Peer Victimization and Friendship Quality: A Moderated Mediational Pathway With Social Withdrawal and Temperament

Victimization, Internalizing Symptoms, and Friendship Quality

Peer victimization is characterized by recurrent exposure to targeted acts of physical, verbal, and psychological abuse from peers with the intent to cause harm to the victim (Graham & Bellmore, 2007; Juvonen & Graham, 2001). This encompasses both direct forms, including physical and verbal harm, and indirect forms, such as relational victimization which targets an adolescents’ relationships and social standing (Cho et al., 2022). Peer victimization during early adolescence is a well-established risk factor for a myriad of internalizing difficulties (e.g., Hanish & Guerra, 2002). It has also been prospectively linked to numerous adverse long-term internalizing outcomes in adulthood, including higher rates of depression and anxiety (e.g., McDougall et al., 2015). Peer victimization likely significantly interferes with adolescents’ feelings of self-worth and social acceptance, which in turn fosters short- and long-term internalized distress (Mullan et al., 2023).

In addition to internalizing outcomes, research findings point to the negative impact of peer victimization on friendships, and specifically friendship quality (Bagwell & Schmidt, 2011). For instance, peer victimization, in its numerous forms, has been linked concurrently and prospectively with lower levels of overall positive friendship quality during early (e.g., Cho et al., 2022; Goldbaum et al., 2003) and middle adolescence (You & Bellmore, 2012). Negative linkages between peer victimization and the specific positive friendship qualities of security, companionship, and closeness have also been reported (e.g., Bagwell & Schmidt, 2011). These findings point to a spillover of distress from negative group-level peer difficulties (peer victimization) to dyadic-level peer difficulties (friendship), likely because the internalizing difficulties associated with peer victimization interfere with offerings of support and opportunities for companionship (Hanish & Guerra, 2002). In support, adolescents who are high in internalizing symptoms, such as anxiety, tend to be timid and verbally reticent with peers, including friends, which can, in turn, constrain social interactions and friendship quality during early adolescence (e.g., Rubin et al., 2006; Schneider, 1999).

In-line with Social Information Processing theory and research (Crick et al., 1994; Lemerise & Arsenio, 2000), it is also likely that experiences with peer victimization bias adolescents’ attention to ambiguous social cues and lead to hostile and rejecting interpretations and eventual avoidant coping. Such avoidant coping may in turn limit time spent with friends and supportive friendship interactions, thereby reducing positive friendship quality. Both interpretations raise the possibility that social withdrawal or socially avoidant behavior rooted in social fears and anxieties may mediate the prospective associations between peer victimization and positive friendship quality, a novel hypothesis that is evaluated for the first time in the present study.

Victimization, Friendship Quality, and Social Withdrawal

Although social withdrawal has not been previously evaluated as a mediator of the relation between peer victimization and friendship quality, past research has provided some support that peer victimization may lead to social withdrawal (path A), which in turn, predicts friendship quality (path B) (Barzeva et al., 2020, 2021). Of note, there are several different forms of social withdrawal, but the present study focuses on social withdrawal that is motivated by social fears and anxiety and is closely related to the constructs of anxious-solitude, shyness-sensitivity, and anxious-withdrawal (Bowker & White, 2021). The focus is on this form of social withdrawal because it is the most commonly studied form and has been consistently linked to a broad range internalizing problems (e.g., anxiety, depressive symptoms, loneliness) as well as such group-level peer difficulties including peer exclusion and peer victimization (Bowker et al., 2013; Bowker & White, 2021).

Regarding the A path, peer victimization is a well-established correlate of social withdrawal (Rubin et al., 2006) and there is growing evidence that peer victimization may lead to increased social withdrawal. For example, Barzeva and colleagues (2020) found that peer victimization predicted subsequent self- and parent-reported social withdrawal during early to middle adolescence. The authors posited that many victimized adolescents may withdraw from peers to cope with or to minimize further victimization, a notion consistent with evidence showing that behavioral avoidance is commonly used by adolescents to cope with interpersonal stressors (e.g., Grennan & Woodhams, 2007; Lee et al., 2016). There is evidence that some adolescents may view social withdrawal as an effective means of self-protection that minimizes potential negative future interactions, thus providing a sense of safety and control over one’s social environment (Bellmore et al., 2013). Additionally, highly socially withdrawn youth may further withdraw when they experience victimization because such peer abuse reinforces their social fears and anxieties (Bowker & White, 2021).

In terms of the B path, while social withdrawal is well-established as a risk factor for group-level peer difficulties (e.g., exclusion, victimization), less is known about its impact on dyadic-level peer experiences such as friendship. Importantly, relative to non-withdrawn youth, it is well-established that socially withdrawn adolescents are as likely to form at least one mutual best friend (Rubin et al., 2006). However, there is some indication that social withdrawal may impair the quality of those friendships. Indeed, findings from several studies have shown that adolescents characterized by high social withdrawal and their friends both perceive their friendships as less fun, helpful, and supportive compared to non-withdrawn youth (Rubin et al., 2006). In addition, observational data suggest that friendship dyads including one socially withdrawn adolescent show more restricted verbal communication relative to dyads without any socially withdrawn adolescents (Schneider, 1999). This may be because withdrawn youth tend to be highly anxious and they appear to self-silence (or intentionally withhold thoughts and feelings) with their friends, likely in an attempt to avoid interpersonal rejection (Bowker et al., 2023). Similarly, anxiety- and fear-driven behavioral withdrawal may limit time spent with peers, including friends, as well as opportunities for fun and companionship, emotional disclosure, and mutual support, all central features of high-quality friendships during early adolescence (Bagwell et al., 2021; Rubin et al., 2009). However, not all studies support this link. For example, Markovic et al. (2017) found no concurrent association between social withdrawal and friendship quality and speculated that such associations may only emerge at high levels of withdrawal severity during early adolescence. Similar non-significant concurrent associations were reported by Fordham and Stevenson-Hinde (1999).

Taken together, there is some indication that peer victimization may lead to reduced positive friendship quality through social withdrawal. However, a direct test of this mediational model is needed, especially since only individual paths have been evaluated in past research and evidence for path B has been mixed. Additional research is also needed to establish the prospective relations between social withdrawal and friendship quality as much of the research in this area has not been longitudinal (e.g., Markovic et al., 2017; Rubin et al., 2006). It is certainly possible that social withdrawal does not immediately impair friendship quality but does so over time as expectations for increased closeness and intimate disclosure grow with relationship length. Evaluation of the proposed mediational model could point to new therapeutic intervention targets with victimized youth. Findings could also enhance knowledge about why some youth struggle to form high quality relationships with friends during early adolescence and answer important unanswered questions about whether social withdrawal negatively impacts friendships over time.

The Role of Temperament

Given the multifinality of peer victimization, consideration of individual factors may help clarify why some adolescents respond to peer victimization with social withdrawal and others do not. Indeed, there are numerous ways in which early adolescents might respond to peer victimization, and social withdrawal or avoidant coping would be considered a more maladaptive response (e.g., Visconti & Troop-Gordon, 2010). Temperament has long been recognized as crucial to understanding the development of adjustment difficulties (Rothbart et al., 2000), and it could potentially function as a moderator of the prospective associations between victimization and social withdrawal. Although there are many models of temperament, this study focuses on the Reinforcement Sensitivity Theory (RST) because it is a learning theory that describes individual differences in reactivity to environmental contexts, such as the peer context, that can motivate approach and avoidance behavior. RST posits that approach and avoidance behaviors emerge from an individual’s sensitivity to both reward and punishment (Corr, 2008; Smillie et al., 2006). According to RST, an individual’s temperament is comprised of three distinct motivational systems, with the most relevant to the current study being the Behavioral Inhibition System (BIS) because it is linked to sensitivity to aversive or potentially aversive consequences (or the expectation of aversive consequences), anxiety, worry, inhibition and internalizing symptoms (Corr, 2008; Muris et al., 2005).

RST is highly relevant to our study because it shares significant overlap with the leading approach and avoidance models of social withdrawal (Asendorpf, 1990, 1993). According to approach and avoidance models, social withdrawal that is rooted in anxiety results from conflicting approach and avoidance social motivations. In this regard, socially withdrawn individuals are thought to desire to approach their peers but ultimately withdraw due to overwhelming fears and anxieties. While RST aligns with this notion of withdrawal being linked to sensitivity to social threats, it differs in its emphasis on reactivity to environmental stimuli rather than on internal motivational conflict as the primary driver of behavior.

Under the framework of RST, we reasoned that peer victimization may prompt youth characterized by high BIS to adopt avoidant coping behaviors due to the aversive and threatening nature of such experiences and strong desire to prevent further aversive experiences (Bellmore et al., 2013; Pérez-Edgar et al., 2011). Although effective in alleviating the initial anxiety within the situation, research has shown that avoidance inadvertently reinforces anxiously-withdrawn behavior (Biggs et al., 2012; Gazelle & Rudolph, 2004) and may increase the likelihood of not only future social withdrawal, but also psychological distress (e.g., Benatov et al., 2020). In sum, according to RST, adolescents have biological propensities that influence how they internalize and respond to contingencies in their environment, such as how they appraise situations to form expectations (e.g., victimization is possible in future interactions due to its past occurrence) and behavioral responses (e.g., socially withdrawing). Given this, we propose that adolescents with high BIS are more likely to develop negative expectations about future interpersonal interactions and withdraw socially following peer victimization, as they tend to interpret social situations as threatening (Miers et al., 2008; Pérez-Edgar et al., 2011; Waite et al., 2015). BIS as a potential moderator of the links between peer victimization and social withdrawal has not been previously evaluated and could inform us about who is at risk to respond with social withdrawal when peer victimization occurs and thus especially in need of intervention. Accordingly, the second aim of the study is to test BIS as a moderator of the association between peer victimization and social withdrawal. The behavioral manifestation of an adolescent’s heightened BIS is understood to be represented by their sensitivity to punishment (e.g., Smillie et al., 2006), which will be used to assess this hypothesis.

The Current Study

This study tested an indirect pathway from peer victimization to friendship quality through social withdrawal, and whether this indirect pathway is moderated by an adolescent’s sensitivity to punishment and used a longitudinal design to expand on past cross-sectional research and to help establish temporal precedence. Furthermore, we focus on the period of early adolescence (ages 10–14 years) because peer relationships become increasingly significant during early adolescence as there is a shift away from parental dependency and a heightened need for peer intimacy (Selman, 1989) and because peer victimization begins to increase in frequency and its impact during early adolescence (Molcho et al., 2009). Specific hypotheses include the following: First, peer victimization will be related positively to later social withdrawal (A path), which will subsequently predict poor friendship quality (B path). Second, the proposed indirect association will be strongest for adolescents exhibiting high levels BIS. See Figure 1 for hypothesized associations. Lastly, we will utilize both adolescent- and parent-report measures of social withdrawal as complementary perspectives to capture a more comprehensive view of the behavior. However, due to their low correlation (see Table 1), they will be examined in separate models.OPEN IN VIEWER

OPEN IN VIEWER

Table 1.

Bivariate Spearman Correlations

Variables	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16
1. PV W1	—	​	​	​	​	​	​	​	​	​	​	​	​	​	​	​
2. PV W2	.55	—	​	​	​	​	​	​	​	​	​	​	​	​	​	​
3. PV W3	.52	.59	—	​	​	​	​	​	​	​	​	​	​	​	​	​
4. SW-A W1	.34	.25	.29	—	​	​	​	​	​	​	​	​	​	​	​	​
5. SW-A W2	.23	.35	.25	.49	—	​	​	​	​	​	​	​	​	​	​	​
6. SW-A W3	.15	.21	.28	.42	.58	—	​	​	​	​	​	​	​	​	​	​
7. SW-P W1	.03	.07	.05	.27	.20	.18	—	​	​	​	​	​	​	​	​	​
8. SW-P W2	−.02	−.02	.02	.22	.17	.19	.75	—	​	​	​	​	​	​	​	​
9. SW-P W3	.05	.00	.04	.18	.13	.19	.63	.72	—	​	​	​	​	​	​	​
10. FQ W1	−.14	−.10	−.06	−.16	−.14	−.04	−.13	−.11	−.06	—	​	​	​	​	​	​
11. FQ W2	−.17	−.14	−.18	−.18	−.21	−.13	−.19	−.18	−.16	.59	—	​	​	​	​	​
12. FQ W3	−.16	−.17	−.16	−.16	−.19	−.11	−.18	−.18	−.18	.56	.71	—	​	​	​	​
13. SP	.03	.01	.00	.28	.22	.23	.54	.51	.52	−.14	−.17	−.14	—	​	​	​
14. Sex	−.12	−.10	−.10	.09	−.07	.05	.01	−.03	.04	.33	.38	.45	.08	—	​	​
15. Race	−.05	−.03	.01	−.10	−.18	−.14	−.12	−.11	−.10	−.07	−.06	−.06	−.11	−.02	—	​
16. Age	−.06	−.11	−.10	.04	−.05	.00	.00	.04	.04	.03	.02	.01	.03	.04	.01	—
Mean	1.38	1.36	1.32	0.29	0.26	0.24	0.30	0.31	0.31	3.75	3.89	3.97	2.67	0.55	0.83	12.09
SD	0.40	0.38	0.38	0.27	0.30	0.29	0.35	0.35	0.34	0.58	0.57	0.57	0.47	0.50	0.37	0.59
Skewness	1.08	0.96	1.45	0.94	1.29	2.15	1.87	1.37	1.48	−0.21	−0.37	−0.56	0.42	−0.20	−1.78	−0.16
Kurtosis	1.36	0.64	3.01	0.40	1.50	6.98	4.58	1.66	2.29	−0.34	−0.14	0.10	0.25	−1.97	1.17	−1.12

Note. Bolded values are statistically significant at p < .05. PV = Peer victimization; -A = Adolescent Report; -P = Parent Report; SW = Social Withdrawal; FQ = Quality of Friendship; SP = Sensitivity to Punishment; Race was dummy coded where white = 0; Age was set only at W1; Sex was coded as Male = 0 and Female = 1.

Participants

Participants were from a longitudinal study which examined risk and protective factors for substance use initiation. Participants were recruited through random digit dialing between 2007 and 2009 in western New York. Data was collected across 9 annual waves. Families were eligible to participate in the study if they had a child between the ages 10 to 12, were fluent in English, and had a caregiver willing to participate. One child/household was invited to participate (see Trucco et al., 2014 for more details about recruitment). The current study used data from the first three assessments (Waves 1, 2, and 3) when data on peer victimization and quality of friendships was collected.

The sample was composed of 387 adolescents (52.4% female) and a parent. The racial and ethnic composition of the target adolescents within the sample were as follows: 83.1% Non-Hispanic Caucasian, 9.1% African American, 2.1% Hispanic, 1.0% Asian, and 4.7% racially/ethnically mixed. The median income for families who participated was $70,000 with a range of $1,500 to $500,000. The mean age for adolescents was 11.61 (0.55) at W1, 12.63 (0.58) at W2, and 13.61 (0.57) at W3. Overall attrition for the sample was low (7.5%) across W1–W3. Adolescent-parent dyads who did not complete W2 or W3 assessments did not significantly differ demographically (i.e., sex, race/ethnicity, education, income, and marital status) from dyads who completed all three assessments. Given the high retention and lack of demographic differences, it is likely that attrition had a minimal impact on results. Nevertheless, hypotheses were tested using full information estimation (Cham et al., 2017), which permits all possible data in the analysis, even cases that are not present at W2 or W3.

Procedures

Adolescents and one of their parents participated in assessments conducted in university research offices. Follow-up waves of data collection (i.e., W2 and W3) took place approximately one year after the preceding wave. Prior to participating in the study, the adolescent and parent provided informed consent and assent, respectively. Dyads were then escorted into different rooms for data collection, which consisted of laboratory tasks and computer administered questionnaires designed to assess a broad spectrum of family, peer, and individual developmental factors. Questionnaires were read aloud to the adolescents by research assistants and adolescents entered their responses into a computer. Families were compensated for their participation in the study ($75 at W1, $85 at W2, $100 at W3). All procedures received approval from the [Removed for Peer Review] Institutional Review Board.

Measures

Peer Victimization (W1–W3; adolescent report) was assessed with four items from the Perceptions of Peer Support Scale (Kochenderfer & Ladd, 1996; Ladd et al., 1996; e.g., “Other kids hit or kick you”) and one item from the Multidimensional-Peer Victimization Scale (Mynard & Joseph, 2000; i.e., “Other kids make fun of you because of your appearance”) to increase the breadth of the items. Response choices ranged from 1 (“never”) through 3 (“a lot”). Items were averaged to create an aggregate score at each wave. While the Perceptions of Peer Support Scale was created for children (ages 6–9), it has been widely utilized in early adolescent samples (e.g., Meisel et al., 2019). Polychoric correlations were used to compute alpha coefficients to accommodate the 3-level ordinal response options. Alphas suggested strong internal consistency at each Wave (α = 0.90 at W1, 0.90 at W2, 0.93 at W3).

Social Withdrawal (W1–W3; parent and adolescent report) was measured using a subset of five items from of the Withdrawn/Depressed subscale from the Youth Self Report (YSR) and Child Behavioral Checklist (CBCL). The YSR (adolescent report) and CBCL (parent report) have been validated in early adolescence samples (Achenbach & Rescorla, 2001) and used in early adolescent research (e.g., Barzeva et al., 2020). The items included where as follows: rather be alone than with others, refuses to talk, secretive or keep things to self, shy or timid, and keeps from getting involved with others. These items have been used previously in studies of anxious-laden social withdrawal (e.g., Barzeva et al., 2020). We did not include the three items that captured depression (e.g., “Unhappy, sad, or depressed”) consistent with prior work that has focused on social withdrawal (see Bagwell & Schmidt, 2011). Responses were averaged to form a composite score at each wave for each reporter. The YSR (α = 0.72 at W1 and 0.83 at W2 and W3) and the CBCL (α = 0.86 at W1, 0.84 at W2, and 0.85 at W3) demonstrated adequate internal consistency across all waves computed with polychoric correlations.

Friendship Quality (W1–W3, adolescent report) was assessed using items from the companionship, intimate disclosure, reliable alliance, affection, and reassurance of worth subscales from the Network of Relationships Inventory-Revised, due to its validation in early adolescent samples (NRI-R; Furman & Buhrmester, 1985). The items reflect positive aspects of friendship quality ((i.e., “How much free time do you spend with this friend?”, “How much do you share your secrets and private feelings with this friend?”, “How sure are you that this relationship will last no matter what?”, “How much does your best friend really care about you?” and “How much does your best friend like or approve of the things you do?”). Target adolescents were instructed to think about their three closest friends (their first, second, and third closest friends) and then they responded to each item for each target friend using a 5-point response scale (1 = “Not a lot” or “Never” to 5 = “Very much” or “Extremely sure”). As done previously (e.g., Barzeva et al., 2021), responses were averaged across the three friends to capture overall positive friendship quality. For adolescents who nominated fewer than three friends (W1: n = 11; W2: n = 15; W3: n = 10), the composite reflected the average of the available reports. Each friend’s score correlated strongly with the others within the wave (i.e., W1 ranged from rs = .61 to .69; W2 ranged from rs = .65 to .71; W3 ranged from rs = .67 to .70). The NRI-R demonstrated adequate internal consistency (α = 0.74 at W1 and 0.79 at W2 and W3) across all waves.

Sensitivity to Punishment (BIS; parent report) was assessed through the Sensitivity to Reward and Punishment Questionnaire (SPSRQ-C), a scale designed for and validated in early adolescence (Colder & O’Connor, 2004). This measure was used as an individual’s sensitivity to punishment (SP) can be understood as the behavioral manifestation of a heightened BIS (e.g., Smillie et al., 2006). This consists of 25 questions (e.g., “In unfamiliar tasks, your child worries about failure” or “Your child often refrain from doing something because they are afraid of getting into trouble”). Items demonstrated good internal consistency (α = 0.87 at W1 and 0.88 at W2 and W3). One-year stabilities for SP (rs = .74 to .81) were high across waves. Given this, SP items were averaged across waves to create an aggregate factor to represent trait levels.

Data Analysis Strategy

Cross-lagged path models with Robust Full-Information Maximum Likelihood estimation (MLR) estimated in Mplus version 8.0 (Muthen & Muthen, 1998–2017) were used to test study aims. Cross-lagged path models were well-suited for this study as they allowed for the examination of bidirectional relationships over time and provide evidence for temporal ordering of the proposed pathways. MLR adjusts fit indices and standard errors for non-normality and is appropriate as some of the included variables were non-normally distributed (see Table 1). Models included within-time covariances among disturbances. Age, child sex at birth, and race were considered as potential statistical control variables and were included as predictors for all endogenous variables. More details on these paths are provided below. Of note, the model depicted in Figure 1 includes stability paths for victimization, friendship quality, and social withdrawal to permit prediction of rank-order change. Moderation was tested by including SP x peer victimization cross-product terms prospectively predicting W2 and W3 social withdrawal. Bias-corrected bootstrap confidence intervals were generated to test indirect effects (Hayes & Scharkow, 2013).Nested tests were used to examine equivalence of paths across time (stability and cross-lagged path coefficients). We estimated separate models for parent- and adolescent-reported social withdrawal to reduce model complexity given the available sample size. Moreover, the low correlation between parent- and adolescent-reports (see Table 1) suggests that they capture different aspects or perspectives of adolescents’ social withdrawal, suggesting limited overlap in the information provided. As such, we were unlikely to lose substantial information by modeling each reporter separately rather than estimating the unique effects of one report above and beyond the other.

Model fit was assessed using the following four indices: Comparative Fit Index (CFI), Standardized Root-Mean-Square Residuals (SRMR), Tucker-Lewis Index (TLI), and Root Mean Square Error Approximation and it’s 90% CI (RMSEA). Fit thresholds were taken from Marsh and colleagues (2004) in addition to Hu and Bentler (1999). For the CFI and TFI, the thresholds were: <0.90 is considered poor, 0.90–0.94 is deemed adequate, and >0.95 is indicative of excellent fit. Thresholds for the SRMR were: .09 is poor, .06–.09 is adequate, and <.06 is excellent. Lastly, thresholds for the RMSEA were: .08 is poor, .05–.07 is adequate, and <.05 is excellent.

Peer Victimization and Social Withdrawal

Contrary to expectation, prospective analyses showed no evidence linking prior levels of peer victimization to later parent- and self-reported social withdrawal when controlling for earlier waves of the behavior. Additionally, prior levels of parent- and self-reported social withdrawal did not predict subsequent levels of peer victimization. These findings are counter to those from several studies suggesting that social withdrawal and peer victimization mutually exacerbate one another over time (e.g., Almeida et al., 2021; Boivin et al., 2010). Additional research is clearly needed, but these discrepant findings may be explained by methodological differences in how peer victimization was measured across studies. For instance, peer nominations were the typical method for assessing peer victimization in the studies that reported reciprocal associations (e.g., Boivin et al., 2010). Peer nominations may capture more “public” victimization and perhaps also more chronic victimization (as it is associated with social reputations for peer victimization) relative to self-report measures. This type of victimization, in turn, may be most likely to contribute to, and result from, socially withdrawn behavior. In contrast, self-reported victimization, as utilized in the current study and in the study by Barzeva and colleagues (2020), may capture emotional distress that is less apparent to peers and, perhaps, less directly linked to social withdrawal. Supporting this interpretation, Barzeva and colleagues also found no evidence of a prospective association between self-reports of social withdrawal and peer victimization when social withdrawal’s stability was significant. In short, peer nominations vs self-reports may identify slightly different victimization experiences that have different implications for understanding social behaviors and psychological well-being.

Another plausible explanation for the null findings could be that many early adolescents who engage in social withdrawal do so effectively, thereby reducing their social interactions and minimizing their exposure to peer victimization. By withdrawing, adolescents may create a protective buffer, in which they intentionally or unintentionally avoid situations where victimization is more likely to occur. It may also be important to consider other types of negative group-level peer experiences such as peer exclusion (Graham & Bellmore, 2007) as exclusion may more directly impact adolescents’ approach and avoidance behaviors (Gazelle & Ladd, 2003). Consistent with this notion, the findings from several studies show that exclusion is a stronger contributor than victimization to other types of adjustment outcomes, such as substance use during early adolescence (e.g., Meisel et al., 2019).

Despite the absence of prospective associations between earlier peer victimization and subsequent social withdrawal, peer victimization was associated with adolescent-reported social withdrawal concurrently, even after accounting for stability. Peer victimization may have an immediate impact on adolescent behavior, influencing functioning in the present rather than predicting changes over time. The stability of social withdrawal indicates that once established, it tends to persist, yet peer victimization may affect adolescents’ behavior more proximally than one-year intervals. There may also be additional factors not included in the current study that contribute to both peer victimization and social withdrawal. Self-esteem is one example as it has been shown to have a robust association with peer victimization (Reijntjes et al., 2011) and links with social withdrawal (Cruz et al., 2023). Additionally, given the ever-growing research on the importance of classroom-level factors and peer responses to victimization, it is also possible that peer victimization only leads to social withdrawal in classroom environments that support or allow for peer victimization and avoidant behaviors or when peers fail to intervene (Salmivalli et al., 2021). In conclusion, while peer victimization has potential proximal impacts on adolescent behavior, the concurrent nature of these effects underscores the need for further exploration into underlying factors that contribute to both victimization and subsequent behavioral outcomes.

It was also hypothesized that an adolescent’s sensitivity to punishment, combined with experiences of peer victimization, shape their expectations regarding peer relationships, which in turn influence their behavior. However, adolescents’ sensitivity to punishment did not moderate the association between prior peer victimization and later social withdrawal. The lack of support suggests that victimization may be such a salient experience that, regardless of temperament, it is aversive with little subtlety. However, results revealed an association between sensitivity to punishment and both parent- and adolescent-reported social withdrawal, indicating that while sensitivity to punishment may not moderate the proposed association, it may still influence specific aspects of social withdrawal. This pattern is consistent with RST’s emphasis on threat-related predispositions toward avoidance behavior (Corr, 2008; Smillie et al., 2006), wherein adolescents with heightened sensitivity to punishment are more likely to engage in withdrawal due to anticipated negative outcomes. It also aligns with past research (Bowker et al., 2017) and Asendorpf’s approach-avoidance model (Asendorpf, 1990, 1993), which conceptualizes anxious-withdrawal as rooted, in part, in strong behavioral inhibition systems. The significant role of temperament to the development of anxious-withdrawn behavioral tendencies during early childhood has been well-established (Rubin et al., 2009). However, the present findings suggest that researchers should perhaps also more carefully consider temperament in studies of social withdrawal during early adolescence.

Social Withdrawal and Friendship Quality

The results of the current study demonstrate a clear prospective association between prior reports of parent- and self-reported social withdrawal and a decline in friendship quality, even after accounting for earlier levels of friendship quality. These findings are notable because, while the negative effects of social withdrawal at the group level of social complexity are well-documented, its impact on dyadic peer experiences has been less explored (Bowker & White, 2021). Therefore, these results extend knowledge by demonstrating that socially withdrawn behavior, particularly when it is anxiety-driven, negatively impacts the positive qualities of young adolescents’ friendships. This may be because social withdrawal involves not only timid and shy behaviors, but also verbal reticence, all of which is likely to interfere with intimate disclosure, offering emotional and social support, and fun and companionship. These findings are novel and suggest although socially withdrawn adolescents may not struggle to form friendships, their friendships may deteriorate over time in ways that hurt rather than help during the critically important early adolescent developmental period. Important next steps will be to determine: (1) which specific friendship processes interfere with positive friendship qualities (such as self-silencing or others; Bowker et al., 2023); (2) whether social withdrawal is especially likely to erode positive friendship qualities over time when the friendship involves friends who are both high in social withdrawal; and (3) whether social withdrawal also prospectively impacts the negative qualities of friendship, such as the extent to which conflict occurs, all of which we were not able to evaluate in our study. In addition, we think it will be important to evaluate whether other forms of social withdrawal, such as unsociability, similarly interfere with both positive and negative friendship quality over time, albeit likely for different reasons.

Notably, a common approach in the broader literature is to assess friendship quality with a single best friend (e.g., Markovic et al., 2017; Schneider, 1999). However, in the current study, an adolescent’s friendship quality was assessed using a composite that combined measures across three of the adolescent’s closest friends. Hence, the index of friendship quality in the current study reflects quality of friendship across multiple friends. This approach offers the distinct advantage of capturing a broader view of the adolescent’s social network, thereby providing valuable insight into their larger social support system. This perspective is particularly important because early to middle adolescents often experience friendship dissolution (Flannery & Smith, 2021), making it crucial to consider the wider social support beyond a single best friend. That said, post hoc analyses indicate that the prospective association between social withdrawal and a decline in friendship quality remained consistent even when focusing solely on the adolescent’s single closest friend.