Abstract
In the current study, associations between individual and friends’ direct and indirect aggression and depressive problems were examined. It was expected that social status would moderate these associations such that low-status preadolescents would be more similar to their unilateral friends with regard to indirect, but not direct, aggression. Furthermore, it was expected that preadolescents’ depressive problems were positively associated to reciprocal friends’ depressive problems, in particular in low-status preadolescents. The hypotheses were tested by studying unilateral and reciprocal friendships in 204 children (Mage=10.90; SD=0.78; 44.1% girls). Direct and indirect aggression and depressive problems were assessed via both self- and peer-reports. Social status was assessed via peer-reported rejection (i.e., dislike) and popularity (i.e., who do others want to be associated with). Analyses showed partial support for the hypotheses, showing that individual and unilateral friends’ self-reported indirect and direct aggression were positively associated in preadolescents who were lower on popularity. Moreover, in preadolescents who were more rejected by peers, depressive problems were positively associated to reciprocal friends’ depressive problems. The current study highlights the importance of including social status and distinguishing between unilateral and reciprocal friendships when examining associations between individual and friends’ behavior.
Introduction
Friendships are important for child and adolescent development and can provide behavioral confirmation, status, and affection (Brendgen, Bowen, Rondeau, & Vitaro, 1999; Hartup, 1996). However, friends may also impair positive emotional and behavioral development. That is, several studies have shown that youth are often in friendships with similarly aggressive or depressive children (Cairns, Cairns, Nechermann, Gest & Gariepy, 1998; Poulin et al., 1997; Stevens & Prinstein, 2005), often select friends similar on behaviors such as aggression and delinquency (Snyder, Horsch & Childs, 1997; Werner & Crick, 2004) and are also influenced by their friends
The wealth of research on youth and their peers has led to at least two important, but understudied, insights. First, the association between individual behavior and friends’ behavior may differ between reciprocal and unilateral friends (see e.g. Scholte et al., 2009; Sijtsema, Lindenberg, & Veenstra, 2010). Reciprocal friends include friendships that are mutual and where both actors indicate the presence of a friendship. In contrast, unilateral friends include friendships that are unreciprocated and where only one actor indicates the presence of a friendship. In previous research, unilateral friendships have often been labeled desired (or choosing) friendships (Scholte et al., 2009; Sijtsema, Lindenberg, et al., 2010), as they may seem to indicate preferred friendships. In the current study, unilateral friendships are also regarded as desired friendships, although it should be noted that they may also indicate affiliations on the border between acquaintance and being friends. Second, social status plays an important role in explaining individual differences in problem behavior in childhood and adolescence (Cillessen & Mayeux, 2004; Dishion & Tipsord, 2011; Newcomb, Bukowski, & Pattee, 1993) and may shed more light on the associations between individual behavior and friends’ behavior. That is, children low on social status (e.g., rejected by peers) may be more inclined to behave similar to their antisocial peers to try to fit in (Dijkstra, Cillessen, Lindenberg & Veenstra, 2010; Dishion & Tipsord, 2011) or, in the case of depressive problems, seek comfort with peers who have had similar negative experiences (see for an overview: Furman & Rose, 2014).
The current study aims to combine these two insights by examining whether the association between individual and friends
Theories of Behavioral Similarity
There are several theories for why peers are similar with respect to problem behavior. Typically, this similarity is thought to be the result of either homophilic selection or socialization (Byrne, 1971; McPherson, Smith-Lovin & Cook, 2001). On the one hand, peers select each other as friends because they are similar in interests, opinions, attitudes and related behaviors and hence are more inclined to meet and become friends (i.e., friendship selection). In contrast to this traditional view of friendship selection, selection is sometimes the result of unwanted processes referred to as default selection. This suggests that some youth are not able to establish the friendships they prefer and settle for friendships with less preferred peers. Such processes seem likely in the context of aggression and depressive problems, and there is some theoretical and empirical support for this notion (Deptula & Cohen, 2004; Hektner, August, & Realmuto, 2000; Sijtsema, Lindenberg, et al., 2010). On the other hand, peers become more similar to each other through repeated interaction and may imitate each other’s behaviors, attitudes, and opinions (i.e., friendship socialization). There is considerable evidence in support of both views with regard to aggression and depressive problems (see for an overview: Dishion & Tipsord, 2011; Furman & Rose, 2014).
However, it has been suggested that behavioral selection and influence processes are often a byproduct of status attainment. For one, there is evidence showing that most variance in friendship selection with regard to aggression is explained by selection on popularity instead of peers’ aggressive behavior (Dijkstra, Berger, & Lindenberg, 2011). Adolescents thus want to befriend popular peers, and similarity in antisocial behavior seems to be a by-product of this selection. The salience of social status for establishing and maintaining friendships is particularly apparent in the early stages of adolescence, including preadolescence (Hawley, 1999). During this period, processes of physical maturing, increasing opportunities for mating, and increasing encounters of competitive contexts are likely to increase the salience of status striving (e.g., De Bruyn, Cillessen, & Weisfeld 2012). Thus, youth want to belong to the in-crowd and, once accepted, to rise in status within the in-crowd, and do whatever is necessary to achieve this. In adolescence, group norms increasingly favor antisocial behavior and these behaviors are likely to become associated with higher social status (Moffitt, 1993; Dijkstra, Lindenberg, & Veenstra, 2008). Moreover, previous research has shown that popular youth often display antisocial behaviors to gain and keep a high peer status (Cillessen & Mayeux, 2004; Rodkin & Berger, 2008; Ojanen & Findley-Van Nostrand, 2014).
Indirect and Direct Aggression
Yet, different forms of antisocial behavior are differently related to behavioral similarity and social status. Specifically, for aggression there is evidence that peers become more similar in indirect forms (e.g., gossiping, ignoring others) and instrumental forms (e.g., bullying) over time (Sijtsema, Ojanen, et al., 2010; Sijtsema, Rambaran, Caravita, & Gini, 2014). Research on deviancy training has shown that youth foster positive attitudes toward deviant behavior and hence provide reinforcement of such behaviors (e.g., Dishion & Tipsord, 2011). Through such processes, antisocial youth may receive status or behavioral confirmation in the peer group. This may also explain why indirect aggression is increasingly associated with higher social status in adolescence (Cillessen & Mayeux, 2004; Heilbron & Prinstein, 2008). As such, low-status children may want to behave more similar to their peers in terms of indirect aggression to fit in. However, low-status youth are likely to be unpopular as friends (e.g., Newcomb et al., 1993) and hence may have a higher chance of ending up in unilateral friendships with indirectly aggressive peers, because the friendship nomination is not reciprocated by their peers.
In contrast, direct forms aggression are more likely to be associated with lower social status in pre- and early adolescence (e.g., Cillessen & Mayeux, 2004). Moreover, previous studies indicate that there are several disadvantages of being friends with directly aggressive peers, which makes it less likely that youth select or maintain friendships with directly aggressive peers. That is, directly aggressive peers easily respond with aggression to their friends (Leary & Katz, 2005) and their friendships dissolve quickly (Dishion, Andrews, & Crosby, 1995). In addition, friendships with directly aggressive peers are characterized by conflict and lower levels of intimacy compared to friendships with nonaggressive peers (e.g., Cillessen, Jiang, West, & Laszkowski, 2005; Grotpeter & Crick, 1996). The behavioral similarity observed in directly aggressive youth may thus be attributable to processes unrelated to status motives and more in line with processes of default selection. In their review, Deptula and Cohen (2004) suggested that default selection processes are likely to occur in rejected children as they have a marginal status in the peer group and hence have a more limited pool of available peers to befriend and thus end up with other rejected peers. A similar process may be observed for directly aggressive youth: aggressive youth low on social status may have to settle for second best friendships with directly aggressive peers, to be able to receive at least some affection from their peers (cf. Sijtsema, Lindenberg, et al., 2010).
Default selection may thus explain why directly aggressive preadolescents with low social status have reciprocal friendships with directly aggressive peers, but have unilateral friendships with peers dissimilar in direct aggression. Previous research indeed supports this notion, showing that there is some evidence that young adolescents characterized by direct aggression, have fewer opportunities to establish reciprocal friendships with nonaggressive, prosocial peers, and in fact end up with other directly aggressive peers (Aboud & Mendelson, 1996; Hektner et al., 2000; Sijtsema, Lindenberg, et al., 2010).
Depressive Problems
The processes that explain similarity in direct aggression may in part also apply to friendship similarity in depressive problems (see Kiuru, Burk, Laursen, Nurmi, & Salmela-Aro, 2012). That is, youth with depressive problems and who are low on social status may have fewer friends to choose from and hence are limited to befriending peers with depressive problems in order to receive some affection. Research focusing on friendship development of adolescents with depressive problems found some support for this, showing that these friendships are more likely to be broken off over time (Van Zalk, Kerr, Branje, Stattin, & Meeus, 2010), hence limiting the pool of remaining friends. A similar process was also observed with regard to the friendships of victimized children (Sijtsema, Rambaran, & Ojanen, 2013). Furthermore, previous studies have shown that friendships with depressive peers are often characterized by conflict and place individuals at risk for developing depressive problems (Nangle, Erdley, Newman, Mason, & Carpenter, 2003; Rubin, Wojslawowicz, Rose-Krasnor, Booth-LaForce, & Burgess, 2006). Hence, the assumption that children actively select depressive friends on the basis of behavioral similarity may in some cases be unrealistic, in particular for those with low social status.
Alternatively, processes of co-rumination may also be at play. In this regard, youth are likely to be similar because they select peers on the basis of depressive problems and influence each other’s depressive problems through rehashing negative events and feelings (Furman & Rose, 2014; Rose, Carlson, & Waller, 2007). Despite the finding that high-quality friendships can buffer for internalizing problems, in pre- and early adolescence co-rumination processes may exacerbate depressive symptoms (Rose et al., 2007). Reciprocal friendships, more than unilateral friendships, may thus be characterized by youth similar in depressive symptoms. In sum, default selection and co-rumination processes both suggest that youth are similar to their reciprocal friends in depressive problems. Moreover, from a default selection perspective, this association may depend upon individual social status.
The Present Study
In sum, it is thus examined to what extent preadolescents’ aggression and depressive problems are associated with friends’ aggression and depressive problems. From a behavioral similarity perspective, individual behavior is expected to be associated with friends’ behavior, but this association is contingent upon friendship reciprocity and individual social status.
First, indirect aggression in preadolescence is positively associated with social status (Cillessen & Mayeux, 2004). Hence, indirect aggression is more likely to be adopted from friends and can be a focal selection criterion for selecting friends. It is thus hypothesized that the preadolescent indirect aggression is positively associated with reciprocal friends’ indirect aggression and that this association is stronger in individuals who are less rejected or higher on popularity, in those who show more indirect aggression (hypothesis 1a). In contrast, low-status preadolescents may be more inclined to fit in and hence imitate potentially rewarding peer behavior, in line with a need for status perspective. As such, in low-status preadolescents, the positive association between indirect aggression and unilateral, or desired, friends’ indirect aggression may be stronger. It is thus hypothesized that the positive association between individual and unilateral friends’ indirect aggression is stronger in youth higher on peer rejection or lower on popularity, in those who show more indirect aggression (hypothesis 1b).
Second, with regard to direct aggression, it is hypothesized that individual social status moderates the association between friends’ aggression and individual aggression. Based on default selection processes, it is hypothesized that the positive association between individual and reciprocal friends’ direct aggression is stronger in individuals who are more rejected or lower on popularity, in those who show more direct aggression (hypothesis 2a). That is, aggressive preadolescents with a low social status may have fewer friendship alternatives and as a result are more likely to have reciprocal friendships with directly aggressive peers. With regard to unilateral friendships, it is hypothesized that the positive association between individual and unilateral friends’ direct aggression is weaker in youth higher on peer rejection or lower on popularity, in those who show more direct aggression (hypothesis 2b).
Third, two sets of hypotheses with respect to depressive problems were formulated. From a co-rumination perspective, it is hypothesized that preadolescent depressive problems are positively associated with reciprocal friends’ depressive problems (hypothesis 3a). That is, these youth may seek more comfort with their close friends who have had similar negative peer experiences. Moreover, the default selection perspective further suggests that this positive association is stronger for individuals low on social status (hypothesis 3b). Furthermore, for unilateral friends, it is explored whether preadolescents’ depressive problems are positively associated with unilateral friends’ depressive problems (hypothesis 4). It will also be explored whether this association depends on social status.
All hypotheses were tested for self- and peer-reported behaviors separately. Previous studies have shown only small to moderate associations between self- and peer-reported behaviors (Noordhof, Oldehinkel, Verhulst, & Ormel, 2008; Sijtsema, Verboom, Penninx, Verhulst, & Ormel, 2014), but there are no reasons to assume that the hypothesized processes would differ depending upon reporter. In addition, social status was assessed by popularity (i.e., who do others want to be associated with; cf. Dijkstra, Lindenberg, Verhulst, Ormel, & Veenstra, 2009) and peer rejection (i.e., dislike).
Method
Participants and Procedure
Data were collected at five elementary schools in the south of the Netherlands. At each school, fifth and sixth graders were included in the study. In total, 231 children were invited to participate of which one participant did not fill out its sex. Two weeks before data collection, participants and their parents were informed about the study and were asked to provide consent. In five cases, parents did not consent to the participation of their child. During data collection an additional five children were absent and hence could not participate in the study. Moreover, in one classroom peer nominations were not reliably assessed, thereby excluding 16 participants. Independent sample t-tests showed that these participants did not differ significantly on sex, age, self-reported indirect and direct aggression and depressive problems from participants with peer nomination information. The final sample thus consisted of 204 (Mage = 10.96, SD = 0.75; range 10–13 years) preadolescents (44.1% girls; 88.7% Dutch nationality) distributed over eight classrooms ranging from 20 to 31 in size. In all classrooms there was a participation rate of 90% or more, which is well above the recommended rate of 60% that is required for peer nominations (Cillessen, 2009).
After schools, parents, and children provided informed consent, children filled out a paper and pencil questionnaire. Children filled out the questionnaires and peer nominations in their classroom under the supervision of a graduate student. During the test administration, there was sufficient space between each of the children. Furthermore, it was explained to the children that the data would be stored anonymously and that parents and teachers would not be informed about their answers. In addition to self-reports and peer nominations, teachers also filled out a short behavioral questionnaire about each child. The study was conducted in accordance with the ethical guidelines as provided by the American Psychological Association (APA, 2010).
Self-reported Measures
Self-reported aggression
To assess direct and indirect aggression, participants filled out the Direct and Indirect Aggression Scales (DIAS) (Bjorkqvist, Lagerspetz, & Kaukiainen, 1992). This questionnaire consists of 17 items asked under the framework ‘when I am angry or mad at someone, I …’ and each item was rated on a 4-point scale, ranging from ‘never’ (1) to ‘very often’ (4). Direct aggression was assessed with 6 items, such as ‘curse at the other’ and ‘take things from the other’, and showed good internal consistency (α = 0.84). Indirect aggression was assessed with 11 items, such as ‘try to make the other jealous’ and ‘gossip about the other’, and showed acceptable internal consistency (α = 0.71). Scores were averaged across the items to construct mean levels of direct and indirect aggression per participant.
Self-reported depressive problems
To assess depressive problems, participants filled out the Center for Epidemiologic Studies Depression Scale (CES-D) (Radloff, 1977). This questionnaire consisted of 12 items and related to how often they felt down or depressed during the last week. Answers were rated on a 4-point scale, ranging from ‘seldom/ never (less than 1 day)’ (1) to ‘often or always (5 to 7 days)’ (4). Items included, amongst others, ‘I slept badly’ and ‘I had the feeling everyone hated me’. The scale showed good internal consistency (α = 0.82). Scores were averaged across the items to construct mean levels of depressive problems per participant.
Friends’ self-reported behavior
Friendship was assessed via peer nominations. Participants were provided with a roster including the names of their classmates and were asked ‘who are your best friends?’ Participants could provide unlimited same- and cross-sex nominations. Based on this information, unilateral (i.e., only an outgoing nomination) and reciprocal (i.e., in- and outgoing nomination) friendships were determined. Next, for each participant, friends’ self-reported behavior was averaged over unilateral and reciprocal friendships separately. This was done for self-reported direct aggression, indirect aggression, and depressive problems and thus yielded three measures of friends’ self-reported behavior.
Peer-reported Measures
Peer-reported aggression
Direct and indirect aggression were also assessed via peer nominations. Participants were provided with a roster including the names of their classmates and were asked ‘who fights or quarrels often?’ and ‘who gossips or excludes others often?’ to assess direct and indirect aggression, respectively (see e.g. Dijkstra et al., 2009, 2010; Sijtsema, Lindenberg, et al., 2010). Participants could provide unlimited same- and cross-sex nominations. To calculate scores per participant, incoming nominations were standardized per classroom to a mean of 0 and a standard deviation of 1 to account for different classroom sizes.
Peer-reported depressive problems
Depressive problems were also assessed via peer nominations. Participants were provided with a roster including the names of their classmates and asked ‘who is sad or down often?’ (see Dijkstra et al., 2009). Participants could provide unlimited same- and cross-sex nominations. To calculate scores per participant, incoming nominations were standardized per classroom to a mean of 0 and a standard deviation of 1 to account for different classroom sizes.
Friends’ peer-reported behavior
Similar to the procedure for self-reported friends’ behavior, for each participant, friends’ behavior was averaged over unilateral and reciprocal friendships separately. This was done for peer-reported direct aggression, indirect aggression, and depressive problems and thus yielded three measures of friends’ peer-reported behavior.
Social status
To assess popularity, participants were provided with a roster including the names of their classmates and could nominate with whom others want to be associated (see Dijkstra et al., 2009; 2010). To assess peer rejection, a similar procedure was used, but participants were instead asked ‘who do you dislike?’ To calculate scores per participant, incoming nominations were standardized per classroom to a mean of 0 and a standard deviation of 1 to account for different classroom sizes.
Analysis Strategy
First, means, standard deviations, and ranges of all study variables were calculated. Second, Spearman rho correlations were calculated of social status with individual and friends’ behavior. Moreover, intra-class correlations (ICC) were calculated to assess the degree of congruence between self- and peer-reports of direct and indirect aggression and depressive problems. Third, to test the hypotheses, several multiple regression analyses were conducted for peer- and self-reported friends’ behavior separately. In the first step, friends’ behavior was predicted by sex, individual peer- and self-reported behavior, peer rejection, and popularity. In addition, three control variables were included to account for the dependencies in the data. Because averaging behavior across friends does not consider the differences between participants’ friends, the standard deviation of friends’ average behavior was included to accounted for the behavioral similar within friendship groups. Moreover, the number of friends differed per participant and hence a control was included for the number of unilateral and reciprocal best friends, respectively. Finally, participants varied in how often they were nominated as friends by their peers and could thus be included multiple times in the analyses. To account for this, the number of times participants were nominated as a best friend was also included in the analysis. Note that for reciprocal friendships, this is identical to the number of reciprocal best friends (i.e., the second control variable). In the second step, interactions of peer rejection and popularity with individual peer- and self-reported behavior were included.
In total, six analyses were performed for peer- and self-reported friends’ direct and indirect aggression and depressive problems, for reciprocal and unilateral friends separately. Significant interactions were plotted using simple slope analysis and it was tested whether the slopes differed significantly from zero (Aiken & West, 1991). To reduce problems with multicollinearity and to ensure that the values plotted in the figures are accurate representations of the data, independent variables were standardized to a mean of 0 and a standard deviation of 1 (Frazier, Tix, & Barron, 2004). All analyses were performed in IBM SPSS 19.0 and hypotheses were tested two-sidedly using a p-value of < 0.05 to indicate significance.
Results
Descriptive Analyses
Table 1 reports means, standard deviations, and ranges of all study variables. Correlations between the study variables are discussed in four parts (not reported). First, ICCs between self- and peer-reported behavior were 0.16, 95% CI = −0.11 to 0.36, for depressive problems, and 0.22, 95% CI = −0.03 to 0.41, and 0.42, 95% CI = 0.24 to 0.56, for indirect and direct aggression respectively. Spearman correlations between direct and indirect aggression were 0.22 for peer-reported behavior and 0.63 for self-reported behavior. Second, within reporter, individual behavior was positively correlated with friends’ with friends’ behavior (range rs: 0.16 to 0.41). There was one exception: individual peer-reported depressive problems were not significantly correlated with unilateral friends’ peer-reported depressive problems. Third, unilateral and reciprocal friends’ aggression and depressive problems were largely uncorrelated. Only unilateral and reciprocal friends’ peer-reported indirect aggression were positively correlated. Fourth, peer rejection was positively associated with individual and friends’ direct and indirect aggression. Correlations between popularity and direct and indirect aggression were mixed, although most correlations indicated that higher popularity was related to more individual and friends’ direct and indirect aggression. Depressive problems were negatively related with popularity, suggesting that lower popularity was associated with more individual and friends’ depressive problems.
Means, standard deviations, and range of self- and peer-reported aggression and depressive problems and peer rejection and perceived popularity.
Regression Analyses
Indirect aggression
To test hypothesis 1a that preadolescents with high social status are more similar to their reciprocal friends in indirect aggression, it was analyzed whether popularity and peer rejection moderated the association between individual indirect aggression and friends’ indirect aggression. This hypothesis was not supported, because interactions between social status and individual indirect aggression were not significant. Hence, only main effects are reported in Table 2.
Regression analyses of individual indirect aggression on friends
*p<0.05, **p<0.01, ***p<0.001.
Next, to test hypothesis 1b that the association between individual and unilateral friends’ indirect aggression is moderated by peer rejection and popularity, Table 2 reports the main and interaction effects with respect to unilateral friends. Popularity significantly moderated the association between individual self-reported indirect aggression and unilateral friends’ self-reported indirect aggression (see Figure 1A). Simple slope analysis showed that the slope for children high on popularity was not significantly different from zero, simple slope: b = −0.02, 95% CI = −0.06 to 0.03, p = 0.42. By contrast, for children low on popularity, there was a statistically significant association between individual indirect aggression and friends’ indirect aggression, simple slope: b = 0.04, 95% CI = 0.01 to 0.06, p < 0.01, such that children high on indirect aggression were more likely to have indirectly aggressive unilateral friends. No significant interactions were found between peer rejection and individual indirect aggression.
Association between self-reported indirect aggression and unilateral friends’ self-reported indirect aggression and at low (−1 SD) and high (+1 SD) levels of popularity (n = 154).
Direct aggression
To test hypothesis 2a that the association between individual and reciprocal friends’ direct aggression is stronger in individuals who are more rejected or lower on popularity, interactions were tested between social status and individual direct aggression. Table 3 shows that there was one significant interaction with regard to reciprocal friends. However, simple slopes for children high and low on rejection were not significantly different from zero, simple slopes respectively: b = 0.12, 95% CI = −0.06 to 0.29, p = 0.20 and b = −0.12, 95% CI = −0.34 to 0.10, p = 0.27. Hence, hypothesis 2a is not supported.
Regression analyses of individual direct aggression on friends
*p<0.05, **p<0.01, ***p<0.001.
To test hypothesis 2b that children with low social status are less similar to their unilateral friends in direct aggression, it was analyzed whether popularity and peer rejection moderated the association between individual direct aggression and friends’ direct aggression. Table 3 reports the main and interaction effects with respect to unilateral and reciprocal friends. Popularity significantly moderated the association between individual peer-reported direct aggression and unilateral friends’ peer-reported direct aggression. However, simple slopes for children high and low on popularity were not significantly different from zero, simple slopes respectively: b = −0.16, 95% CI = −0.41 to 0.10, p = 0.23 and b = 0.28, 95% CI = −0.04 to 0.60, p = 0.09. A similar interaction was found with regard to the association between individual self-reported direct aggression and unilateral friends’ self-reported direct aggression (see Figure 2A). Simple slope analysis showed that the slope for children high on popularity was not significantly different from zero, simple slope: b = 0.03, 95% CI = −0.02 to 0.08, p = 0.28. However, for children low on popularity, there was a statistically significant association between individual direct aggression and friends’ direct aggression, simple slope: b = 0.15, 95% CI = 0.08 to 0.22, p < 0.001, such that children with high levels of direct aggression were more likely to have unilateral friends who displayed direct aggression.
a: Association between self-reported direct aggression and unilateral friends’ self-reported direct aggression at low (−1 SD) and high (+1 SD) levels of popularity (n = 154); b: Association between self-reported direct aggression and unilateral friends’ self-reported direct aggression at low (−1 SD) and high (+1 SD) levels of peer rejection (n = 154).
Peer rejection also significantly moderated the association between individual self-reported direct aggression and unilateral friends’ self-reported direct aggression (see Figure 2B). Simple slope analysis showed that the slope for children high on peer rejection was not significantly different from zero, simple slope: b = 0.04, 95% CI = −0.02 to 0.09, p = 0.19. By contrast, for children low on peer rejection, there was a statistically significant association between individual direct aggression and friends’ direct aggression, simple slope: b = 0.14, 95% CI = 0.06 to 0.22, p < 0.001, such that children with high levels of direct aggression were more likely to have directly aggressive unilateral friends.
Depressive problems
To test hypothesis 3a and hypothesis 3b, main effects and interactions are calculated between social status and individual depressive problems on reciprocal friends’ depressive problems (see Table 4). There was a significant and positive association between individual self- and peer-reported depressive problems and reciprocal friends’ self- and peer-reported depressive problems. In line with a default selection perspective, these associations were significantly moderated by peer rejection. For one, peer rejection significantly moderated the association between individual self-reported depressive problems and reciprocal friends’ peer-reported depressive problems (see Figure 3A). Simple slope analysis showed that the slope for children low on peer rejection was not significantly different from zero, simple slope: b = −0.08, 95% CI = −0.21 to 0.05, p = 0.22. By contrast, for children high on peer rejection, there was a statistically significant association between individual depressive problems and friends’ depressive problems, simple slope: b = 0.22, 95% CI = 0.09 to 0.36, p < 0.01, such that children with high levels of depressive problems were more likely to have reciprocal friends with depressive problems.
Regression analyses of individual depressive problems on friends
*p<0.05, **p<0.01, ***p<0.001.
a: Association between self-reported depressive problems and reciprocal friends’ peer-reported depressive problems at low (−1 SD) and high (+1 SD) levels of peer rejection (n = 187); b: Association between peer-reported depressive problems and reciprocal friends’ self-reported depressive problems at low (−1 SD) and high (+1 SD) levels of peer rejection (n = 187).
Similar interactions were found with regard to the associations between individual peer-reported depressive problems and reciprocal friends’ peer- and self-reported depressive problems. Simple slopes with regard to reciprocal friends’ peer-reported depressive problems were not significantly different from zero: b = −0.10, 95% CI = −0.34 to 0.14, p = 0.41 for low peer rejection, and: b = 0.17, 95% CI = −0.02 to 0.37, p = 0.08 for high peer rejection. Simple slopes with regard to reciprocal friends’ self-reported depressive problems were: b = −0.14, 95% CI = −0.25 to −0.02, p < 0.05 for low peer rejection, and: b = 0.09, 95% CI = −0.01 to 0.18, p = 0.08 for high peer rejection (see Figure 3B).
Finally, it was explored whether youth high on depressive problems have unilateral friends’ with similar levels of depressive problems (hypothesis 4). Analyses showed that individual peer-reported depressive problems were positively associated with friends’ self- and peer-reported depressive problems. These associations were not dependent upon social status (not reported in table).
Discussion
In the current study, associations were examined between individual and friends’ direct and indirect aggression and depressive problems in preadolescence. Specifically, the study focused on the role of friendship reciprocity and individual social status. There are three main findings. First, in line with hypothesis 1b, associations between self-reported indirect aggression and self-reported friends’ indirect aggression were stronger when preadolescents were lower on popularity. This finding seems to offer circumstantial support the need for status notion, indicating that those low on social status behave more similar to peers they want to befriend by displaying behaviors associated with a popular status (see also Cillesen & Mayeux, 2004; Dijkstra et al., 2008). In order to receive status or approval from the peer group, preadolescents who are perceived as less popular may use indirect aggression to fit in with other, potentially more popular, peers (see also Dijkstra et al., 2008; Dishion & Tipsord, 2011).
Second, with regard to self-reported direct aggression it was found that the association between individual and friends’ direct aggression was stronger for those lower on popularity (hypothesis 2b). Yet, for preadolescents lower on peer rejection, the association between individual direct aggression and unilateral friends’ self-reported direct aggression was stronger. The latter findings contrast the hypotheses related to default selection and suggest that peer rejection and popularity differently modify the association between individual and friends’ direct aggression. One explanation could be that direct aggression is still associated with social status in preadolescence and that this association decreases only later in adolescence (see also Cillessen & Mayeux, 2004). As such, direct aggression may be imitated from unilateral friends to gain access to potentially more popular peer groups.
Of note, social status did not modify the association between peer-reported measures of individual and friends’ direct and indirect aggression. Caution is thus warranted as these findings were not replicated for the peer-reported measures and it suggests that there are important differences with regard to the informant of aggression.
Third, in line with hypothesis 3b and quite consistently across reporters, it was found that peer rejection modified the association between individual and reciprocal friends’ depressive problems. Such associations were not found with regard to unilateral friends’ depressive problems. Specifically, the association between individual and reciprocal friends’ depressive problems was stronger in those who were more rejected by their peers, but weaker for those who were less rejected by their peers. These findings could offer indirect support for default selection and co-rumination processes, suggesting that youth co-ruminate more in higher quality friendships (here: reciprocal friendships), and in turn foster more similarity between friends in depressive problems (see e.g. Furman & Rose, 2014). At the same time, the findings also suggest that such processes are stronger in youth who are more rejected by their peers. Yet, recent evidence from a study among monozygotic twins, showed that co-rumination only predicted increases in anxiety, but not in depressive problems (Dirghangi et al., 2015). This seems thus to contrast the notion of co-rumination as a valid explanation for the current findings. Hence, future research needs to examine the longitudinal changes in both friendships and depressive problems to shed more light on the underlying processes of similarity in depressive problems.
Although several hypotheses were (partly) supported, for other hypotheses no support was found. In contrast to hypothesis 1a regarding reciprocal friends’ indirect aggression and higher levels of social status, the association between individual and reciprocal friends’ indirect aggression was modified by popularity or peer rejection. This finding contradicts previous studies that showed associations between indirect aggression and high social status, in particular in adolescence (Card, Stucky, Sawalani, & Little, 2008; Cillessen & Mayeux, 2004; Heilbron & Prinstein, 2008). An explanation for this contrasting finding could be that in the current study most participants were still in preadolescence. As such, the observed rewards of indirect aggression may have been less prominent and not a meaningful criterion for forming friendships or behaving similar to friends.
In addition, hypothesis 2a, which was related to default selection and direct aggression was not supported. That is, low-status individuals were not more likely have reciprocal friends similarly high in direct aggression. In general, there was little similarity between individual and friends’ behavior in direct aggression. These findings contrast earlier studies that have reported default selection processes (e.g., Hektner et al., 2000; Sijtsema, Lindenberg, et al., 2010; Sijtsema, Rambaran, et al., 2013). One explanation could be that these previous studies focused on more extreme groups of aggressive youth who were also lacking prosocial skills.
Finally, the exploration that youth high on depressive problems have unilateral friends’ with similar levels of depressive problems (hypothesis 4), showed that preadolescents who are seen by their peers as sad seem to be in unilateral (and reciprocal) friendships with similar peers. Moreover, social status did not moderate this association. This finding is somewhat in line with a co-rumination perspective; that is, that youth who experience depressive problems have a need to share their negative thoughts and experiences with similar peers, and hence nominate depressed peers as friends.
Reconciling these findings, the role of social status on behavioral similarity in preadolescents seems to be different for aggressive and depressive behaviors. Whereas similarity in indirect and direct aggression was in particular observed in unilateral friendships in youth lower on popularity, similarity in depressive problems was mostly observed in reciprocal friendships in youth who were more rejected by their peers. Friendship reciprocity and the type of social status thus seem to be related to different behavioral outcomes (i.e., externalizing versus internalizing problems). Moreover, it should be noted that there were important differences regarding self- and peer-reported individual and friends’ behavior, especially with regard to direct and indirect aggression.
However, the current findings should be interpreted with caution for several reasons. First, the current study is cross-sectional in nature and therefore it was not possible to determine whether the associations were attributable to processes related to default selection, homophilic selection, a need for status, or co-rumination. Although it is tempting to assume that the current findings support one or more of these processes, only in a longitudinal design it can be tested whether friendships precede behavior or the other way around. Therefore, future research needs to investigate the current questions in more detail in a longitudinal study.
Second, there are multiple dependencies in the current data. That is, participants were both treated as individual and as potential friends, different individuals could have similar friends, and some individuals were included in the data set many times and others only a few. Therefore, assumptions of nonindependence were violated and individuals are unequally represented in the analyses. However, previous research has shown that these dependencies do not necessarily affect traditional multivariate analyses, as used in the current study (Scholte et al., 2009). In addition, friendships often do not exist as isolated dyads but these dyads are integrated in larger peer clusters (see e.g. Veenstra & Steglich, 2012). The current analyses did not account for these structural friendship clustering effects (e.g., transitivity: the tendency of friends to become friends with their friends’ friends) and as such did not fully account for the complexity of friendship relationships. To overcome part of these limitations, several controls were included in the regression analyses to account for the disparity in friendship nominations, the varying number of friends per individual, and the behavioral range in each cluster of friends.
Third, the current sample was relatively small. Although the sample size warrants sufficient power for detecting two-way interactions, it becomes more problematic when investigating three-way interactions. Therefore, interactions with sex were not included although there is ample evidence that the current findings may be attenuated by sex (Furman & Rose, 2014). Such interactions could be tested in a larger sample of preadolescents. Another issue related to hypothesis testing and power is the lack of a Bonferroni adjustment. Although a number of analyses were conducted, the p-value was kept at < 0.05, which may have increased Type I errors. However, significant interactions were only interpreted when simple slopes were also significant from zero.
Fourth, the popularity measure in the current study differed slightly from more traditional measures using most and least popular nominations. As argued by Dijkstra et al. (2010: 795–796), ‘[the current measure] reflects the extent to which adolescents want to affiliate with certain peers.’ As such, this measure assessed the power to attract peers. Although this reduces the comparability with other research, the current measure assessed a specific aspect of popularity (i.e., the power to attract) rather than the broad multidimensional notion that encompasses influence, social power as well as the power to attract (e.g., De Bruyn & Cillessen, 2006; Parkhurst & Hopmeyer, 1998).
Another issue is the meaning of self-reported behaviors and similarity between individuals and their unilateral friends. It may be that preadolescents report similar behaviors to their unilateral friends because they want to be like them or overestimate their similarity. A strength of the current findings was thus that support was found for the hypotheses regarding indirect aggression with regard peer-reported individual behavior and that the findings with regard to direct aggression were similar for self- and peer-reported individual behavior. Moreover, the self-reported measure of aggression was asked under the stem ‘when I am angry or mad at someone, I…’, which measure mostly reactive aggression. Future research may thus want to include additional assessments of aggression that also take proactive aggression into account.
In addition, future research may want to examine to what extent it can be assumed that unilateral friendships represent desired friendships. In line with previous research (e.g., Scholte et al., 2009; Sijtsema, Lindenberg, et al., 2010), much of the theoretical rationale in the current study hinges on the assumption that unilateral friendships are indeed preferred, but unobtainable, friendships. However, it could be that unilateral friendships are friendships or acquaintances that are less intense and intimate, and hence only observed by one individual in the dyad. Future research may want to assess desired friendships more directly by asking with whom youth want to be friends, but are not friends yet.
Despite these limitations, the current study extended previous research in several ways. The findings showed that it is important to consider friendship reciprocity and social status in explaining the association between individual and friends’ aggressive and depressive behavior. That is, unilateral friendships were an important context of behavioral similarity in direct and indirect aggression for preadolescents lower on popularity, whereas reciprocal friendships were an important context of behavioral similarity in depressive problems for preadolescents who were more rejected by their peers. Future research may thus benefit from these differential associations and potentially different underlying motivations for behavioral similarity.
Footnotes
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.