Introduction

Victimization has been broadly defined as harm caused by human agents acting in violation of social norms (Finkelhor 2011). In schooling contexts, victimization encompasses specific and non-specific actions such as physical acts, abuse, mistreatment or bullying intended to cause pain (Finkelhor 2008). Individuals victimized by peers at school are prone to suffer from mental health problems that affect their psychological wellbeing. Victimized adolescents are also particularly likely to be affected by several internalizing (depression, anxiety) and externalizing problems such as aggressive behavior and substance use (Graham et al. 2003; Schwartz et al. 2015; Zeman et al. 2002). Victimization impacts both on the psychological (i.e., intraindividual) and the social (i.e., interpersonal) experiences of harassed individuals (Berger and Rodkin 2009; Schwartz et al. 2013); victims are likely to suffer from long-lasting social isolation, loneliness, and anxiety, and their social competences are also compromised (Berger and Rodkin 2009; Copeland et al. 2013; Fleming and Jacobsen 2009; Veenstra et al. 2005). Victims frequently engage in suicidal ideation, suicide attempts, and actual suicide (Winsper et al. 2012). Their academic achievement is commonly affected too (Guay et al. 1999).

Globally, rates of victimization vary from country to country, most likely reflecting societal and cultural conditions and different ways of operationalizing this phenomenon (Cook et al. 2010). However, data suggest that, at the very least, around 15 % of the world’s schoolers have been victimized (Jimerson et al. 2010). In most countries, boys report higher levels of victimization than girls. Girls, however, tend to feel more vulnerable when victimized, reporting higher levels of relational victimization possibly because of the importance girls place on social relationships as compared to boys (Casey-Cannon et al. 2001). This trend is relatively stable across age groups (Zimmer-Gembeck et al. 2013). Particularly in Chile, where this study was conducted, the National Survey on Violence in the School Context (2009) found that 14.5 % of the student population reported being bullied, following strict criteria.

Still, not all students experience victimization in the same way or present the same symptoms (Fitzpatrick and Bussey 2014). Adolescent interpersonal relationships, and especially friendships, have been advanced as one most important factor diminishing the negative consequences of victimization (Boivin et al. 2001). This protective factor does not merely consist of having friends per se, but also encompasses the quality of these friendships (Hartup 1996; Hodges et al. 1997, 1999; Kendrick et al. 2012). Research has shown that adolescents with close friendships are less likely to be victimized by peers (Hodges et al. 1999) and perceive friendship self-efficacy (i.e., adolescents’ confidence in their own abilities to engage and communicate with a friend, to resolve conflicts, and to manage emotions) as a tool enabling them to deal with the negative effects of social victimization (Fitzpatrick and Bussey 2014; Bagwell and Schmidt 2011; Bollmer et al. 2005; Schwartz et al. 2000). Within school contexts, friendships are thus one salient developmental dimension related with peer victimization (Ostrov and Kamper 2015). Nevertheless, no systematic approach has been implemented to study the influence of the quality of friendships and their dimensions on the social experience of adolescents, particularly on peer victimization.

Adolescent peer relationships may take several forms: victimization is one of such forms, and so is friendship. Both are considered different ways to engage with the social world, in accordance with prevailing and valued cultural systems allowing individuals to simultaneously establish different forms of relationships, whether positive or negative (Ladd et al. 1996).

Close and significant peer relationships are related to sociometric group-localization (Marks et al. 2013), affecting individuals’ status and their acceptance or rejection among peers (Wentzel and Caldwell 1997). However, studies have usually overlooked the fact that being accepted by peers is different from being accepted by friends; peers refer to a wider social ecology in which the individual unfolds and defines his or her social standing, whereas friends constitute an emotional environment in which the individual searches for close, secure, and nurturing relationships. A measure of most intimate acceptance can be found in mutual friendship and reciprocity (Newcomb and Bagwell 1995), which provide high-quality social resources for social adjustment. Literature has shown that quality friendships impact on instrumental support, on the experience of group-belonging, and on the ability to establish intimacy (Steinberg 2010; Ryan and Deci 2009; Furman and Robbins 1985). High-quality friendships, which translate into experiences of support, intimacy, self-disclosure, and acceptance, can neutralize or inhibit a variety of symptoms related with victimization. Depending on the context, involvement in friendship relationships can mitigate negative attributions and emotions, and help develop coping strategies (Troop-Gordon et al. 2015; Steinberg 2010).

Friendship-quality adds to socioemotional wellbeing, especially during adolescence (Huebner and Gilman 2003; Elmore and Huebner 2010). When adolescents spend time with friends, they share experiences in an intimate and safe context, which allows them to elaborate on these experiences and learn from them, consequently improving their feelings of self-efficacy and subjective wellbeing (Caprara et al. 2006).

Friendship networks provide social support and protection, improving situational assessment and the subsequent behavioral responses (Burgess et al. 2006). Friendship intimacy is conceived as a central transactional process whereby two key components–self-disclosure and perceived responsiveness–facilitate or hinder a close connection for individuals to become friends (Chow et al. 2013). Intimacy involves having partners mutually disclose personal information and feelings. In order to establish a bond that allows the emergence of feelings and beliefs about being validated and cared for, friends need to perceive that their exchange of personal information is in fact successful (Shelton et al. 2010). Evidence has shown that adolescents with high friendship quality relationships are better at handling peer pressure and are more likely to resolve future conflicts in a positive manner (Hartup 1996; Berndt and Murphy 2002; Barry et al. 2009; Oriña et al. 2011). When they perceive that their mutual relationship is stable, self-esteem improves, they are better at perspective taking and decision making, are able to express disagreement, and develop verbal skills to constructively discuss with peers (Baumeister et al. 2003). Studies have also reported potential benefits linked with high levels of intimacy and self-disclosure, especially among adolescent girls (Bauminger et al. 2008). Even if consensus is not complete, literature suggests that girls share more time talking with same-sex friends than boys do (Rose and Rudolph 2006), and that boys spend more time involved in physical activities, such as games or sports, than girls (Rose 2002; Rudiger and Winstead 2013).

Current Study

Evidence suggests that high-quality friendships involve higher levels of peer acceptance, social support, self-esteem, and belonging, while victimization conversely affects those very same dimensions. In this sense, friendship quality may constitute a buffer against the negative consequences of peer harassment (Saarento et al. 2013). The main objective of this study was to explore the relationship between friendship quality, victimization and wellbeing. Specifically, we aimed to (1) assess the correlation of adolescents’ friendship quality, victimization and wellbeing outcomes considering the effects of gender and grade; (2) confirm the effects of peer victimization on adolescents’ wellbeing; and (3) test whether perceived friendship quality and its dimensions moderate the impact of peer victimization on wellbeing. In particular, we expected that victimized adolescents who perceive high quality in their friendships would experience less harm in their wellbeing. No specific hypotheses by friendship quality dimensions were formulated. In order to test these hypotheses, we adopted a longitudinal design with three consecutive assessments waves (with a 6-months interval). Wellbeing was assessed in time 1 (baseline score) and time 3 (1 year later). Victimization and friendship quality were assessed at time 2.

Also, considering gender differences in victimization (Zimmer-Gembeck et al. 2013), wellbeing (Bergman and Scott 2001) and friendship quality (Rose et al. 2016; Valkenburg et al. 2011) reported by earlier studies, we tested separate models for boys and girls. However, we did not expect different moderation processes by gender.

Method

Participants

Participants were 614 young adolescents (4th, 5th, and 6th graders, 50.1 % girls) from four middle-low socio-economical-status (SES) schools in Santiago, Chile. They were taking part of a 3-year research project (5 assessments, with 6 months interval) focused on peer relations (reports on other aspects of this project can be found in Berger et al. 2015; Berger and Caravita 2016; Berger and Palacios 2014; Palacios and Berger 2016). Schools were selected through convenience sampling; all students were invited to participate (according to the Chilean national socioeconomic classification, two schools corresponded to middle, one upper-middle and one lower-middle SES). Participants with data in the three assessments waves were included in the present analyses; 68 participants (11.1 %) did not have complete data, thus having a final sample of 546 adolescents. Attrition analyses showed that participating and excluded students had no significant differences in any of the study variables (all ts < 1.41, ps < 0.15). Parental consent and children’s assent were obtained for all participants following the ethical standards of the funding institution and the local university.

Measures

Victimization

Victimization was assessed using the Illinois Bullying and Fighting Scale during the second assessment wave (April 2013). This scale is an 18-item self-report measure consisting of three subscales that address how often adolescents engage in bullying, fighting, and peer victimization (Espelage and Holt 2001). For this study, we used a translated version of the four items of the victimization subscale (alpha = .84): “Other students called me names”; “Other students made fun of me”; “Other students picked on me”; and “I got hit and pushed by other students”. Participants were asked to report on a 4-point Likert scale (1 = never to 4 = almost always). Given its unidimensional structure, this measure was set as an observed variable.

Friendship Quality

Friendship quality was assessed in the second assessment wave (April 2013) with a 9-item self-reported questionnaire translated and adapted from Shelton et al. (2010), and based on Reis and Shaver’s intimacy model (1988) addressing interpersonal intimacy in friendship. Participants completed the nine items using a 5-point scale (1 = very little/very distant, 5 = very much/very close). A maximum likelihood CFA was conducted to check the factorial structure of friendship quality. Results confirmed the four dimensions structure reported for the original scale. Factor loadings ranged from 0.58 to 0.86 (standardized), with a maximum standard error of 0.08 for item 7. Model Fit Indexes indicated an appropriate goodness of fit (N = 547, χ 2 = 57.632 (21 df), p = 0.000; CFI = 0.98, TLI = 0.97, RMSEA = 0.05, SMRM = 0.03), and an adequate reliability (Cronbach´s alpha = .88). However, some items loaded in different dimensions compared to the original scale. Therefore, subscales were renamed according to the research questions of the present study: Perceived Affection (“How much of your feelings do you express to your friends?”, and “How much of their feelings do your friends express to you?”; alpha = .80); Disclosure (“How much personal information (e.g., information about you personally and your views) do you disclose to your friends during your interactions?”, and “How much personal information [e.g., information about they personally and their views] do your friends disclose during your interactions?”; alpha = .79); Perceived Support (“During your interactions, how much do you feel that you are accepted by your friends?”, and “During your interactions, how much do you feel that you are cared and supported by your friends?”; alpha = .77); and Closeness (“How well do you feel your friends understand you when you disclose issues about your private life?”, “Relative to all your other relationships, how much do you appreciate your relationship with your best friends?”, and “How close would you assess the relationship with your best friends?”; alpha = .67).

Gender invariance was tested with equality constraints for factor loadings, intercepts, and residuals for boys (N = 273) and girls (N = 273). Models were inspected for goodness of fit on every step and indices were consistently good. Indices for the final invariance model tested were χ 2 = 10.285 (10 df), p = 0.41. CFI = 0.99, TLI = 1.0, RMSEA = 0.01, SMRM = 0.02. ANOVA for absolute differences between χ2 for all models of invariance can be seen in Table 1.

Table 1 ANOVA: χ2 difference test for invariance models with restrictions based on gender
Full size table

Fit indices and χ 2 difference test show that the Friendship Quality Scale measures the same construct for boys and girls.

Socioemotional Wellbeing

Wellbeing was assessed with a reduced version of the self-reported Social Emotional Wellbeing scale in the first (October 2012) and third (October 2013) assessment waves. This scale was developed for the Chilean population by Arab (2009) as a screening tool for 8–12 years schoolers. The original scale consists of 52 items with Likert response of four options (1 = it happens to me all the time, 2 = it usually happens to me, 3 = it happens to me rarely, 4 = it never happens to me). Global internal consistency was high (alpha = 0.94; Milicic et al. 2013). We used a 19-items reduced version (alpha = 0.82; Berger et al. 2011) including items such as “I am alone at recess,” “I am confident that I can achieve my goals,” “When I have a problem I find ways to solve it”, “when I’m angry I know how to calm down”. This variable was set as observed, given its unidimensional structure.

Procedure

Data were gathered as part of a wider study in three consecutive assessments over a year. Instruments were completed during regular class hours through a group administration by trained assistants, taking approximately 45 min per application. Analyses were run using SPSS software version 20.0 and Mplus 7. Because of non-normality on some variables, we used a MLR estimator (Mplus option for maximum likelihood estimation with robust standard errors, Muthén and Asparouhov 2003). Latent Moderated Structural Equation modeling (LMS) in Mplus was used to test the study hypotheses. As of the date on which these analyses were conducted, Chi square and related fit statistics were not available on Mplus. Model adequacy was tested by firstly obtaining a loglikelihood difference between the null model and the proposed model and subsequently multiplying this difference by −2. This indicator follows a Chi square distribution (Muthén 2008). Also, since Mplus did not allow conducting multi-group analysis for LMS, gender-related model adequacy was tested separately for boys and girls. Two full models were compared with their respective base null model.

Results

Results are reported following the structure of our research questions and hypotheses. Table 2 shows overall descriptives for baseline scores for wellbeing, victimization, and friendship quality dimensions, by gender. Correlation coefficients (disaggregated by gender) show that wellbeing, victimization and friendship quality factors are significantly and moderately correlated, consistent with prior literature (see Table 3).

Table 2 Means and standard deviations of wellbeing, victimization, and friendship quality dimensions, by gender
Full size table
Table 3 Correlations between wellbeing, victimization, and friendship quality dimensions (Rho) by gender
Full size table

A MANOVA gender × grade (2 × 3) was conducted on 6 dependent variables: wellbeing, victimization, and the four chosen friendship-quality dimensions: affection, support, disclosure, and closeness. A significant effect was observed for gender (F(7,661) = 11.76 p < .001; Wilk’s Λ = .89, partial η2 = .11) and grade (F(14,1322) = 3.0 p < .001; Wilk’s Λ = .94, partial η2 = .03).

As for friendship quality (FQ), both gender and grade show a statistically significant effect on the four individual dimensions inspected. Gender on FQ (F (1, 667) = 67.04; p < .0001; partial η2 = 0.09); Affection F (1, 667) = 75.37; p < .0001; partial η2 = 0.10); Disclosure (F (1, 667) = 54.95; p < .0001; partial η2 = 0.07); Support (F (1, 667) = 13.94; p < .0001; partial η2 = 0.02); and Closeness (F (1, 667) = 39.52; p < .0001; partial η2 = 0.05). Grade on FQ (F (2, 667) = 13.70; p < .0001; partial η2 = 0.04); Affection (F (2, 667) = 8.06; p < .0001; partial η2 = 0.024); Disclosure (F (2, 667) = 8.87; p < .0001; partial η2 = 0.02); Support (F (2, 667) = 7.52; p < .0001; partial η2 = 0.02); and Closeness (F (2, 667) = 12.73; p < .0001; partial η2 = 0.04). Neither gender nor grade showed significant effects on wellbeing or victimization; no interaction were found.

Latent Moderated Structural Equations (LMS)

In order to assess the structural and moderation relations in the proposed model, we have disaggregated the structural and moderation multi-group modeling for boys and girls. Thus, Figs. 1 and 2 present structural models for boys and girls respectively, including standardized factor loadings and standardized regression coefficients for linear relationships between variables. As can be seen, goodness of fit of the two gender-related model is good (χ2 = 142.69, df = 82, p = 0.000; RMSEA = 0.05, CFI = 0.97, TLI = 0.96; SMRS = 0.03).

Fig. 1
figure 1

Model fit and factor loadings of the linear model for Boys. Note All factor loadings and regression coefficients are standardized. *p < .05

Full size image
Fig. 2
figure 2

Model fit and factor loadings of the linear model for Girls. Note All factor loadings and regression coefficients are standardized. *p < .05

Full size image

Individual effects of studied variables on wellbeing are summarized in Table 4. As expected, victimization negatively predicted wellbeing on both models (boys and girls). Baseline wellbeing positively and significantly predicted wellbeing in both models. As for friendship quality of boys, only disclosure and closeness had a direct and positive effect on wellbeing. For girls, no friendship quality dimensions significantly predicted wellbeing.

Table 4 Unstandardized regression coefficients and standard errors for the structural equation model for boys and girls
Full size table

Moderation Model

Interaction effects between victimization, friendship-quality dimensions and wellbeing were tested by means of an LMS. Figures 3 and 4 present the moderation model for boys and girls, respectively, with unstandardized scores (standardized coefficients were not available for LMS random type analysis on Mplus). In girls’ model, only disclosure and support significantly moderate the effect of victimization on socioemotional wellbeing. In boys’ model, only support moderates the effect of victimization on wellbeing. Both LMS moderation models were compared with their respective null model, using the above mentioned log-likelihood difference test. Results indicate that both moderation models fit data significantly better than their respective null models (log-likelihood difference for boys: Δχ2=4.36, df = 4, P < 0.05; log-likelihood difference for girls: Δχ2=8.09, df = 4, p < 0.05).

Fig. 3
figure 3

Simplified moderator model for Boys. Note Regression coefficients are unstandardized. *p < .05

Full size image
Fig. 4
figure 4

Simplified moderator model for Girls. Note Regression coefficients are unstandardized. *p < .05

Full size image

Discussion

The present study contributes to the existing literature on the protective role of friends, by zooming in on friendship quality as a main feature of adolescent friendships. Previous studies in this field have remarked the importance of high-quality friendship for the development of communication skills, emotional self-regulation and decision making to deal with the negative effects of social victimization (Fitzpatrick and Bussey 2014; Bagwell and Schmidt 2011; Bollmer et al. 2005; Schwartz et al. 2000). In particular, we show that the negative effect of peer victimization on wellbeing is moderated by friendship quality, thus stressing the relevance of positive and quality peer relations, based on communication and support, during this developmental phase.

Overall, results support our hypotheses. Consistent with previous findings, data show that victimization has a direct, negative effect on wellbeing, and that victimized adolescents reporting high-quality friendships do better on wellbeing. This adds to the large body of empirical evidence focusing on the role of high-quality friendships as a protective factor against conflicts arising from social interactions at school (Malti et al. 2015; Collibee et al. 2014; Fitzpatrick and Bussey 2014).

We believe that one important contribution of our study may lie in the observed gender differences and the way they relate with friendship-quality. These differences are most visible when comparing structural models and moderation models. We will first focus on the former.

Two of the four selected dimensions significantly influence wellbeing in boys: closeness and disclosure. However, these two dimensions act in opposite directions: while closeness coefficients are positive, disclosure coefficients are negative. Consequently, the higher the perception of disclosure, the greater the negative impact on wellbeing. While this finding might appear counterintuitive, previous research shed light on potential explanations. Schwartz-Mette and Rose (2012) remark that talking about problems can help in fostering intimacy with friends and in problem-solving, but can also lead to negative outcomes when friends engage in intense self-disclosure. When this is the case, disclosure tends to converge on negative feelings, promoting a negative assessment of the experienced situations leading to emotional difficulties (e.g., internalizing problems) (Alieske et al. 2014; Haggard et al. 2011; Rose 2002). High disclosure may translate the sharing of one’s problems and feelings with friends into collectively dwelling on them (Waller and Rose 2013). In any case, these results should be further explored to test this or alternative hypotheses. As for girls, none of selected the friendship-quality dimension significantly impacts on well-being, whether positively or negatively. Note that the effect of friendship quality on well-being was assessed after controlling for the stability path (well-being at time 1) and the negative effect of victimization, which may explain in part these findings.

The results of the moderation models for boys and girls are also different. Recall, however, that these models inspected the influence of victimization on wellbeing. As such, they are different from the structural models inasmuch as they do not reflect the influence of friendship-quality on the overall wellbeing, but the role played by friendship-quality when victimization is at play. When observing boys, only support significantly –and positively– moderates victimization effects on well-being. As compared to boys’ previous model, disclosure’s significance is lost. Most likely, this happens because boys have a tendency to receive less support for their expression of emotions than girls (Legerski et al. 2015). When boys experience victimization situations, they tend to refrain from publicly sharing their problems (Hébert et al. 2014; Leavitt et al. 2013). For them, disclosure not only entails reenacting unpleasant experiences and reinforcing the awareness of victimization (Almquist et al. 2013; Haggard et al. 2011), but also exposing themselves (Frijns et al. 2010; Pellegrini 2001). This, in turn, might make them vulnerable to mocking or teasing (Hashem 2015), or make them expect to feel weird and like they were wasting time (Rose et al. 2012).

As for girls, the influence of victimization on wellbeing is moderated by disclosure and support, which played no significant role in their structural model. In other words, results of the structural model suggest that friendship-quality does not significantly impact on girls’ general wellbeing. However, under victimization circumstances both disclosure and support do positively influence wellbeing. These results might be in line with gender stereotype theories. For girls, the social interactions included in friendship-quality might be a given (Berndt 1982). As such, they may be part of an evolutive process in which they engage naturally as they grow (Pratt and George 2005). Thus, girls tend to change their friends more frequently than boys do, expecting to obtain the satisfaction to their social needs through of these changes (Besag 2006). If this is indeed the case, none of the inspected friendship-quality dimensions should be expected to play a particularly salient role on general wellbeing. Nonetheless, girls do seem to ponder differently disclosure and support when they are a faced with stressful, unpleasant situations like victimization (Rose et al. 2016).

Differences between boys and girls could be related to the developmental requirements of social interaction. Boys and girls satisfy their social needs differently, as has been advanced by gender stereotypes theories (Guhn et al. 2013). Self-disclosure (verbally sharing personal, private information) is a more frequent and more spontaneous strategy among girls. Boys tend to resort to more direct and less verbal strategies (Valkenburg et al. 2011). Girls may display relational and linguistic skills in order to gain social support and sustain friendship intimacy, whereas boys may display direct physical interactions, which offer them feelings of closeness and support. Way (2004) argued that boys, while transiting through adolescence, have to respond to conventional notions of masculinity that reinforce autonomy and independency and exclude emotional intimacy. From this perspective, disclosure becomes a threat for boys, whereas it is a central feature of girls’ friendship experiences.

Finally, some remarks about the scope of our study are in order. The measure we used to assess wellbeing did not include mental health, internalizing and externalizing problems, and other indicators such as parental influence and support. Future research should include these dimensions. Also, in our study we did not consider specific adolescent profiles. Dimensions such as the duration of victimization or social integration, among others, may influence the role of friendships. In the same vein, future studies might consider specific attributes of mutual friendship such as the stability of relationships, duration, and friends’ similarity.

Conclusion

While literature on the relationship between friendship-quality does consider fine-grained analysis of the influence of some specific dimensions of friendship-quality on wellbeing, studies focused on victimization and wellbeing typically measure friendship in terms of the presence/absence of friends or gauge friendship discreetly (high or low, for instance). We tried to contribute to the field by closely inspecting the differential roles that the several aspects of friendship-quality may play on wellbeing.

We believe that our study contributes to the growing body of literature on peer relations, specifically on how different peer networks (victimization and friendships) overlap. Thus, our study supports the findings by earlier studies regarding the association between intimacy and interpersonal skills (Buhrmester 1990), and consequently as a buffer against peer victimization. However, our results broaden the understanding of developmental process regarding peer relations, suggesting that these associations are explained by gender patterns regarding socio-emotional development and skills during adolescence.

Our study also deeps the understanding of adolescent relationships by considering the perceived quality of friendships. These findings should inform educational interventions to prevent victimization’s negative outcomes and to promote quality relationships. In particular, teachers and school staff should foster activities (both curricular and extracurricular) and contexts in which friendships can raise and evolve, and where adolescents can explore and share their interests with peers in a secure and nurturing environment.