Introduction
Children and adolescents spend many hours a day interacting with same-aged peers in the school context, and considerable research effort has been directed at investigating the short- and long-term effects of these interactions. For example, a broad literature has examined the spreading of aggressive behavior in peer groups (e.g., Busching and Krahé
2015; Laninga-Wijnen et al.
2019). However, reviews have concluded that there is a lack of longitudinal studies focusing on the positive effects of peer interactions, for example in promoting prosocial behavior. Prosocial behavior is defined as behavior intended to benefit another, which includes helping, donating, sharing, and comforting (Eisenberg et al.
2015). Such studies are needed because although prosocial behavior and aggressive behavior are inversely related, they are not merely two opposite anchors of the same construct, but represent independent dimensions underlying adolescent friendship relations (Farrell et al.
2017). Therefore, the present study investigated how the development of adolescents’ prosocial behavior is shaped by the prosocial behavior of the peers in their classroom, applying multilevel analysis to data from a two-wave longitudinal study.
The Influence of Peers on the Development of Prosocial Behavior
During adolescence, the primary caregivers become less important, and peers gain more and more influence as socializing agents (Lam et al.
2014). A broad research literature has demonstrated the influence of peers on adolescents’ behavior in a variety domains. The focus of this research has been on problem behaviors, such as smoking and drinking (Ragan
2020; Vitória et al.
2020), problematic social media use (Marino et al.
2020), and aggressive and antisocial behavior (Jung et al.
2019).
Evidence on how the prosocial behavior of peers affects the development of individual prosocial behavior is scarce by comparison. A study found that young adolescents (aged 12-14 years) changed the probability that they would show a broad range of prosocial behaviors in the direction of the (fictitious) probabilities of showing these behaviors indicated by others, including both peers and adults (Foulkes, Leung, Fuhrmann, Knoll, and Blakemore,
2018). Another study found a significant association between the prosocial behavior of adolescents and their friends (Farrell et al.
2017). In a longitudinal study, the degree to which their best friends engaged in prosocial behavior predicted adolescents’ pursuit of prosocial goals, which in turn predicted their prosocial behavior, especially when the affective quality of the friendship and the interaction frequency were high (Barry and Wentzel
2006). In a cross-sectional study with Dutch adolescents, positive correlations between best friends’ engagement in voluntary work and participants’ readiness to volunteer were found (van Goethem et al.
2014). Also studying volunteering as a form of prosocial behavior, an experimental study showed that the impact of prosocial peer behavior on adolescents’ prosocial behavior was moderated by the perceived social status of the peer (Choukas-Bradley et al.
2015). This study used a chat-room paradigm in which participants communicated about volunteering in response to different scenarios with three peers from their grade, who were introduced as being high vs. low in social status. After exposure to the volunteering behavior of the peers, participants showed more prosocial behavior compared to a baseline assessment, but only if they believed the peers had a high social status. The increase remained significant in a subsequent phase when participants indicated their willingness to volunteer in private, knowing that it would not be communicated to others.
Studying Peer Effects on Prosocial Behavior in Classroom Communities
There is plenty of evidence that children and adolescents choose friends who are similar to themselves, including homophily with regard to prosocial behavior (Shin et al.
2019). Therefore, observed similarity among friends may be due to selection effects, socialization effects, or a combination of both. A three-wave longitudinal study of peer influences on adolescents’ smoking behavior used a cross-lagged design to separate selection from socialization effects and found evidence for both paths (Vitória et al.
2020). Adolescents’ smoking behavior in their younger cohort, similar in age to the present sample, predicted the choice of friend who smoked at Time 2, and friends’ smoking behavior at Time 2 predicted participants’ smoking behavior at Time 3. Neither selection nor socialization effects were found for same-grade students who were not friends. However, classroom effects were treated as control variables in this analysis and not examined in their own right as main effects or in interactions. A two-wave study from South Korea did not find an influence of classmates on prosocial behavior as assessed by teacher reports (Shin
2017).
In a sample of 51 classes, the overall level of prosocial behavior did not predict later prosocial and aggressive behavior of the individual class members (Laninga-Wijnen et al.
2018). However, when classroom characteristics were taken into account, a significant increase in prosocial behavior was found in classes in which there was an association between prosocial behavior and social status. This finding provides indirect evidence for the importance of normative beliefs at the classroom level because it indicates that only in classrooms where prosocial behavior was linked to high status did classroom prosociality affect the individual development of prosocial behavior. Further studies have shown individual prosocial behavior to be more closely linked with social status in classes with a high level of overall prosocial behavior compared to classes with a lower level of prosocial behavior (Dijkstra and Gest
2015; Torrente et al.
2014). However, other studies did not support this relationship (Boor-Klip et al.
2017; Stormshak et al.
1999).
The question whether the effect of the overall level of prosocial behavior in a classroom is moderated by individual differences in prosocial behavior has received little attention so far. Analyses of aggressive behavior have shown that initially nonaggressive students were influenced to a greater extent than initially more aggressive individuals by the collective level of aggression in their class (Busching and Krahé
2015; Rohlf et al.
2016). As parallel processes of observational learning may be assumed for negative and positive forms of social behavior, this evidence may be used to predict that initially less prosocial individuals would be influenced by the prosocial behavior of their peers to a greater extent than individuals with a higher initial level of prosocial behavior.
To provide a test of peer socialization effects on prosocial behavior, the present study was conducted in a setting in which selection effects are minimized and peer groups stay together over an extended period of time. The peers with whom adolescents arguably interact most in terms of time spent together are their classmates (Dishion
2014). Moreover, in contrast to their choice of friends, adolescents have no influence over the peers with whom they are in the same class, as students are assigned to classrooms by the school administration. This provides an excellent research opportunity for investigating the socializing effect of stable peer groups in which selection effects are minimized.
To investigate how the collective behavior of the members of a classroom shapes the development of individual students, the most suitable approach is multilevel analysis. In this approach, it is possible to disentangle effects due to individual characteristics, effects due to the classroom as a whole, and their interactions (Hox et al.
2017). Several studies have used multilevel analysis to demonstrate that peers who are surrounded by aggressive classmates also show more aggressive behavior over time (Henry et al.
2000; Henry et al.
2004). However, classmates’ behavior can also be considered as a resource. Some studies have shown that in classrooms with a higher level of prosocial behavior, individual students show less aggressive and antisocial behavior over time (Hofmann and Müller
2018) and show fewer teacher-rated problem behaviors (O’Brennan et al.
2014), but another study found no effect of the classroom level of prosocial behavior, assessed through peer nominations, on either class members’ aggressive behavior or their victimization by other classmates (Mercer et al.
2009).
The organization of secondary education in Germany provides the structural requirements for investigating peer influences in classroom communities from a longitudinal perspective. Schools typically have a number of parallel classes in each year, and students are assigned to a class by the school authorities without a say by the students themselves and their parents. Classes stay together for several consecutive school years, and students spend most of the school day in the community of their class. In combination, these features mean that self-selection into classroom peer groups is minimized, and classrooms provide a stable context for social learning experiences to occur. Although students are, of course, free to selectively interact with the other students in their class based on friendship and shared interests, they are nonetheless exposed to the behavior of all students in their class over extended periods of time. Therefore, studying the impact of the collective level of prosocial behavior within classrooms is a meaningful approach for understanding peer influences on the development of prosocial behavior. The measure of prosocial behavior used in this study was specifically designed to capture behavior shown in a school context so that it would be observable by all members of the class.
Social Learning as a Basis for Peer Influences on (Pro)Social Behavior
The psychological processes underlying the influence of peer groups on social behavior in general and prosocial as well as aggressive behavior in particular may be explained by social learning theory (Bandura
1977). This theory highlights observational learning as a key mechanism by which individuals acquire both positive and negative forms of social behavior (van Hoorn et al.
2016). Observational learning is not limited to the acquisition of patterns of behavior but also contributes to the development of social cognitions, such as cognitive scripts and normative beliefs about certain types of behavior (Huesmann
2018). For example, a longitudinal study showed that adolescents became more aggressive over time when they were surrounded by classmates who believed that aggression was an acceptable form of resolving interpersonal conflicts (Busching and Krahé
2015). In classes with a high proportion of aggressive students, aggressive behavior was found to be positively linked to popularity, which contributed to the normative acceptance of aggression in the class over time (Laninga-Wijnen et al.
2020). The link was reduced the higher the number of prosocial students in the class. In a study on intergroup contact between Catholics and Protestants in Northern Ireland, adolescents who perceived their peers to hold positive views about intergroup contact, defining a pro-outgroup norm, were more likely to engage in prosocial behavior toward outgroup members (McKeown and Taylor
2018).
In the case of prosocial behavior, the norm of reciprocity plays a crucial role. The norm of reciprocity refers to the belief that individuals should help others if these had helped them (Penner et al.
2005). Thus, being surrounded by prosocial peers, from whom individuals receive help, should increase their willingness to show prosocial behavior towards these others. This norm should be more salient if many adolescents show prosocial behavior and should facilitate the adoption of prosocial behavior through the process of observational learning. Social learning theory posits that similar models are more likely to serve as a source of observational learning than dissimilar models, so age-homogeneous groups, such as classroom communities, should facilitate observational learning. A three-wave study with young adolescents supports this line of reasoning by showing that Time 1 reports of receiving help from classmates predicted prosocial behavior to classmates at Time 2, which in turn predicted help received from classmates at Time 3 (Stotsky et al.
2019).
Gender Differences in Prosocial Behavior and Peer Influences
Several studies have shown gender differences in the level and development of prosocial behavior in adolescence. A peer nomination study found that girls were more often named as helpers than boys (van Rijsewijk et al.
2016), and there is evidence that the peak in prosocial behavior is higher and reached earlier in girls than in boys (van der Graaff et al.
2018).
Few studies have investigated gender as a moderator of peer influence effects. Early experimental studies of prosocial behavior found that the impact of a role model showing prosocial behavior was greater on girls than on boys (Grusec and Skubiski
1970). More recent studies have shown gender differences in the appreciation of prosocial behavior. While girls prefer friends with similar levels of prosocial behavior to their own, prosocial behavior does not seem to play a role in friendship selection for boys (Hsiao et al.
2019). However, in the study by Boor-Klip et al. (
2017), the relationship between individual prosocial behavior and social status was closer for boys than for girls.
Three different possibilities for gender to affect social influence in peer groups may be distinguished (Brechwald and Prinstein
2011): (1) the gender of the target person, for example whether girls are more easily influenced than boys. Relevant to this question is a finding that boys were more susceptible than girls to peer pressure to engage in antisocial behavior, but no such difference was found for prosocial behavior (Farrell et al.
2017); (2) the gender of the influencer, for example whether girls as a group have a stronger impact on their classmates than boys as a group. Such a main effect of influencer gender was observed for aggressive behavior in a study that found girls to have a greater impact as a group compared to boys on both their male and their female classmates (Busching and Krahé
2015); (3) the interaction of influencer and target gender, for example whether individuals are more influenced by same-gender than by opposite-gender peers. The latter possibility is suggested by the finding that more than 80% of the nominated helpers had the same gender as the nominating person (van Rijsewijk et al.
2016). To address these potential moderation effects, separate class-level scores of prosocial behavior at T1 based on the male and female class members, respectively, were calculated in the present study.
The Current Study
To investigate the impact of classroom prosocial behavior on the development of prosocial behavior in individual students over time, the current study was designed to test two predictions. The first prediction was that individuals would show more prosocial behavior over time if they were in a class in which the overall level of prosocial behavior was high than if they were in a class with a lower overall level of prosocial behavior (Hypothesis 1). This hypothesis predicts a main effect of the classroom level of prosocial behavior on individual class members.
Based on previous evidence concerning aggressive and deviant behavior, the second prediction was that classroom effects of prosocial behavior would be more pronounced on class members with lower prosocial behavior at T1 (Hypothesis 2). Specifically, it was expected that the initially less prosocial adolescents in a prosocial classroom would show a greater increase in prosocial behavior over time than those in the same classroom who were more prosocial to begin with. This hypothesis postulates a cross-level interaction between classroom level and individual level of prosocial behavior at T1 on individual prosocial behavior at T2. Multilevel analysis was employed as the correct approach for testing these hypotheses. This approach partitions the dependent variable into variation due to individual class members (individual level), differences between the classrooms (class level), and differences between schools (school level).
In addition to these two hypotheses, the present study investigated the role of gender as a potential moderator of classroom effects on prosocial behavior. Because results from past studies have been inconclusive, this issue was addressed as a research question rather than a directed hypothesis. Because gender was treated as a binary construct in the current data set, gender effects could only be examined for males and females.
Discussion
The development of prosocial behavior is a critical challenge in adolescence. Social learning theory conceptualizes prosocial behavior as a form of social behavior that is learned in the course of socialization, not only through direct reinforcement, but also through observational learning (Bandura
1977). Adolescents spend an extensive amount of time in the company of their peers, with ample opportunities for observational learning. Although past research has mainly focused on peer influences on the development of aggressive and antisocial behavior, positive peer behavior may be conceptualized as a learning resource that operates in a parallel way to affect individuals’ behavior. Accordingly, the current study examined the impact of classroom communities on facilitating the social learning of prosocial behavior in adolescence.
Exploiting the fact that the German school system minimizes self-selection into classrooms, and classroom communities remain together over consecutive school years, data from a large nationwide sample were used to test two hypotheses: The first hypothesis predicted that individuals surrounded by classmates with a higher collective level of prosocial behavior would be more prosocial about two years later than individuals surrounded by less prosocial classmates. The second hypothesis predicted that the impact of classmates’ prosocial behavior would be stronger on individuals starting off with a lower level of prosocial behavior than for individuals already reporting high prosocial behavior at the first data wave. Both predictions received support by the findings, consistent with the view that peers are influential socialization agents in adolescence. Whereas the negative influence of peers in increasing problem behavior has been studied extensively, scholars agree that the potential of peers to promote positive social behaviors and outcomes has received insufficient attention in past research (Barry and Wentzel
2006). The present study addressed this gap by demonstrating that individuals who are part of classroom communities with a high collective level of prosocial behavior showed an increase in prosocial behavior over time. In contrast to previous studies that failed to find a main effect of classroom-level prosocial behavior on the development of individual prosocial behavior (e.g., Laninga-Wijnen et al.
2018), the present data showed that individuals’ prosocial behavior increased if they were surrounded by prosocial classmates. A notable difference between the two studies is that the number of classes included in the present study was much higher (1308 vs. 51), yielding a more powerful basis for the estimation of class-level effects.
However, the main effect of the class level was qualified by a significant cross-level interaction with individual levels of prosocial behavior at Time 1. Whereas previous studies found evidence for class characteristics as moderators of class-level effects, such as the degree to which social status or popularity were tied to prosocial or aggressive behavior (Laninga-Wijnen et al.
2020), the present results showed that class-level effects were moderated by individual differences in the form of class members’ initial levels of prosocial behavior. In parallel to findings for aggressive behavior (Busching and Krahé
2015,
2018), being in a prosocial class community had a greater impact on the development of prosocial behavior of the initially less prosocial members than on those members who were more prosocial at the first data wave. Because the mean level of prosocial behavior in the present sample was below the midpoint of the response scale, this finding cannot be attributed to a ceiling effect. It is consistent with social learning theory, which sees the learning of social behavior as a result of both direct and vicarious reinforcement. Because prosocial behavior tends to be rewarded, observing peers’ prosocial behavior being followed by positive consequences should have a greater impact on initially less prosocial individuals because these individuals experience less direct reinforcement in response to their own prosocial actions. A combination of longitudinal and multilevel designs is particularly suited to detect these interactions of collective and individual prosocial behavior and is recommended for use in future studies.
The findings have theoretical as well as practical implications. At a theoretical level, they mirror the pattern of classroom main and cross-level effects found for deviant and aggressive behavior (Busching and Krahé
2015,
2018; Rohlf et al.
2016). This evidence is consistent with a conceptualization of peer influences based on social learning theory, which identifies observational learning and script learning as general mechanisms by which social behavior is shaped through social interactions. These learning principles are thought to operate in the same way across different types of behavior. Therefore, demonstrating similar effects for both anti- and prosocial behavior is a critical result. Future studies should extend these multilevel analyses to other domains of social behavior for which peer influences have been shown to be relevant, such as eating behavior (e.g. Salvy et al.
2012), smoking (e.g. Hoffman et al.
2006), or sexual behavior (Widman et al.
2016).
A noteworthy finding of the present study is that highly prosocial classrooms promoted prosocial behavior in initially less prosocial individuals, but classrooms with a low level of prosocial behavior did not make the initially more prosocial individuals less prosocial over time. This pattern mirrors the findings from previous studies of aggressive behavior, where aggressive individuals placed in classroom communities with low collective aggression did not become less aggressive over time (Busching and Krahé
2015,
2018). The finding that parallel patterns emerged with regard to two different types of social behavior suggests an explanation in terms of a general process rather than a process specific to prosocial behavior. Based on social learning theory, one could argue that classroom effects depend on the behavior being shown as a prerequisite for observational learning to occur. If a behavior is rarely shown, be it prosocial or aggressive, no learning opportunities are provided that could lead to changes in the individual’s behavioral repertoire.
Finally, the findings provide evidence on possible moderating effects of gender. This question has rarely been addressed within a multilevel framework, and past results have been inconsistent. In the present study, the collective prosocial behavior of both same-gender and opposite-gender peers predicted individual prosocial behavior at T2, but the influence of peers of the same gender was stronger. This pattern is consistent with the finding that same-gender friendships are by far more common than cross-gender friendships in adolescence. A study in six European cities found rates of other-gender friendships of only 21% for boys and 13.2% for girls (Grard et al.
2018). It is further in line with evidence that the vast majority of nominations of prosocial peers are made within gender groups (van Rijsewijk et al.
2016). This means that opportunities for learning prosocial behavior from same-gender peers are likely to be greater than learning opportunities involving other-gender peers, even within the context of mixed-gender classrooms.
This moderation effect differs from a previous study with adolescents that investigated aggressive behavior. In that study, both girls and boys were more influenced by the girls’ than by the boys’ normative acceptance of aggressive behavior in their class, and no same-gender effects were found for boys (Busching and Krahé
2015). The authors argued that the finding that girls define the boundaries for the aggressive behavior of girls and boys alike may be due to the increased importance of romantic relationships in adolescence, which may lead boys to conform to the norms about aggression defined by the girls. Why no parallel effect was found for prosocial behavior cannot be explained conclusively on the basis of the current data and needs to be addressed in future studies.
A further task for future studies is to integrate the different forms of pro- and antisocial behavior for which peer group effects have been demonstrated into a common design (Laninga-Wijnen et al.
2019). For example, can a high level of prosocial behavior in a class reduce individual class members’ aggressive behavior over time or is there a risk for a high collective level of aggression in a class to reduce individual class members’ prosocial behavior in a longitudinal perspective? There is some evidence that a high level of prosocial behavior may reduce individuals’ aggression (Hofmann and Müller
2018), but it is yet unclear whether the reverse link can also be found and whether these peer group effects are moderated by individual characteristics.
The findings also have implications for intervention measures designed to promote prosocial behavior. Because a higher level of prosocial behavior is associated with positive outcomes at the group level, such as a better school climate (O’Brennan et al.
2014), seeking to create a high collective level of prosocial behavior in classrooms is an important goal. Based on the present research, increasing the level of prosocial behavior by targeting all students may be expected to raise the overall level of prosocial behavior in the classroom (see Laninga-Wijnen et al.
2020, for a similar argument). As a result, the initially less prosocial students may shift in the direction of the higher level of prosocial behavior in their (Conklin et al.
2017). This could be a better strategy than targeting only the less prosocial students, whose behavior may be harder to change by explicit intervention efforts.
The gender effects found in the present study indicate that both same-gender and other-gender peers are influential, even if the influence of same-gender peers seems to be stronger. Interventions seeking to promote prosocial behavior at the classroom level should seek to address gendered norms and behaviors regarding prosocial acts and address the fact that cross-gender interactions become more important in the course of adolescence.
In evaluating the present findings, both strengths and limitations should be noted. The main strengths are the representativeness of the sample, the large number of classrooms and the use of state-of-the-art multilevel analysis to detect both main effects and cross-level interactions on changes in prosocial behavior from Time 1 to Time 2. A first limitation of this study is that it did not take teachers’ perspectives and behaviors into account. Teachers are an important source for the development of prosocial behavior in classrooms (Jennings and Greenberg
2009), and their reactions toward the overall prosocial behavior in a class could influence students’ behavior. However, if the classroom effects were due to teachers’ behavior, the parallel pattern for aggression and prosocial behavior would be hard to explain, since teachers may be assumed to react differently to the two forms of behavior, encouraging prosocial behavior and discouraging aggression. A second limitation is that prosocial behavior was assessed only by self-reports. However, if social desirability had led to an overreporting of prosocial behavior, this would have only affected the class-level means but cannot explain the class-level effects on individual behavior and the cross-level interactions. Nevertheless, future studies should include alternative measures of prosocial behavior, such as teacher ratings (Krahé and Möller
2011) or peer nominations (Mercer et al.
2009). Furthermore, only prosocial behavior was examined in this study. Social learning and the acquisition of prosocial scripts also involves prosocial norms, which have a strong evaluative component. Previous research on classroom effects on aggressive behavior showed that individual aggressive behavior increased not only in classrooms with a high level of aggressive behavior but also in classes with a high normative acceptance of aggression (Busching and Krahé
2015). From a theoretical point of view, a parallel finding should be expected for shared normative beliefs about prosocial behavior in a classroom. A final limitation is that the data set on which the present findings are based adopted a gender-binary definition that only offered the response categories of male and female. Broadening the analysis of gender effects to include non-binary categories of gender identification is a task for future research.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.