Article Critique 1

Adolescent Internalizing Symptoms and the “Tightknittedness” of

Friendship Groups

Sonja E. Siennick and Mayra Picon Florida State University

Adolescents with depression have lower peer status overall, but tend to befriend each other. We examined the “tightknittedness” of their friendship groups by testing whether adolescent friendship groups’ average levels of or vari- ability in internalizing symptoms predict group cohesiveness. We used four waves (9th–12th grades) of survey and social network data on 3,013 friendship groups from the PROmoting School-Community-University Partnerships to Enhance Resilience study. Friendship groups with higher average depressive symptoms were less cohesive; groups with higher average anxiety symptoms had greater reciprocity. Groups with greater variability in depressive symptoms had greater density; variability in anxiety symptoms was not consistently associated with cohesion. The friendship groups of depressed adolescents appear less cohesive than the “typical” adolescent friendship group.

When compared with their peers, adolescents with more depressive symptoms have fewer friends, have less stable friendships, and are more often victimized and rejected by their peers (Chan & Poulin, 2009; Kochel, Ladd, & Rudolph, 2012; Rose et al., 2011; Stice, Ragan, & Randall, 2004). Yet depressive symptoms also are a basis for friend- ship formation, such that adolescents experiencing these symptoms tend to be friends with each other (Cheadle & Goosby, 2012; Hogue & Steinberg, 1995; Schaefer, Kornienko, & Fox, 2011). This means that even if they have lower status in their larger peer networks, many adolescents with depressive symptoms do have friends, and those friends often have depressive symptoms them- selves. Thus many youth with depressive symp- toms are likely embedded in friendship groups of adolescents with similar symptoms. Yet we do not know whether these friendship groups are as cohe- sive as the groups formed by youth without depressive symptoms, or whether they are

structurally weaker and thus not comparable sub- stitutes, at least in terms of cohesion, for typical friendship groups. Most studies of depressive symptoms and peer networks have focused on dyadic interactions or on individual adolescents’ status within entire social networks, rather than on friendship groups.

This study examined whether friendship groups comprised of adolescents with more depressive symptoms are smaller and “looser,” or less tight- knit, than groups characterized by fewer depres- sive symptoms. It also examined whether groups whose members vary more in their levels of depressive symptoms are smaller and less tight- knit. Finally, it distinguished between symptoms of depression and symptoms of anxiety, which stud- ies suggest may have opposite effects on friendship cohesion (Rose et al., 2011). To our knowledge, this is the first paper to describe the internal cohesive- ness of friendship groups with members who have varying constellations of depressive and anxiety symptoms, as well as one of the only papers to examine characteristics of depressed youths’ friendships at the group or “clique” level, versus at the dyad level or at the ego level (examining individual youths’ status in the larger friendship network).

How Do Depressed Friendship Groups Form?

Research has uncovered four microlevel processes that appear to underlie the similarity between

We thank Derek Kreager, Jim Moody, and David Schaefer for their helpful comments. Grants from the W. T. Grant Foundation (8316), National Institute on Drug Abuse (R01-DA018225), and National Institute of Child Health and Development (R24- HD041025) supported this research. The analyses used data from PROSPER, a project directed by R. L. Spoth, funded by the National Institute on Drug Abuse (RO1-DA013709) and the National Institute on Alcohol Abuse and Alcoholism (AA14702). The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Requests for reprints should be sent to Sonja E. Siennick, Col-

lege of Criminology and Criminal Justice, Florida State Univer- sity, 112 S. Copeland Street, Tallahassee, FL 32306. E-mail: ssiennick@fsu.edu

© 2019 Society for Research on Adolescence DOI: 10.1111/jora.12484

JOURNAL OF RESEARCH ON ADOLESCENCE, 30(S2), 391–402

friends on various emotional and behavioral dimensions: selection, de-selection, influence, and social withdrawal. The first, selection or homo- phily, is based on the notion that people are attracted to similar others (Berger & Calabrese, 1975) and implies that friends’ similarity in depres- sive symptoms occurs because adolescents choose to befriend youth who have levels of symptoms similar to theirs. Indeed, an early longitudinal net- work study of high school students found that selection partly explained adolescent friends’ simi- larity in internalized distress (Hogue & Steinberg, 1995). These findings have been replicated by more recent studies on depressive symptoms (Kiuru, Burk, Laursen, Nurmi, & Salmela-Aro, 2012; Van Zalk, Kerr, Branje, Stattin, & Meeus, 2010) and on social anxiety symptoms as well (Van Zalk, Van Zalk, Kerr, & Stattin, 2011). Moreover, one of these studies found evidence of a second process—de- selection. Specifically, this study found that youth were more likely to terminate friendships with adolescents who became more dissimilar in terms of depressive symptoms over a 1-year period (Kiuru et al., 2012). Thus, one reason why adoles- cent friends have similar levels of depressive symptoms is that adolescents are more likely to become and stay friends with other adolescents who resemble them on this dimension.

Another reason involves a third process. Influ- ence, or socialization, suggests that friends have similar levels of depressive symptoms because they influence each other’s symptoms and become increasingly similar over time. This occurs through mechanisms such as co-rumination, or excessively discussing one’s problems within a friendship (Rose, 2002). Hogue and Steinberg’s (1995) early network study examined influence processes as well, and found that males’, but not females’, inter- nalized distress grew more similar to their friend- ship groups’ levels over time. A later study using a sample of third graders similarly found that chil- dren’s levels of depressive and social anxiety symptoms became more similar to the average level of their friends’ symptoms over a 1-year per- iod (Mercer & Derosier, 2010). Further evidence of depression socialization comes from two longitudi- nal studies of dyads that found that adolescents’ depressive symptoms were predicted by their best friends’ symptoms, controlling for their initial levels of depressive symptoms (Prinstein, 2007; Ste- vens & Prinstein, 2005). A more recent study found similar evidence of socialization of depressive symptoms, but only among female best friend dyads (Giletta et al., 2011). Other internalizing

symptoms may be transmitted in the same way. One study, for example, found that friends, espe- cially girls, tended to influence each other’s levels of social anxiety over time (Van Zalk, Van Zalk, Kerr, et al., 2011). This influence may be especially strong in certain friendship groups: These same authors also found that youth who affiliated with “Radical” crowds, those comprised of Goths and Punks, were most influenced by their peers’ social anxiety (Van Zalk, Van Zalk, & Kerr, 2011).

Finally, a fourth process of social withdrawal may contribute to the resemblance among friends in depressive symptoms. Under this mechanism, adolescents with depressive symptoms may with- draw from peer interaction to avoid experiencing stigma and negative interactions with others, or because they perceive themselves as socially incompetent (Altmann & Gotlib, 1988; Rudolph, Hammen, & Burge, 1994). This social withdrawal may exclude adolescents with depressive symp- toms from normative peer networks, leaving them to befriend other marginalized youth, including those who are also experiencing symptoms. Consis- tent with this, one study found that youth with few friends, who often also had symptoms of depression, tended to befriend each other, and this indirectly created depression similarity among friends (Schaefer et al., 2011).

In this paper, we do not test these microlevel processes. Instead, we take as our starting point the friendship groups that are formed by these pro- cesses, and we examine the characteristics of those groups. As we next describe, the processes that make friends similar on depressive symptoms might also create cohesive, tight-knit friendship groups. However, a separate literature on group functioning suggests that these symptoms instead might undermine group cohesiveness.

Potential Associations Between Depressive Symptoms and Group Structure

Membership in cohesive friendship groups could offer many potential benefits to adolescents who are experiencing depressive symptoms. These groups are important contexts for development and for flows of influence, information, and sup- port among peers (Brechwald & Prinstein, 2011; De�girmencio�glu, Urberg, Tolson, & Richard, 1998; Ellis & Zarbatany, 2007; Simons-Morton & Farhat, 2010). Tight-knit groups of friends with depressive symptoms could offer group members more oppor- tunities for supportiveness, intimacy, and sharing, which could enhance their well-being (Buhrmester,

392 SIENNICK AND PICON

1990). They could also provide unified buffers against the bullying and peer victimization that adolescents with depressive symptoms are at risk for experiencing (Kaltiala-Heino, Fr€ojd, & Mart- tunen, 2010; Kochel et al., 2012).

Although research suggests that adolescents with depressive symptoms have lower quality peer relations in general, it is possible that these adoles- cents still manage to form strong, cohesive friend- ship groups. Theoretically, they have friendship problems because their peers find some of their behaviors aversive, such as their excessive reassur- ance-seeking, their redirecting conversations to focus on their problems, and their acting socially helpless (Agoston & Rudolph, 2013; Coyne, 1976; Prinstein, Borelli, Cheah, Simon, & Aikins, 2005; Schwartz-Mette & Rose, 2016). However, if pro- cesses of selection are at play—that is, if adoles- cents with depressive symptoms choose to form friendship groups with each other despite these behaviors—then the behaviors may not undermine the internal cohesiveness of those groups. Indeed, there is evidence that some level of emotional dis- tress may actually enhance these friendships by prompting social support and self-disclosure (Hill & Swenson, 2014). In addition, if processes of socialization are at play, then the increased similar- ity among members of depressed groups may enhance group cohesiveness. In fact, research sug- gests that processes such as co-rumination serve as protective factors for friendship problems, even when they worsen depressive and anxiety symp- toms (Rose, 2002; Rose, Carlson, & Waller, 2007). It thus is possible that groups comprised of adoles- cents with depressive symptoms have strong inter- nal connections despite their members’ symptomatology and lower status within the larger peer network.

Yet that possibility is in contrast with predic- tions from research on the impact of emotions on small group functioning, which provides some of the only existing group-level evidence on this topic. That work, which tends to use adult samples, has shown that group members’ affect and moods do influence group performance and cohesion (Kelly & Jones, 2012). Much of this research has focused on group members’ positive emotions, which have been found to facilitate within-group communica- tion, reduce within-group conflict, and promote within-group cooperation and bonding (Spoor & Kelly, 2004). If negative emotions have the opposite effect, depressive symptoms could worsen these features of adolescent friendship groups. This may be especially likely if adolescents with depressive

symptoms befriend each other less because they like each other than because they have few alterna- tive friendship opportunities, as under the social withdrawal mechanism (Schaefer et al., 2011). Without the positive mood and strong mutual lik- ing that promote group cohesiveness, friendship groups composed of adolescents with depressive symptoms may be less tight-knit than the friend- ship groups formed by their less depressed peers. Research has yet to determine whether these symp- toms enhance or undermine the internal cohesive- ness of adolescent friendship groups.

Variation by Type of Internalizing Symptom and Extent of Similarity

Two considerations complicate studies of this topic. First, many past studies of depressive symptoms and peer relations have not distinguished between depressive and anxiety symptoms. Depression and anxiety have distinct developmental courses (Cum- mings, Caporino, & Kendall, 2014) and could have different associations with peer problems. Yet in this literature, depressive symptoms are often examined as part of a broader “internalizing” or “distress” construct that includes general anxiety. Notably, this combined construct does not consis- tently predict peer problems (Hill & Swenson, 2014; Hogue & Steinberg, 1995; Mcleod & Uemura, 2012). Indeed, studies that have separated the two suggest that depressive symptoms drive the harm- ful effects of internalizing symptoms on peer prob- lems (de Matos, Barrett, Dadds, & Shortt, 2003; Kennedy, Spence, & Hensley, 1989; Strauss, Lahey, Frick, Frame, & Hynd, 1988). In fact, youth with some forms of anxiety may have few peer prob- lems at all (Chen, Cohen, Johnson, & Kasen, 2009; de Matos et al., 2003). For example, general anxiety symptoms have been found to be associated with having more and higher quality friendships (Rose et al., 2011) and less hostile peer interactions (Rudolph et al., 1994). The differential effects of depressive and anxiety symptoms could stem from the fact that although both are characterized by negative affect, only depression is characterized by low positive affect (Rose et al., 2011), potentially making adolescents’ anxiety less aversive to, or even eliciting sympathy from, their peers.

A second consideration is whether heterogeneity in, versus simply the average level of, depressive symptoms within a friendship group is associated with lower group cohesiveness. A group with a moderate average level of depressive symptoms could be composed of several adolescents who

DEPRESSION AND FRIENDSHIP GROUPS 393

each are experiencing moderate depression, or a mixture of adolescents who are not depressed and adolescents who are highly depressed. In the latter case, the weaker homophily among group mem- bers in depressive symptoms could undermine group cohesiveness. Past research shows that peers who are dissimilar in mood or behavior are more likely to argue with, reject, and defriend each other, perhaps because they experience relation- ships with dissimilar others as less validating (Bl€ote, Bokhorst, Miers, & Westenberg, 2012; Chow, Tan, & Ruhl, 2015; Hafen, Laursen, Burk, Kerr, & Stattin, 2011; Van Zalk et al., 2010; Wright, Giam- marino, & Parad, 1986). In experimental studies as well, people who are experiencing depression, anx- iety, or bad moods prefer others who have the same feelings over others who do not (Baker, Hud- son, & Taylor, 2014; Rook, Pietromonaco, & Lewis, 1994; Rosenblatt & Greenberg, 1991). If dissimilar- ity has the same impact on the multiple relation- ships within friendship groups, then groups with more internal variation on depressive or anxiety symptoms could be less cohesive.

This Study

This study used survey and social network data from a large community sample of adolescents to examine the friendship groups of distressed adoles- cents. First, we examined whether groups with greater average within-group distress are smaller and more loosely connected. In doing so, we distin- guished between depressive and anxiety symp- toms, which are often comorbid but which may have opposite effects on adolescent friendships. Second, we also examined whether groups whose members have more dissimilar levels of distress are smaller and more loosely connected. Based on past work, we expected that group depressive symptoms would be negatively associated with measures of group cohesion; group anxiety symp- toms would be unassociated or positively associ- ated with group cohesion; and within-group variability in these symptoms would be negatively associated with measures of group cohesion.

METHOD

Data

Sample. Our data were from PROmoting School-Community-University Partnerships to Enhance Resilience (PROSPER), a place-rando- mized substance abuse prevention trial in 28 public

school districts in rural Pennsylvania and Iowa (Spoth, Greenberg, Bierman, & Redmond, 2004; Spoth et al., 2013). The districts enrolled between 1,300 and 5,200 students each, were at least 95% English-speaking, were economically diverse (with an average of 29% of families eligible for free or reduced cost school lunches), and were predomi- nantly white (61–96%). For the prevention trial, dis- tricts were blocked on size and location and randomly assigned to have universal substance abuse prevention programming implemented by community teams, or to be in the control condition. In this study, we found that intervention condition did not moderate any of our results, so we used data from both conditions.

PROSPER’s sample began with two successive cohorts of adolescents who were in sixth grade in the Fall semesters of 2002 and 2003 (N = 10,849). Half were female and nearly all (98%) were age 11 or 12. Adolescents completed machine-scored paper-and-pencil surveys about their attitudes, behaviors, and friendships during class sessions. Follow-up surveys were administered each spring from 6th–12th grades; at each wave, students who had moved into study schools were recruited to join the study. At each survey, adolescents pro- vided assent, and parents and guardians were con- tacted by mail and had the opportunity to return a form excluding their child from the study. At any given wave, approximately 3% of adolescents and 4% of parents declined to participate, and approxi- mately 5% of adolescents were absent. We used survey and social network data from 9th–12th grades (four waves), when adolescents reported on their depressive and anxiety symptoms. Because cohort did not predict any of our outcomes, we used combined data from both cohorts in our anal- yses. Our analytical sample was 48% male and 84% white; 76% lived in two parent households; 22% were eligible for free or reduced-cost lunch; and at ninth grade, 94% were age 14 or 15 (mean = 14.8).

Friendship identification and friendship group creation. The surveys asked adolescents to write the names of up to two best friends and five addi- tional close friends in their school and grade. Most adolescents (94%) nominated at least one friend; adolescents who made no nominations were more likely to be male, non-White, from single-parent households, and to be eligible for free lunch, and had higher depressive and anxiety symptom scores than adolescents who nominated friends. The PROSPER staff was able to match over 83% of friendship nominations to students on the schools’

394 SIENNICK AND PICON

class rosters. Only 2% of nominations matched multiple names on the rosters and thus could not be matched with certainty, and <1% were inappro- priate choices (e.g., pop stars). The remaining 15% matched no name on the school grade rosters and presumably were not adolescents’ grademates. Since the entire grade-level was targeted for partic- ipation in the study, these data allowed us to map out adolescents’ own personal friendship ties and the complete within-grade school friendship net- works that existed at the time of the survey, and to link adolescents with their friends’ survey responses.

PROSPER staff identified distinct mutually exclu- sive friendship groups within the larger friendship networks using computer algorithms designed to delineate groups by maximizing modularity scores, which are weighted functions of within-group com- pared to cross-group ties (Moody, 2001). The algo- rithm began with starting values based on factor analysis (Gest, Davidson, Rulison, Moody, & Welsh, 2007), and iteratively evaluated whether the modu- larity score would be improved by reassigning each student to another group, by merging any groups, or by splitting any groups. When no more improve- ments to the score were found, that set of friendship group assignments was kept. This approach identi- fied 3,090 friendship groups with at least three mem- bers. We excluded 16 (1% of) groups with 40 or more members because “group” likely means something different for groups of this size (cf. Kreager, Rulison, & Moody, 2011; Rubin, Bukowski, & Bowker, 2015). Group sizes ranged from 3 to 38 members, with an average of approximately 10. We also excluded 61 (2% of) groups that were missing information on cohesiveness. The remaining 3,013 unique friendship groups were our analytical sample.

Measures

Friendship group cohesiveness. Once discrete friendship groups were identified and the friend- ship ties between individual group members were mapped out, measures of group structure were cal- culated. Our measures of cohesiveness follow those used in past work (Gest et al., 2007; Kreager et al., 2011; Rubin et al., 2015). Group size was a count measure of the number of adolescents in the group. Density was the proportion of pairs of adolescents within the group who were friends with each other based on their nomination lists. Reciprocity was the proportion of friendship nominations from one group member to another that were reciprocated (i.e., the proportion of dyads who both included

each other on their friendship lists). Transitivity was the proportion of triads in the group in which all three members were friends (i.e., where the friend of an adolescent’s friend also was friends with that adolescent). Together these measures cap- ture the size and “tight-knittedness” of the groups.

Creation of group-level predictor variables. Our group-level predictors and control variables were aggregated versions of the survey responses given by the individual group members. For example, to create a group’s average depressive symptoms score for a given wave, we computed symptoms scale scores for the indi- vidual adolescents within the group and then averaged them. Prior to doing this, we addressed item-level miss- ing data at each wave from individual adolescents’ sur- veys using multiple imputation (Rubin, 1987). Two percent of adolescents were missing information on one or more depression or anxiety items, and 6%were miss- ing information on one or more control variables. There were onlymodest differences between these adolescents and thosewith complete data (e.g., those groups differed by an average of 1/6 of a standard deviation on the focal variables).

Depressive symptoms. The surveys included items tapping symptoms of depression, such as depressed mood (e.g., feeling sad or appearing tear- ful), feelings of guilt and worthlessness, and suicidal- ity (American Psychiatric Association, 2000). Individual adolescents’ depressive symptoms scores were the average of five items assessing whether in the past 6 months they had experienced such depres- sive symptoms (with each item scored 0–2; range of a across waves = .83–.85). As described above, group members’ scores were then averaged to create group- level average depressive symptoms scores. The within-group standard deviation of group members’ scores was our measure of the group’s variability in depressive symptoms.

Anxiety symptoms. The surveys also included items tapping symptoms of anxiety (e.g., excessive anxiety and worry, muscle tension; American Psy- chiatric Association, 2000). Individual adolescents’ anxiety symptoms scores were the average of three items assessing such symptoms (each scored 0–2; range of a across waves = .80–.83). Measures of groups’ average anxiety symptoms and groups’ variability in anxiety symptoms were created using the same strategies as for depression.

Control variables. Past research has identified demographic correlates of group structure, such as

DEPRESSION AND FRIENDSHIP GROUPS 395

gender composition and group members’ family structures and socioeconomic statuses, which may confound the associations between group structure and behavioral risk factors (Kreager et al., 2011). We thus controlled for grade level (9–12), interven- tion condition, group gender composition (all-male, all-female, or mixed gender), and the proportion of group members who were White, from two-parent families, and eligible for free or reduced-cost school lunch. Because reciprocity and transitivity depend on the number of friendship ties within a group, we also controlled for density in models of those outcomes. Descriptive statistics for these and other study variables are shown in Table 1.

Analytical Strategy

Our analyses were multilevel random effects regression models predicting each measure of group cohesiveness from the control variables and either average levels of or variability in groups’ depressive and anxiety symptoms. The adolescents’ friendship groups were clustered within 29 schools, violating the independence assumption of regres- sion analysis. For instance, as shown in Table 1, between 2% and 7% of the variance in each of our

outcomes fell between schools (vs. within schools). Our use of multilevel regression models adjusted our standard errors for this clustering through the inclusion of variance components for school (Rau- denbush & Bryk, 2002). Tests indicated that addi- tional variance components were not needed for grade (to address potential temporal autocorrela- tion in the data) or for the main independent vari- ables (to address potential variability in effects across schools). We used linear models for all group-level outcomes except the skewed count variable group size, for which we used negative binomial models.

RESULTS

We first examined the associations between groups’ average internalizing symptoms and their size and structure. Table 2 shows the results of multilevel negative binomial and linear regressions predicting each group cohesiveness outcome from group-level depressive and anxiety symptoms. Groups with higher average depressive symptom scores were significantly smaller in size, less dense, and had lower reciprocity of friendships, but did not have significantly lower transitivity.

TABLE 1 Descriptive Statistics for Study Variables (N = 3,013 Groups)

Variable

Mean/%, by Grade

Range ICC9th 10th 11th 12th

Dependent variables Group size 10.54 10.11 9.48 9.07 3–38 .02 Group density 0.29 0.30 0.30 0.29 0.04–1 .07 Group transitivity 0.42 0.43 0.41 0.42 0–1 .05 Group reciprocity 0.39 0.40 0.37 0.37 0–1 .03

Independent variables Group’s average depressive symptoms 0.30 0.28 0.27 0.23 0–1.16 .02 Within-group variation in depressive symptoms 0.34 0.34 0.33 0.30 0–0.94 .02 Group’s average anxiety symptoms 0.48 0.49 0.49 0.45 0–1.75 .01 Within-group variation in anxiety symptoms 0.49 0.49 0.48 0.47 0–0.98 .01

Control variables Grade 9 10 11 12 9–12 – Intervention condition 46% 47% 49% 50% 0–1 – All male group 25% 26% 20% 21% 0–1 .02 All female group 25% 23% 25% 21% 0–1 .01 Mixed gender group 50% 51% 56% 58% 0–1 .04 Percent of group that is White 83% 86% 86% 87% 0–1 .62 Percent of group that lives with two parents 77% 77% 76% 76% 0–1 0 Percent of group eligible for free or reduced-price lunch 26% 22% 21% 18% 0–1 .03 N 885 751 738 639

Notes. Intraclass correlation coefficients (ICC) represent the proportion of variance in the variable that is between (vs. within) schools. ICCs are based on linear models for continuous outcomes and logistic models for dichotomous outcomes. Source: PROSPER Peers Study.

396 SIENNICK AND PICON

Examination of predicted values revealed that com- pared with groups whose members had no depres- sive symptoms, groups with elevated average symptoms (1 standard deviation above the mean) had on average 0.36 fewer members (9.38 vs. 9.74), 2% lower density (0.29 vs. 0.31), and 5% less reciprocity (0.37 vs. 0.42). Supplemental analyses revealed that depressive symptoms and transitivity were negatively associated when density was removed from the model, indicating that in groups with higher average depressive symptoms there were fewer ties to friends-of-friends because there were fewer pairs of friends to begin with.

To further gauge the magnitude of the observed associations, we compared the X-standardized coef- ficients, which give the change in Y for a one-stan- dard deviation increase in X, for group-level depressive and anxiety symptoms with those for two known correlates of group cohesiveness: the proportion of members from two-parent families and the proportion eligible for free or reduced- price school lunch (Kreager et al., 2011). In the group size model, the X-standardized coefficient for depressive symptoms (bstdX = �.030) was 48% larger than the absolute value of the analogous coefficient for family structure (bstdX = .020, p < .05) and 40% smaller than the analogous coefficient for free lunch status (bstdX = �.048, p < .001). In the density model, the three coefficients were compara- ble in size (bstdX = �.010 for depressive symptoms, .008 (p < .01) for family structure, and �.009 (p < .01) for free lunch status). In the reciprocity model, the X-standardized coefficient for depres- sive symptoms (bstdX = �.015) was more than dou- ble the analogous coefficient for free lunch status (bstdX = �.007, p < .05); family structure was not a significant predictor of reciprocity.

In contrast with the findings on depressive symptoms, groups with higher average anxiety symptoms had greater reciprocity of friendships. The X-standardized version of this coefficient (bstdX = .014) was nearly double the absolute value of that for free lunch status (bstdX = �.007). Although the coefficients predicting group size and density from group anxiety symptoms also were positive, neither was statistically significant.

We next examined the associations of within- group variation in internalizing symptoms with groups’ size and structure. Table 3 shows that groups with more variability among members in depressive symptoms were less cohesive in terms of density, but depressive symptom variability did not signifi- cantly predict group size, reciprocity, or transitivity. The findings for anxiety symptom variability were very mixed. Specifically, anxiety symptom variabil- ity positively predicted group size, negatively pre- dicted density, did not predict reciprocity, and positively predicted transitivity.

Sensitivity Analyses

We ran a series of ancillary analyses to determine whether our main findings varied across grade level or gender composition of the friendship group. For example, to determine whether the group-level effects of average distress on group size differed by grade, we added two interaction terms to the first model in Table 2, namely average depressive symptoms 9 grade and average anxiety symptoms 9 grade. Of the 48 resultant interaction terms, only one was statistically significant: Vari- ability in anxiety symptoms was negatively associ- ated with group density only for all-female groups. Because of the large number of tests run we urge

TABLE 2 Negative Binomial and Linear Random Effects Regression Coefficients Predicting Adolescent Friendship Groups’ Cohesiveness From

Group Members’ Average Levels of Depressive and Anxiety Symptoms (N = 3,013 Groups)

Predictor

Outcome

Group sizea Group density Group reciprocity Group transitivity b (SE) b (SE) b (SE) b (SE)

Average depressive symptoms �.15 (.07)* �.05 (.02)* �.08 (.02)** �.04 (.03) Average anxiety symptoms .03 (.06) .03 (.02) .06 (.02)** �.01 (.02)

Notes. All models included a variance component for school and controlled for grade, intervention condition, group gender compo- sition, and group members’ race/ethnicity, family structure, and free lunch eligibility. The reciprocity and transitivity models also con- trolled for density. aNegative binomial coefficients shown; linear coefficients shown for all other models. *p < .05, **p < .01, ***p < .001. Source: PROSPER Peers Study

DEPRESSION AND FRIENDSHIP GROUPS 397

caution in interpreting this single significant inter- action.

DISCUSSION

Although adolescents with depressive symptoms have more than their share of peer problems (Chan & Poulin, 2009; Kochel et al., 2012; Rose et al., 2011), they also tend to cluster together in friendship groups, and we know little about the qualities of those groups. This study went beyond past work on depressed adolescents’ dyadic friendships, and their individual statuses in larger social networks, to reveal that friendship groups comprised of adoles- cents with more depressive symptoms are smaller and “looser,” or less tight-knit, than groups charac- terized by lower depressive symptoms. That is, their groups are composed of fewer adolescents, and the members are less likely to all be friends with each other. These associations were modest, but the effect sizes were comparable to those of known demo- graphic correlates of friendship quality, namely family structure and socioeconomic status.

Our findings perhaps contradict what might be expected based on theories of friendship formation between adolescents with depressive symptoms. For example, these friendships may form because adolescents with these symptoms have a preference for each other as friends (selection or homophily; Kiuru et al., 2012; Van Zalk et al., 2010). Friends also might change over time to become more like each other in terms of depressive symptoms (so- cialization or influence; Mercer & Derosier, 2010; Prinstein, 2007; Stevens & Prinstein, 2005). Indeed, past research suggests that both processes are likely at play (Hogue & Steinberg, 1995). With friends specifically choosing each other on the basis

of their depression and then changing to further mirror each other’s depression, we might predict that their friendship groups would be no less cohe- sive than the typical adolescent friendship. Yet our findings suggest that this is not the case.

Our findings also contrast with a small body of literature suggesting that interpersonal processes associated with depression, such as co-rumination, can enhance adolescent friendships (Rose, 2002; Rose, Carlson, & Waller, 2007). They especially contrast with a recent study that found that inter- nalizing symptoms themselves, and not necessarily any specific behavior associated with those symp- toms, predicted higher dyadic friendship quality (Hill & Swenson, 2014). We offer two possible explanations for the divergent findings. First, whereas other studies have examined perceived dyadic friendship quality, we examined friendship group size and structure. It is possible that even if their groups’ structures are weaker, adolescents with depressive symptoms can still experience sub- jectively high friendship quality in specific friend- ships within their friendship groups.

Second, many of the past studies that found benefits of depressive symptoms for friendships examined broad scales of internalizing symptoms (Hill & Swenson, 2014; Rose, 2002). We separated depressive from anxiety symptoms, and found that the two had distinct associations with group cohesiveness. First, anxiety symptoms were gener- ally less predictive of group cohesiveness than depressive symptoms. Second, whereas higher average depressive symptoms were negatively associated with cohesion, higher average anxiety symptoms were associated with greater reciprocity of group members’ friendships. This indicates that anxiety symptoms may not be as damaging to

TABLE 3 Negative Binomial and Linear Random Effects Regression Coefficients Predicting Adolescent Friendship Groups’ Cohesiveness From

Within-Group Variation in Depressive and Anxiety Symptoms (N = 3,013 Groups)

Predictor

Outcome

Group sizea Group density Group reciprocity Group transitivity b (SE) b (SE) b (SE) b (SE)

Variation in depressive symptoms .08 (.06) �.07 (.02)*** �.04 (.02) .01 (.02) Variation in anxiety symptoms .28 (.07)*** �.07 (.02)** .04 (.02) .06 (.03)*

Notes. All models included a variance component for school and controlled for grade, intervention condition, group gender compo- sition, and group members’ race/ethnicity, family structure, and free lunch eligibility. The reciprocity and transitivity models also controlled for density. aNegative binomial coefficients shown; linear coefficients shown for all other models. *p < .05, **p < .01, ***p < .001. Source: PROSPER Peers Study.

398 SIENNICK AND PICON

friendship group structure as are depressive symp- toms, and it is consistent with past work showing that anxiety may not harm, and may under some conditions actually enhance, adolescents’ friend- ships (Chen et al., 2009; de Matos et al., 2003; Rose et al., 2011).

Our findings are consistent with work showing that group emotions, or the “content” of groups— here, members’ internalizing symptoms—have implications for the “structure” of those groups (Rubin et al., 2015, p. 350). Theoretically, this is because group affect can undermine group pro- cesses such as communication and cooperation, thus interfering with group bonding (Kelly & Jones, 2012; Spoor & Kelly, 2004). The logical next ques- tions for research are (1) do these group processes in fact mediate the association of depressive symp- toms with group structure, and (2) what are the implications of these processes and structure for key functions of peer groups, such as peer support and peer influence? For example, friendship groups characterized by weaker internal ties may be less effective at protecting and defending their members against victimization by other peers, which depressed youth are more likely to experience (Kaltiala-Heino et al., 2010; Kochel et al., 2012). Future studies should examine whether friendship groups composed of more depressed adolescents indeed appear worse on dimensions such as com- munication, cooperation, conflict, and supportive- ness. Future research also should test whether the weaker group bonds among adolescents with depressive symptoms paradoxically have protective effects on other forms of problem behavior by undermining the group processes that facilitate that behavior. For instance, poor within-group commu- nication could hamper the within-group spread of attitudes favorable toward substance use. Although depressive symptoms in adolescents are associated with such attitudes (Siennick, Widdowson, Woess- ner, Feinberg, & Spoth, 2017), perhaps the associa- tion would be stronger if these adolescents were embedded in tight-knit groups.

We found only partial support for the idea that friendship groups with greater variability in inter- nalizing symptoms would be less internally cohe- sive. Variability in depressive symptoms was negatively associated with group density, but not with group size, reciprocity, or transitivity. Vari- ability in anxiety symptoms did not consistently predict group cohesiveness in one direction or the other. The support for this hypothesis was weaker than we would expect given the general finding that adolescents tend to prefer friends with similar

levels of depression (Kiuru et al., 2012; Van Zalk et al., 2010) and social anxiety (Van Zalk, Van Zalk, Kerr, et al., 2011). It also is weaker than we would expect given past work showing that people prefer others with similar moods and affect (Bl€ote et al., 2012; Chow et al., 2015; Hafen et al., 2011). Perhaps at the friendship group level, dissimilarity in inter- nalizing symptoms does not consistently make group interactions less validating or interfere with communication, cooperation, and other group bonding processes, even if global depressive symp- toms do.

Our study had limitations. First, there is the pos- sibility of reverse causality, such that qualities of friendship groups predict group internalizing symptoms rather than the other way around. We believe that this concern is mitigated by our mea- surement of symptoms of depression and anxiety during the 6 months prior to the survey and of qualities of friendship groups at the time of the survey. In addition, our proposed direction of causality is consistent with recent longitudinal find- ings that depression influences friendship forma- tion (Schaefer et al., 2011). Still, future studies should examine the impact of friendship group tight-knittedness on internalizing symptoms, and test for possible reciprocal associations. Second, our sample was from rural and predominately White school districts with large proportions of low- income families. The advantage of this sample is that the communities included in the study were served by one public high school each, increasing the chances that the social networks we mapped captured adolescents’ entire friendship pools. Still, future research should examine whether our results generalize to different populations.

Another limitation of this study is that the items we used to create our measures were only subsets of the items that could possibly have been included and thus our measures did not capture the full spectrum of mood and anxiety symptoms. More- over, we were unable to examine potentially rele- vant problems such as social anxiety, which might affect friendship group structure differently than general anxiety. Indeed, one study found that youth with social anxiety disorders had fewer close friends, greater difficulty making friends, and poorer quality peer interactions, compared to their nonanxious counterparts (Alfano, Beidel, & Wong, 2011). In contrast, youth with general anxiety disor- ders appeared more similar than not to the control group, with the exception that anxious youth had fewer friends overall than nonanxious youth (Alfano et al., 2011). This suggests that social

DEPRESSION AND FRIENDSHIP GROUPS 399

anxiety might be more damaging than general anx- iety, and thus groups comprised of socially anxious youth might potentially be less cohesive than groups formed by youth with general anxiety.

Our study is also limited by the use of self-report measures of depressive and anxiety symptoms. Self- report measures are prone to error due to over or underreporting of behaviors, which could potentially bias the results. Future studies should consider using parent or teacher ratings to determine whether our findings differ based on how adolescents’ depressive or anxiety symptoms are measured. Also, because our measures were not based on clinical diagnoses of depression or anxiety, our results do not necessarily apply to adolescents with clinical levels of depression or anxiety. Lastly, the friendship groups we created were mutually exclusive, and it is possible that some adolescents may identify with multiple groups. These “bridges” or “liaisons” are believed to play a critical role in the diffusion of behaviors and informa- tion between groups (Ennett & Bauman, 1996). Thus, the groups they bridge could grow more similar to each other in terms of depressive or anxiety symp- toms. Moreover, it is possible that these adolescents were less integrated into the groups to which they were assigned. To the extent that they have different levels of anxiety and depressive symptoms than do core group members, this could have influenced our results. Future research should examine whether such friendship group liaisons do indeed differ in levels of symptoms compared to core group mem- bers, and whether their social position influences the cohesiveness of the groups with which they identify.

Friendship groups are important because they provide adolescents with sources of belongingness and identity, and serve as channels for the spread of information and peer influence (Brechwald & Prin- stein, 2011; Ellis & Zarbatany, 2007; Simons-Morton & Farhat, 2010). Our findings suggest that even when adolescents with depressive symptoms, who often are already socially marginalized, manage to form friendship groups, those groups are not exact substitutes for “typical” adolescent friendship groups in terms of cohesiveness. This study extends to the group level past work on the harmful effects of depressive symptoms on peer relations, and in doing so highlights the relevance of group structure for our understanding of the social network statuses of adolescents with depression.

REFERENCES

Agoston, A. M., & Rudolph, K. D. (2013). Pathways from depressive symptoms to low social status. Journal of

Abnormal Child Psychology, 41(2), 295–308. https://doi. org/10.1007/s10802-012-9675-y

Alfano, C., Beidel, D., & Wong, N. (2011). Children with generalized anxiety disorder do not have peer prob- lems, just fewer friends. Child Psychiatry and Human Development, 42(6), 712–723. https://doi.org/doi:10. 1007/s10578-011-0245-2

Altmann, E. O., & Gotlib, I. H. (1988). The social behav- ior of depressed children: An observational study. Journal of Abnormal Child Psychology, 16(1), 29–44. https://doi.org/10.1007/BF00910498

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., text rev.). Washington, DC: Author.

Baker, J. R., Hudson, J. L., & Taylor, A. (2014). An inves- tigation into the lower peer liking of anxious than nonanxious children. Journal of Anxiety Disorders, 28(6), 599–611. https://doi.org/10.1016/j.janxdis.2014.06.007

Berger, C. R., & Calabrese, R. J. (1975). Some explorations in initial interactions and beyond: Toward a develop- mental theory of interpersonal communication. Human Communication Research, 1(2), 99–112. https://doi.org/ 10.1111/j.1468-2958.1975.tb00258.x

Bl€ote, A. W., Bokhorst, C. L., Miers, A. C., &Westenberg, P. M. (2012). Why are socially anxious adolescents rejected by peers? The role of subject-group similarity character- istics. Journal of Research on Adolescence, 22(1), 123–134. https://doi.org/10.1111/j.1532-7795.2011.00768.x

Brechwald, W. A., & Prinstein, M. J. (2011). Beyond homophily: A decade of advances in understanding peer influence processes. Journal of Research on Adoles- cence, 21(1), 166–179. https://doi.org/10.1111/j.1532- 7795.2010.00721.x

Buhrmester, D. (1990). Intimacy of friendship, interper- sonal competence, and adjustment during preadoles- cence and adolescence. Child Development, 61(4), 1101–1111. https://doi.org/10.1111/j.1467-8624.1990.tb 02844.x

Chan, A., & Poulin, F. (2009). Monthly instability in early adolescent friendship networks and depressive symp- toms. Social Development, 18(1), 1–23. https://doi.org/ 10.1111/j.1467-9507.2008.00461.x

Cheadle, J. E., & Goosby, B. J. (2012). The small-school friendship dynamics of adolescent depressive symp- toms. Society and Mental Health, 2(2), 99–119. https://d oi.org/10.1177/2156869312445211

Chen, H., Cohen, P., Johnson, J. G., & Kasen, S. (2009). Psychiatric disorders during adolescence and relation- ships with peers from age 17 to age 27. Social Psychia- try and Psychiatric Epidemiology, 44(3), 223–230. https://doi.org/10.1007/s00127-008-0421-2

Chow, C. M., Tan, C. C., & Ruhl, H. (2015). Misery loves company: A dyadic approach to examining the effects of depressive symptoms on friendship discord. Journal of Social and Clinical Psychology, 34(9), 774–787. https://doi.org/10.1521/jscp.2015.34.9.774

Coyne, J. C. (1976). Toward an interactional description of depression. Psychiatry: Interpersonal and Biological

400 SIENNICK AND PICON

Processes, 39(1), 28–40. https://doi.org/10.1080/ 00332747.1976.11023874

Cummings, C. M., Caporino, N. E., & Kendall, P. C. (2014). Comorbidity of anxiety and depression in chil- dren and adolescents: 20 years after. Psychological Bul- letin, 140(3), 816–845. https://doi.org/10.1037/ a0034733

de Matos, M. G., Barrett, P., Dadds, M., & Shortt, A. (2003). Anxiety, depression, and peer relationships during adolescence: Results from the Portuguese national health behaviour in school-aged children sur- vey. European Journal of Psychology of Education, 18(1), 3–14. https://doi.org/10.1007/BF03173600

De�girmencio�glu, S. M., Urberg, K. A., Tolson, J. M., & Richard, P. (1998). Adolescent friendship networks: Continuity and change over the school year. Merrill- Palmer Quarterly, 44(3), 313–337. https://doi.org/10. 4319/lo.2013.58.2.0489

Ellis, W. E., & Zarbatany, L. (2007). Peer group status as a moderator of group influence on children’s deviant, aggressive, and prosocial behavior. Child Development, 78(4), 1240–1254. https://doi.org/10.1111/j.1467-8624. 2007.01063.x

Ennett, S. T., & Bauman, K. E. (1996). Adolescent social networks: School, demographic, and longitudinal con- siderations. Journal of Adolescent Research, 11(2), 194– 215. https://doi.org/10.1177/0743554896112003

Gest, S. D., Davidson, A. L., Rulison, K. L., Moody, J., & Welsh, J. A. (2007). Features of groups and status hier- archies in girls’ and boys’ early adolescent peer net- works. New Directions for Child and Adolescent Development, 118, 42–60. https://doi.org/10.1002/cd

Giletta, M., Scholte, R. H. J., Burk, W. J., Engels, R. C. M. E., Larsen, J. K., Prinstein, M. J., & Ciairano, S. (2011). Similarity in depressive symptoms in adolescents’ friendship dyads: Selection or socialization? Develop- mental Psychology, 47(6), 1804–1814. https://doi.org/10. 1037/a0023872

Hafen, C. A., Laursen, B., Burk, W. J., Kerr, M., & Stattin, H. (2011). Homophily in stable and unstable adoles- cent friendships: Similarity breeds constancy. Personal- ity and Individual Differences, 51, 607–612. https://doi. org/10.1016/j.paid.2011.05.027

Hill, E. N., & Swenson, L. P. (2014). Perceptions of friendship among youth with distressed friends. Child Psychiatry and Human Development, 45(1), 99–109. https://doi.org/10.1007/s10578-013-0381-y

Hogue, A., & Steinberg, L. (1995). Homophily of internal- ized distress in adolescent peer groups. Developmental Psychology, 31, 897–906. https://doi.org/10.1037/0012- 1649.31.6.897

Kaltiala-Heino, R., Fr€ojd, S., & Marttunen, M. (2010). Involvement in bullying and depression in a 2-year follow-up in middle adolescence. European Child and Adolescent Psychiatry, 19(1), 45–55. https://doi.org/10. 1007/s00787-009-0039-2

Kelly, J. R., & Jones, E. (2012). Investigating emotions and affect in groups. In A. B. Hollingshead & M. S.

Poole (Eds.), Research methods for studying groups and teams: A guide to approaches, tools, and technologies (pp. 154–172). New York: Taylor & Francis/Routledge.

Kennedy, E., Spence, S. H., & Hensley, R. (1989). An examination of the relationship between childhood depression and social competence amongst primary school children. Journal of Child Psychology and Psychia- try, and Allied Disciplines, 30(4), 561–573. https://doi. org/10.1111/j.1469-7610.1989.tb00268.x

Kiuru, N., Burk, W. J., Laursen, B., Nurmi, J. E., & Sal- mela-Aro, K. (2012). Is depression contagious? A test of alternative peer socialization mechanisms of depres- sive symptoms in adolescent peer networks. Journal of Adolescent Health, 50(3), 250–255. https://doi.org/10. 1016/j.jadohealth.2011.06.013

Kochel, K. P., Ladd, G. W., & Rudolph, K. D. (2012). Lon- gitudinal associations among youth depressive symp- toms, peer victimization, and low peer acceptance: An interpersonal process perspective. Child Development, 83(2), 637–650. https://doi.org/10.1111/j.1467-8624. 2011.01722.x

Kreager, D. A., Rulison, K., & Moody, J. (2011). Delin- quency and the structure of adolescent peer groups. Criminology, 49(1), 95–127. https://doi.org/10.1111/j. 1745-9125.2010.00219.x

Mcleod, J. D., & Uemura, R. (2012). Substance use, dis- tress, and adolescent school networks. Journal of Research on Adolescence, 22(3), 438–452. https://doi. org/10.1111/j.1532-7795.2012.00787.x

Mercer, S. H., & Derosier, M. E. (2010). Selection and socialization of internalizing problems in middle child- hood. Journal of Social and Clinical Psychology, 29(9), 1031–1056. https://doi.org/10.1521/jscp.2010.29.9.1031

Moody, J. (2001). Peer influence groups: Identifying dense clus- ters in large networks. Social Networks, 23(4), 261–283. https://doi.org/10.1016/S0378-8733(01)00042-9

Prinstein, M. J. (2007). Moderators of peer contagion: A longitudinal examination of depression socialization between adolescents and their best friends. Journal of Clinical Child and Adolescent Psychology, 36(2), 159–170. https://doi.org/10.1080/15374410701274934

Prinstein, M. J., Borelli, J. L., Cheah, C. S. L., Simon, V. A., & Aikins, J. W. (2005). Adolescent girls’ interper- sonal vulnerability to depressive symptoms: A longitu- dinal examination of reassurance-seeking and peer relationships. Journal of Abnormal Psychology, 114(4), 676–688. https://doi.org/10.1037/0021-843X.114.4.676

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: Sage.

Rook, K. S., Pietromonaco, P. R., & Lewis, M. A. (1994). When are dysphoric individuals distressing to others and vice versa? Effects of friendship, similarity, and interaction task. Journal of Personality and Social Psychology, 67(3), 548–559. https://doi.org/10.1037/0022-3514.67.3.548

Rose, A. J. (2002). Co-rumination in the friendships of girls and boys. Child Development, 73(6), 1830–1843. https://doi.org/10.1111/1467-8624.00509

DEPRESSION AND FRIENDSHIP GROUPS 401

Rose, A. J., Carlson, W., & Waller, E. M. (2007). Prospec- tive associations of co-rumination with friendship and emotional adjustment: Considering the socioemotional trade-offs of co-rumination. Developmental Psychology, 43(4), 1019. https://doi.org/10.1037/0012-1649.43.4. 1019

Rose, A. J., Carlson, W., Luebbe, A. M., Schwartz-Mette, R. A., Smith, R. R., & Swenson, L. P. (2011). Predicting difficulties in youth’s friendships: Are anxiety symp- toms as damaging as depressive symptoms? Merrill- Palmer Quarterly, 57(3), 244–262. https://doi.org/10. 1353/mpq.2011.0013

Rosenblatt, A., & Greenberg, J. (1991). Examining the world of the depressed: Do depressed people prefer others who are depressed? Journal of Personality and Social Psychology, 60(4), 620–629. https://doi.org/10. 1037/0022-3514.60.4.620

Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. New York, NY: John Wiley & Sons. https://d oi.org/10.1002/SERIES1345

Rubin, K. H., Bukowski, W. M., & Bowker, J. C. (2015). Children in peer groups. In M. H. Bornstein, T. Leventhal & R. M. Lerner (Eds.), Handbook of child psy- chology and developmental science, volume 4, ecological set- tings and processes (Vol. 4(October), pp. 175–222). Hoboken, NJ: John Wiley & Sons. https://doi.org/10. 1002/9781118963418.childpsy405

Rudolph, K. D., Hammen, C., & Burge, D. (1994). Inter- personal functioning and depressive symptoms in childhood: Addressing the issues of specificity and comorbidity. Journal of Abnormal Child Psychology, 22 (3), 355–371. https://doi.org/10.1007/BF02168079

Schaefer, D. R., Kornienko, O., & Fox, A. M. (2011). Mis- ery does not love company: Network selection mecha- nisms and depression homophily. American Sociological Review, 76(5), 764–785. https://doi.org/10.1177/ 00031224 11420813

Schwartz-Mette, R. A., & Rose, A. J. (2016). Depressive symptoms and conversational self-focus in adolescents’ friendships. Journal of Abnormal Child Psychology, 44(1), 87–100. https://doi.org/10.1007/s10802-015-9980-3

Siennick, S. E., Widdowson, A. O., Woessner, M. K., Feinberg, M. E., & Spoth, R. L. (2017). Risk factors for substance misuse and adolescents’ symptoms of depression. Journal of Adolescent Health, 60(1), 50–56. https://doi.org/10.1016/j.jadohealth.2016.08.010

Simons-Morton, B. G., & Farhat, T. (2010). Recent find- ings on peer group influences on adolescent smoking. Journal of Primary Prevention, 31(4), 191–208. https://d oi.org/10.1007/s10935-010-0220-x

Spoor, J. R., & Kelly, J. R. (2004). The evolutionary signif- icance of affect in groups: Communication and group bonding. Group Processes & Intergroup Relations, 7(4), 398–412. https://doi.org/10.1177/1368430204046145

Spoth, R., Greenberg, M., Bierman, K., & Redmond, C. (2004). PROSPER community-university partnerships model for public education systems: Capacity-building for evidence-based, competence-building prevention. Prevention Science, 5(1), 31–39. https://doi.org/10. 1023/B:PREV.0000013979.52796.8b

Spoth, R., Redmond, C., Shin, C., Greenberg, M., Fein- berg, M., & Schainker, L. (2013). PROSPER commu- nity-university partnership delivery system effects on substance misuse through 6 1/2 years past baseline from a cluster randomized controlled intervention trial. Preventive Medicine, 56(3–4), 190–196. https://doi. org/10.1016/j.ypmed.2012.12.013

Stevens, E. A., & Prinstein, M. J. (2005). Peer contagion of depressogenic attributional styles among adolescents: A longitudinal study. Journal of Abnormal Child Psychol- ogy, 33(1), 25–37. https://doi.org/10.1007/s10802-005- 0931-2

Stice, E., Ragan, J., & Randall, P. (2004). Prospective rela- tions between social support and depression: Differen- tial direction of effects for parent and peer support? Journal of Abnormal Psychology, 113(1), 155–159. https://doi.org/10.1037/0021-843X.113.1.155

Strauss, C. C., Lahey, B. B., Frick, P., Frame, C. L., & Hynd, G. W. (1988). Peer social status of children with anxiety disorders. Journal of Consulting and Clinical Psy- chology, 56(1), 137–141. https://doi.org/10.1037/0022- 006X.56.1.137

Van Zalk, M. H. W., Kerr, M., Branje, S. J. T., Stattin, H., & Meeus, W. H. J. (2010). It takes three: Selection, influence, and de-selection processes of depression in adolescent friendship networks. Developmental Psychol- ogy, 46(4), 927–938. https://doi.org/10.1037/a0019661

Van Zalk, N., Van Zalk, M. H. W., & Kerr, M. (2011). Socialization of social anxiety in adolescent crowds. Journal of Abnormal Child Psychology, 39(8), 1239–1249. https://doi.org/10.1007/s10802-011-9533-3

Van Zalk, N., Van Zalk, M., Kerr, M., & Stattin, H. (2011). Social anxiety as a basis for friendship selection and socialization in adolescents’ social networks. Jour- nal of Personality, 79(3), 499–526. https://doi.org/10. 1111/j.1467-6494.2011.00682.x

Wright, J. C., Giammarino, M., & Parad, H. W. (1986). Social status in small groups: Individual group similar- ity and the social “misfit”. Journal of Personality and Social Psychology, 50(3), 523–536. https://doi.org/10. 1037/0022-3514.50.3.523

402 SIENNICK AND PICON