Esther Mertens

| 107 Classmate Influence in Intervention future research examining interpersonal relations in the class should add questions asking to which extent students are affected by the behaviors of other classmates. Limitations and future directions When considering the findings of our study, it is important to note some strengths and limitations. A strength of the study is the use of observations to assess classmates’ influences and mutuality. Using observations we were able to directly code modeling, reinforcement, and dyadic mutuality without depending on subjective perspectives of students. Furthermore, we examined both deviant and prosocial peer influences in not self-selected dyads. This allowed us to examine negative as well as positive peer influences with reduced selection effects. Moreover, our study had an experimental design with a pre- and post-measurement enabling us to examine changes in modeling, reinforcement, and the perceived classroom peer context. A limitation of the present study is the somewhat small sample size. Even though our sample size is rather large for an observation study, it might be that due to the relatively small sample size some relations failed to reach significance. Additionally, we did not examine classroom characteristics as predictors. Due to the limited number of clusters at classroom level the models were kept as simple as possible. However, classmates’ influences might depend on characteristics of the general classroom context such as class size or gender composition. Future research could focus on classroom characteristics and examine whether these characteristics influence students’ perceived classroom peer context. Moreover, the mediators and outcomes were measured at the same time point (i.e., post intervention). We have analyzed these time points because we expected changes in modeling and reinforcement to mediate intervention effects immediately after the intervention rather than intervention effects between post and follow-up measurements (Beauchaine & Slep, 2018). However, this approach limited the extent to infer causal order (Weeland et al., 2018). Furthermore, we measured victimization with only one item. Even though it is common in research concerning bullying to measure (types of) victimization with one item, it might be more reliable to use multiple items. Conclusion The present study showed that increases in prosocial modeling of classmates were related to decreases in students’ experienced victimization, especially when dyadic mutuality between classmates was high. However, changes in deviant and prosocial modeling and reinforcement did not mediate the effect of R&W on the perceived classroom peer context. Whereas prosocial modeling was related to victimization, 5