seventh graders and GI contentedness (p � .01) for eighth graders. In summary, although the statis- tical test of GOF for these parameters suggested significant differences between simulated and observed networks because the red line remained within grey bands, representing 90% CI, we concluded that overall the model produced an adequate fit to the data.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1587PEER INFLUENCE ON GENDER IDENTITY

Discussion

The main goal of this study was to investigate peer influences in between- and within-gender dimensions of GI in young adoles- cents’ peer networks. In doing so, we used a state-of-the-art analytical approach that allowed us to examine whether friends become similar to one another on dimensions of GI over time (peer influence), while statistically controlling for similarities in GI self-concepts that may stem from peers preferring to befriend others with similar levels of GI (peer selection). Consistent with our expectations, we found significant and positive peer socializa- tion effects on between-gender dimensions of GI: intergroup bias among seventh and eighth graders and felt pressure for gender conformity for adolescents in eighth grade. In addition, and as expected, adolescents in both grade cohorts formed friendships with students of the same gender, and among eighth graders, friendship network selection was also based on similarity on within-gender dimensions of GI: gender typicality and contented- ness. Additionally, adolescents who were more content with their gender sent out a higher number of friendship nominations, whereas those seventh graders holding more negative attitudes toward other gender were less attractive as a potential friend. Finally, we did not observe gender differences in the magnitude of peer network influence on GI self-concepts. Despite decades of research highlighting the role of peers in gender socialization, this is the first study, to the best of our knowledge, to document that adolescents become similar to their friends on certain dimensions of GI. These findings make important conceptual contributions to the body of work concerning the role of peers and social context for identity and gender development more broadly (e.g., Egan & Perry, 2001; Yunger et al., 2004). Methodological advances in longitudinal social network analysis (Snijders et al., 2010) enabled the current investigation to provide a nuanced description of the role that the peer context plays in shaping GI.

Our findings regarding peer influence on between-gender di- mensions of GI are that, over the course of the school year, seventh and eighth graders changed their levels of gender intergroup bias (i.e., negative feelings toward the other gender) to become similar to the intergroup bias levels of their friends. That is, youth who reported being low on gender intergroup bias, who were friends with adolescents with high gender intergroup bias, moved toward their friends’ levels of bias, whereas youth who reported initial high levels of gender intergroup bias, whose friends reported low levels of bias, moved their levels of gender bias to better match those of their friends. The observed pattern of peer influence on gender intergroup bias is in line with theoretical proposals that social stereotyping and prejudice develop through establishing psychological salience of the other group, explicit discussions of salient group’s attributes, and development of affective compo- nents of intergroup bias (Bigler & Liben, 2007). Because gender intergroup bias stems from magnification of gender differences and expanded stereotype use (Powlishta, 1995), our findings of significant peer influence effects contribute not only to GI research but also to applications of intergroup theory within the develop- mental literature. Consider, for instance, that for some adolescents, peers appear to make the other gender group more salient and more susceptible to negative comments, but for other adolescents, peers appear to make the other group less salient and less susceptible to negative comments. In keeping with social identity and intergroup

theories (e.g., Bigler & Liben, 2007), we might expect that the adolescents who make the other gender salient by making these negative comments will develop even more negative stereotypes over the course of the year as compared with the group who has not focused such negative attention on the out-group of other- gender peers.

Our results also document evidence of peer influence for eighth graders on felt pressure from peers for gender conformity, such that adolescents changed their levels of peer felt pressure to become similar to the levels of felt pressure experienced by their friends. Because this particular dimension of gender self-concept focuses on the adolescents’ perceptions of what is accepted and enforced by their peer group, it may be less surprising that we found peer influence on this aspect of GI among friends. Felt pressure for gender conformity is a relevant aspect of social cognition for this developmental age because adolescents are par- ticularly sensitive to peer evaluations (O’Brien & Bierman, 1988). This finding takes on added significance, given research showing that experiencing high levels of pressure for gender conformity has detrimental consequences in terms of psychological well-being (Egan & Perry, 2001; Yunger et al., 2004).

The findings concerning intergroup bias and felt pressure for gender conformity were consistent with our expectation that the between-gender group dimensions of GI might be more suscepti- ble to peer influence than the within-gender dimensions of GI. It is interesting to consider how this influence occurs. We hypothesized this pattern because statements about intergroup bias and pressure to conform to gender norms might be shared among friends or overheard, similar to what has been documented for discussions of different ethnic-racial groups (Kiesner et al., 2003). Due to these overt conversations and greater visibility, between-gender dimen- sions of GI appear to be easier targets of peer influence than within-gender GI self-concepts GI (gender typicality and content- ment) which are more internal cognitive and affective states. As a result, both felt pressure and intergroup bias are likely to be influenced through direct peer actions of modeling bias, or rein- forcing, punishing, or ignoring particular statements made in the peer group. Indirect means are also likely effective in promoting socialization as adolescents have strong motivation to “fit in” with their peers by showing similar intergroup biases, recognizing peers’ attitudes about gender nonconformity or conformity, and changing their behavior to better match their friends. Although specific mechanisms of peer influence were not explicitly tested, they likely involve a combination of direct or indirect social transmission of ideas, rewards and punishments, and modeling. Investigation of the specific methods of peer influence and the directionality of these effects are worthy topics for future research.

One of the issues in understanding peer influence is that friends often select others similar to themselves thereby confounding peer influence effects with network selection effects. Our use of longi- tudinal SNA methods permitted controlling for these confounding network selection processes including the role of gender in net- work selection and the contributions of GI dimensions in network selection. Our results on network selection by gender demonstrated that adolescents across both grades tended to befriend each other if they were of the same gender. This observation is in line with nearly universal patterns observed in childhood and adolescence in which preferences are shown for same-gender peers (e.g., Mac-

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1588 KORNIENKO, SANTOS, MARTIN, AND GRANGER

coby, 1998; Mehta & Strough, 2009) as well as with other social network studies of gender development (Martin et al., 2013).

The present study was also the first to examine whether adoles- cents select peers based on the four dimensions of GI. We antic- ipated that adolescents may prefer to form friendships with others who have similar levels of GI dimensions, including both between- and within-gender self-concepts, because past research has shown that homophily or similarity on gender-typed behaviors drives peer affiliation dynamics among young children (Martin et al., 2013). Consistent with these expectations, we found that eighth graders were more likely to select each other if they reported similar levels of gender typicality. This finding is intriguing since it suggests that, although gender typicality is an internal and global part of personal identity, there may be external manifestations of typical- ity that are apparent to adolescents at least by eighth grade. Egan and Perry (2001) found that gender typicality was positively cor- related with gender-typed activities, so it could be that adolescents use gender typed activity engagement as signals or cues of simi- larity in gender typicality. However, if that is the case, it is surprising that younger adolescents do not also use engagement in gender-typed activities as signals. Further research is needed to explore whether younger adolescents are less motivated to select friends based on similar levels of gender typicality or whether they are less susceptible to the cues that suggest gender typicality.

As anticipated, we found that homophily on gender contented- ness also increased the likelihood of friendship ties among seventh and eighth graders. Being content with one’s gender appeared to be a salient cue for friend selection. Contentment may be marked by acceptance of one’s gender category and may not be visibly manifested; however, lack of contentment, that is, expressions or actions that signal a lack of happiness with being of one’s gender, may be more salient (Carver et al., 2003; Yunger et al., 2004). Because past research has shown that gender contentedness and typicality are positively correlated with social skills, self-esteem, and greater peer acceptance (Carver et al., 2003; Smith & Leaper, 2006; Yunger et al., 2004), future research would benefit from examining these associations within mediation analytical frame- work to elucidate the mechanisms linking within-gender self- concepts, peer relationships, and wellbeing.

We expected that between-gender aspects of GI would be del- eterious for friendship selection dynamics and subsequently neg- atively associated with the number of friendship nominations sent out to and received from the grade-mates because the past research has shown that intergroup bias and pressure for gender conformity often translate into externalizing problems and difficulties with peers (Carver et al., 2003; Smith & Leaper, 2006; Yunger et al., 2004). Our findings provided partial support to these hypotheses and revealed that in seventh grade, intergroup bias (i.e., holding negative feelings toward the other gender) was negatively associ- ated with the likelihood of being nominated as a friend by others. It is plausible that holding more negative attitudes toward the other gender constrains friendship formation opportunities (i.e., across gender lines), which may result in lower attractiveness of an individual with higher levels of intergroup bias as a friend. This pattern replicates a previously suggested idea that intergroup bias may be linked to peer relationship difficulties (Powlishta, 1995) because youth with higher negative attitudes toward the other gender grouping may be perceived as immature and hostile. Our finding confirms such a view because youth with higher levels of

intergroup bias are less attractive as a friend to their peer group. In contrast, this pattern indicates that greater openness to the other gender may be adaptive in terms of allowing for greater opportu- nities to form friendships among youth during middle school, when adolescents’ social, emotional, cognitive, and physical attri- butes change and their potential pool of friends increases com- pared with elementary school. It is interesting, however, that this feature of network selection did not continue to be of importance in older middle schoolers.

Turning to association between within-gender self-concepts and friendships network selection, we expected positive associations between gender typicality and contentedness with the number of friendship nominations sent and received because these aspects of GI have been linked to greater peer acceptance and social skills. Our results provided partial support and showed that gender con- tentedness was also positively associated with the number of friendship nominations that seventh and eighth-grade students sent out suggesting a better social integration in peer networks. Al- though these findings are intriguing, the question of how adoles- cents evaluate and determine others’ gender contentedness is an important direction for future research. Contrary to our expecta- tions, we documented that gender typicality was negatively asso- ciated with the number of friendship ties that eighth graders sent out. It is plausible that having higher levels of perceived proto- typicality with one’s gender group might operate similar as inter- group bias in that it might also restrict the range of opportunities afforded to form friendships with peers who possess low forms of typicality (this pattern is supported in our models by a significant preference to select friends who have similar levels of gender typicality). For example, it could be that adolescents who view themselves as very typical of their gender group prefer to affiliate with others who also view themselves that way, thus potentially restricting the range of friendship formation options.

Our final exploratory goal was to examine whether the magni- tude of peer influence differed between boys and girls. Our results do not reveal any gender differences. This finding is somewhat surprising given that other studies have suggested boys experience stronger peer pressures to conform to gender norms than do girls (Bussey & Bandura, 1992) and given that boys report higher levels of felt pressure from parents, teachers, peers, and self than do girls (Egan & Perry, 2001; Carver et al., 2003; Yunger et al., 2004). What is not clear from earlier studies that ask adolescents and older children to report on felt pressure is the source of felt pressure. That is, the felt pressure scale combines items about pressure from parents, peers, and self; therefore, it is unclear which of these sources is stronger for boys than girls. It may be that parents hold more strict expectations in terms of gender norm conformity for boys than for girls, or that boys hold themselves to stricter standards for conformity than do girls. Future research exploring gender differences and similarities in actual versus per- ceived peer pressures for conformity, and whether conformity pressures are explicit or implicit may provide additional insight into this finding.

Overall, our study adds to a growing body of research highlight- ing the impact that peers have on gender development. Previous research using the same approach to the study of social network influence on gender development illustrated that younger children became more similar to peers in their gender-typed behaviors over a few months time (Martin et al., 2013). The present study was the

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1589PEER INFLUENCE ON GENDER IDENTITY

first to explore peer effects on GI explicitly, and showed that peers influence one another on several dimensions of GI. The latter is an important addition to this body of work as behaviors may be relatively easily influenced among peers because they are overt (Martin et al., 2013); however, the present study shows that peers also exert influence even on less visible processes such as the development of one’s representations of gender. Our research further illustrates the dialectic and social nature of how GI devel- ops within the context of peer networks. Thus, we expand the current understanding of multiple socialization contexts of GI development (i.e., family, peers, mass media; Galambos, 2004; Leaper & Friedman, 2007).

Limitations and Future Directions

Although the present study has numerous strengths including the use of longitudinal SNA and employing a large sample of ethnically diverse students, it also has limitations. One is that the data on friendship networks was constrained to each grade rather than to the whole school. This choice of network boundary was informed by prior research demonstrating that grade level is the relevant sampling frame for middle school students (Cairns, Xie, & Leung, 1998). Nonetheless, if children have broader networks spanning beyond the grade level, then we may have missed older or younger peers who also could be salient sources of influences on GI. Also, we did not attempt to gather information about adolescents’ out-of-school networks in their neighborhoods, again potentially missing some potential sources of peer influence. Sec- ond, some might question our use of self-report measures; how- ever, Egan and Perry (2001) have argued forcefully that GI must be self-reported as these are personal, internal summary judgments about one’s view of self and others. Third, we modified the Egan and Perry’s (2001) scale and complemented it with Sellers et al.’s (1998) measure of identity, which may have subtly altered how participants responded to them. However, we believe that it is unlikely that these subtle variations influenced how adolescents responded to these scales.

Future directions to pursue include exploring how and through which mechanisms (i.e., adherence to group norms or reinforce- ment) GI attitudes and feelings of pressure are transmitted (Brech- wald & Prinstein, 2011). Another line of related research could focus on identifying the role of individual differences in socio- evaluative concerns and rejection sensitivity (Downey, Lebolt, Rincón, & Freitas, 1998) as contributing factors to make some adolescents more or less susceptible to peer influence. Because between- and within-gender self-concepts are posited to be shaped by sociocognitive processes transpiring within own- and other- gender collectives (Pauletti et al., 2014), future work would be prudent to examine the role of proportion of same-gender friends in peer influence and network selection on GI dimensions. It would also be interesting to explore other methods assessing GI that might provide somewhat different findings during adolescence, such as separating own-gender and other-gender typicality. Fur- thermore, research on related aspects of GI, such as centrality or salience, may provide additional insights into how adolescents develop their identities and how they influence their peers during this developmental transition.

Future research would benefit from use of diverse methodolo- gies and conceptualizations of peer relationships (Gifford-Smith &

Brownell, 2003) in identifying the mechanisms through which peers influence GI. The present study employed a longitudinal SNA approach to study peer influence to allow the modeling of peer effects of multiple friends of the focal individual while statistically controlling for how these friendships have been cre- ated and are being maintained. Assessing the influence of multiple friends has advantages of capturing peer group dynamics, how- ever, if the close or best friends are the primary source of peer influence on GI, then dyadic, or actor-partner interdependence modeling may provide the appropriate analytical framework to study peer influence on GI (Kenny, Kashy, & Cook, 2006). Close friendships, as enduring and high quality reciprocated relationships (for reviews, see Bagwell & Schmidt, 2011; Berndt, 2002) also represent a rich social context in which identity evolves. Finally, using experimental paradigms would be beneficial in establishing causal mechanisms of peer influence. For instance, young chil- dren’s choices of toys have been influenced by how peers respond to those choices in experimental studies (e.g., Langlois & Downs, 1980). Although it might be challenging to design studies that manipulate potential peer responses to adolescents’ aspects of GI, it would be possible to develop scenarios in which adolescents indicate how they would respond to unfamiliar peers given differ- ent types of GI, similar to the study conducted by Jewell and Brown (2014).

Conclusion

To the best of our knowledge, the present study is the first to assess how peers influenced changes in adolescents’ GI. Using a large sample of ethnically diverse seventh and eighth graders, we found evidence that adolescents use gender typicality, contented- ness, and gender when they select their friends. The evidence concerning selection based on GI dimensions was relatively lim- ited, although gender selection played a large role in structuring adolescents’ networks. Most interestingly, we provided strong evidence that friends’ self-reported levels of felt pressure and their intergroup attitudes were influential in changing adolescents’ GI, such that youth became similar to their friends in these beliefs over time. This research extends prior research on peer influence in adolescence by illustrating that attitudes about gender and internal pressures to conform to gender norms also have the potential to be transmitted among youth.

References

Aboud, F., & Mendelson, M. (1996). Determinants of friendship selection and quality: Developmental perspectives. In W. Bukowski, A. New- comb, & W. Hartup (Eds.), The company they keep: Friendships in childhood and adolescence. Cambridge studies in social and emotional development (pp. 87–113). Cambridge, UK: Cambridge University Press.

Bagwell, C. L., & Schmidt, M. E. (2011). Friendships in childhood and adolescence. New York, NY: Guilford Press.

Berndt, T. J. (2002). Friendship quality and social development. Current Directions in Psychological Science, 11, 7–10. http://dx.doi.org/10 .1111/1467-8721.00157

Bigler, R. S., & Liben, L. S. (2007). Developmental intergroup theory explaining and reducing children’s social stereotyping and prejudice. Current Directions in Psychological Science, 16, 162–166. http://dx.doi .org/10.1111/j.1467-8721.2007.00496.x

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1590 KORNIENKO, SANTOS, MARTIN, AND GRANGER

Brechwald, W. A., & Prinstein, M. J. (2011). Beyond homophily: A decade of advances in understanding peer influence processes. Journal of Re- search on Adolescence, 21, 166–179. http://dx.doi.org/10.1111/j.1532- 7795.2010.00721.x

Bussey, K., & Bandura, A. (1992). Self-regulatory mechanisms governing gender development. Child Development, 63, 1236–1250. http://dx.doi .org/10.2307/1131530

Bussey, K., & Bandura, A. (1999). Social cognitive theory of gender development and differentiation. Psychological Review, 106, 676–713. http://dx.doi.org/10.1037/0033-295X.106.4.676

Cairns, R., Xie, H., & Leung, M. C. (1998). The popularity of friendship and the neglect of social networks: Toward a new balance. New Direc- tions for Child and Adolescent Development, 109, 25–53. http://dx.doi .org/10.1002/cd.23219988104

Carver, P. R., Yunger, J. L., & Perry, D. G. (2003). Gender identity and adjustment in middle childhood. Sex Roles, 49, 95–109. http://dx.doi .org/10.1023/A:1024423012063

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed., pp. 284–285). Hillsdale, NJ: Erlbaum.

DeLay, D., Ha, T., Van Ryzin, M., Winter, C., & Dishion, T. J. (2016). Changing friend selection in middle school: A social network analysis of a randomized intervention study designed to prevent adolescent problem behavior. Prevention Science, 17, 285–294. http://dx.doi.org/10.1007/ s11121-015-0605-4

Dishion, T. J., & Tipsord, J. M. (2011). Peer contagion in child and adolescent social and emotional development. Annual Review of Psy- chology, 62, 189–214. http://dx.doi.org/10.1146/annurev.psych.093008 .100412

Downey, G., Lebolt, A., Rincón, C., & Freitas, A. L. (1998). Rejection sensitivity and children’s interpersonal difficulties. Child Development, 69, 1074–1091. http://dx.doi.org/10.1111/j.1467-8624.1998.tb06161.x

Drury, K., Bukowski, W. M., Velásquez, A. M., & Stella-Lopez, L. (2013). Victimization and gender identity in single- and mixed-sex schools: Examining variations in pressure to conform to gender norms. Sex Roles, 69, 442–454. http://dx.doi.org/10.1007/s11199-012-0118-6

Egan, S. K., & Perry, D. G. (2001). Gender identity: A multidimensional analysis with implications for psychosocial adjustment. Developmental Psychology, 37, 451– 463. http://dx.doi.org/10.1037/0012-1649.37.4 .451

Galambos, N. L. (2004). The gender and gender role development in adolescence. In R. M. Lerner & L. Steinberg (Eds.) Handbook of adolescent psychology (2nd ed., pp. 233–262). Hoboken, NJ: Wiley. http://dx.doi.org/10.1002/9780471726746.ch8

Gibbons, F. X., Gerrard, M., & Lane, D. J. (2003). A social reaction model of adolescent health risk. Social psychological foundations of health and illness, 107–136. http://dx.doi.org/10.1002/9780470753552.ch5

Gifford-Smith, M. E., & Brownell, C. A. (2003). Childhood peer relation- ships: Social acceptance, friendships, and peer networks. Journal of School Psychology, 41, 235–284. http://dx.doi.org/10.1016/S0022- 4405(03)00048-7

Harris, J. R. (1995). Where is the child’s environment? A group socializa- tion theory of development. Psychological Review, 102, 458–489. http://dx.doi.org/10.1037/0033-295X.102.3.458

Horn, S. S. (2008). The multifaceted nature of sexual prejudice: How adolescents reason about sexual orientation and sexual prejudice. In S. R. Levi & M. Killen (Eds.), Intergroup attitudes and relations in childhood through adulthood (pp. 173–188). New York, NY: Oxford University Press.

Hunter, D. R., Goodreau, S. M., & Handcock, M. S. (2008). Goodness of fit of social network models. Journal of the American Statistical Asso- ciation, 103, 248–258. http://dx.doi.org/10.1198/016214507000000446

Jewell, J. A., & Brown, C. S. (2014). Relations among gender typicality, peer relations, and mental health during early adolescence. Social De- velopment, 23, 137–156. http://dx.doi.org/10.1111/sode.12042

Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis. New York, NY: Guilford Press.

Kiesner, J., Maass, A., Cadinu, M., & Vallese, I. (2003). Risk factors for ethnic prejudice during early adolescence. Social Development, 12, 288–308. http://dx.doi.org/10.1111/1467-9507.00234

Kornienko, O., & Santos, C. E. (2014). The effects of friendship network popularity on depressive symptoms during early adolescence: Modera- tion by fear of negative evaluation and gender. Journal of Youth and Adolescence, 43, 541–553. http://dx.doi.org/10.1007/s10964-013- 9979-4

Langlois, J. H., & Downs, A. C. (1980). Mothers, fathers, and peers as socialization agents of sex-typed play behaviors in young children. Child Development, 51, 1237–1247. http://dx.doi.org/10.2307/1129566

Larson, R., & Richards, M. H. (1991). Daily companionship in late childhood and early adolescence: Changing developmental contexts. Child Development, 62, 284–300. http://dx.doi.org/10.2307/1131003

Leaper, C., & Brown, C. S. (2008). Perceived experiences with sexism among adolescent girls. Child Development, 79, 685–704. 0009-3920/ 2008/7903-0013

Leaper, C., & Friedman, C. (2007). The socialization of gender. In J. Grusec & P. Hasting (Eds.), Handbook of socialization: Theory and research (pp. 561–587). New York, NY: Guilford Press.

Lee, E. E., & Troop-Gordon, W. (2011). Peer processes and gender role development: Changes in gender atypicality related to negative peer treatment and children’s friendships. Sex Roles, 64, 90–102. http://dx .doi.org/10.1007/s11199-010-9883-2

Lospinoso, J. A., Schweinberger, M., Snijders, T. A. B., & Ripley, R. M. (2011). Assessing and accounting for time heterogeneity in stochastic actor oriented models. Advances in Data Analysis and Classification, 5, 147–176. http://dx.doi.org/10.1007/s11634-010-0076-1

Maccoby, E. (1998). The two sexes: Growing up apart, coming together. United States: First Harvard University Press.

Martin, C. L., & Fabes, R. A. (2001). The stability and consequences of young children’s same-sex peer interactions. Developmental Psychol- ogy, 37, 431–446. http://dx.doi.org/10.1037/0012-1649.37.3.431

Martin, C. L., Kornienko, O., Schaefer, D. R., Hanish, L. D., Fabes, R. A., & Goble, P. (2013). The role of sex of peers and gender-typed activities in young children’s peer affiliative networks: A longitudinal analysis of selection and influence. Child Development, 84, 921–937. http://dx.doi .org/10.1111/cdev.12032

McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27, 415– 444. http://dx.doi.org/10.1146/annurev.soc.27.1.415

Mehta, C. M., & Strough, J. (2009). Sex segregation in friendships and normative contexts across the life span. Developmental Review, 29, 201–220. http://dx.doi.org/10.1016/j.dr.2009.06.001

Michaels, M. L., Barr, A., Roosa, M. W., & Knight, G. (2007). Self- esteem: Assessing measurement equivalence in a multiethnic sample of youth. The Journal of Early Adolescence, 27, 269–295. http://dx.doi .org/10.1177/0272431607302009

O’Brien, S. F., & Bierman, K. L. (1988). Conceptions and perceived influence of peer groups: Interviews with preadolescents and adoles- cents. Child Development, 59, 1360–1365.

Parker, J. G., Rubin, K. H., Erath, S. A., Wojslawowicz, J. C., & Buskirk, A. A. (2006). Peer relationships, child development, and adjustment: A developmental psychopathology perspective. In D. Cicchetti & D. J. Cohen (Eds.), Developmental psychopathology: Vol. 1. Theory and methods (2nd ed., pp. 419–493). New York, NY: Wiley.

Pauletti, R. E., Cooper, P. J., & Perry, D. G. (2014). Influences of gender identity on children’s maltreatment of gender-nonconforming peers: A person � target analysis of aggression. Journal of Personality and Social Psychology, 106, 843–866. http://dx.doi.org/10.1037/a0036037

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1591PEER INFLUENCE ON GENDER IDENTITY

Poteat, V. P. (2007). Peer group socialization of homophobic attitudes and behavior during adolescence. Child Development, 78, 1830–1842. http:// dx.doi.org/10.1111/j.1467-8624.2007.01101.x

Powlishta, K. K. (1995). Intergroup processes in childhood: Social cate- gorization and sex role development. Developmental Psychology, 31, 781–788. http://dx.doi.org/10.1037/0012-1649.31.5.781

Priess, H. A., Lindberg, S. M., & Hyde, J. S. (2009). Adolescent gender- role identity and mental health: Gender intensification revisited. Child Development, 80, 1531–1544. http://dx.doi.org/10.1111/j.1467-8624 .2009.01349.x

Ripley, R. M., Snijders, T. A. B., Boda, Z., Voros, A., & Preciado, P. (2016). Manual for RSiena. Oxford, UK: University of Oxford, Depart- ment of Statistics; Nuffield College. Retrieved from http://www.stats .ox.ac.uk/~snijders/siena/RSiena_Manual.pdf

Ruble, D. N., & Dweck, C. S. (1995). Self-perceptions, person concep- tions, and their development. In N. Eisenberg (Ed.), Review of person- ality and social psychology: Vol. 15. Social development (pp. 109–139). Thousand Oaks, CA: Sage.

Ruble, D. N., Martin, C. L., & Berenbaum, S. A. (2006). Gender devel- opment. In W. Damon & R. M. Lerner (Series Eds.) & N. Eisenberg (Vol. Ed.), Handbook of child psychology: Vol 3. Social, emotional, and personality development (6th ed., pp. 858–932). Hoboken, NJ: Wiley.

Santos, C. E., Kornienko, O., & Rivas-Drake, D. (in press). Peer influence on ethnic-racial identity development: A multi-site investigation. Child Development.

Santos, C. E. & Updegraff, K. A. (2014). Feeling typical, looking typical: Physical appearance and ethnic identity among Mexican-origin youth. Journal of Latina/o Psychology, 2, 187–199. http://dx.doi.org/10.1037/ lat0000023

Schweinberger, M. (2012). Statistical modelling of network panel data: Goodness of fit. The British Journal of Mathematical and Statistical Psychology, 65, 263–281. http://dx.doi.org/10.1111/j.2044-8317.2011 .02022.x

Sellers, R. M., Smith, M. A., Shelton, J. N., Rowley, S. A. J., & Chavous, T. M. (1998). Multidimensional model of racial identity: A reconceptu- alization of African American racial identity. Personality and Social Psychology Review, 2, 18 –39. http://dx.doi.org/10.1207/s15327957 pspr0201_2

Smith, T. E., & Leaper, C. (2006). Self-perceived gender typicality and the peer context during adolescence. Journal of Research on Adolescence, 16, 91–104. http://dx.doi.org/10.1111/j.1532-7795.2006.00123.x

Snijders, T. A. B., Steglich, C. E. G., & Schweinberger, M. (2007). Modeling the co-evolution of networks and behavior. In K. van Mont- fort, H. Oud, & A. Satorra (Eds.), Longitudinal models in the behavioral and related sciences (pp. 41–71). Mahwah, NJ: Erlbaum.

Snijders, T. A., van de Bunt, G. G., & Steglich, C. E. (2010). Introduction to stochastic actor-based models for network dynamics. Social Net- works, 32, 44–60. http://dx.doi.org/10.1016/j.socnet.2009.02.004

Spence, J. T. (1985). Gender identity and its implications for concepts of masculinity and femininity. In T. B. Sonderegger (Ed.), Nebraska Sym- posium on Motivation: Psychology and gender (Vol. 32, pp. 59–96). Lincoln, NE: University of Nebraska Press.

Veenstra, R., Dijkstra, J. K., Steglich, C., & Van Zalk, M. H. W. (Eds.). (2013). Network and behavior dynamics in adolescence. Journal of Research on Adolescence, 23. http://dx.doi.org/10.1111/jora.120700

Walton, G. M., Cohen, G. L., Cwir, D., & Spencer, S. J. (2012). Mere belonging: The power of social connections. Journal of Personality and Social Psychology, 102, 513–532. http://dx.doi.org/10.1037/a0025731

Wimmer, A., & Lewis, K. (2010). Beyond and below racial homophily: ERG models of a friendship network documented on Facebook. Amer- ican Journal of Sociology, 116, 583–642. http://dx.doi.org/10.1086/ 653658

Yee, M. D., & Brown, R. (1992). Self-evaluations and intergroup attitudes in children aged three to nine. Child Development, 63, 619–629. http:// dx.doi.org/10.2307/1131350

Young, R., & Sweeting, H. (2004). Adolescent bullying, relationships, psychological well-being, and gender-atypical behavior: A gender diag- nostic approach. Sex Roles, 50, 525–537. http://dx.doi.org/10.1023/B: SERS.0000023072.53886.86

Yunger, J. L., Carver, P. R., & Perry, D. G. (2004). Does gender identity influence children’s psychological well-being? Developmental Psychol- ogy, 40, 572–582. http://dx.doi.org/10.1037/0012-1649.40.4.572

Zosuls, K. M., Miller, C. F., Ruble, D. N., Martin, C. L., & Fabes, R. A. (2011). Gender development research in Sex Roles: Historical trends and future directions. Sex Roles, 64, 826–842. http://dx.doi.org/10.1007/ s11199-010-9902-3

Received May 17, 2015 Revision received June 17, 2016

Accepted June 29, 2016 �

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1592 KORNIENKO, SANTOS, MARTIN, AND GRANGER