When Does the In-Group Like the Out-Group? Bias Among Children as a Function of Group Norms

Research indicates that in-group favoritism is prevalent among both adults and children. Although research has documented that individuals do not consistently display an in-group bias, the conditions under which out-group preference exists are not well understood. In this study, participants (N = 462) aged 9 to 16 years judged in-group deviant acts that were either in line with or counter to a generic norm shared by both groups. The findings demonstrated, for the first time, that children preferred out-group over in-group deviance only when the in-group peer’s deviance was in line with the generic norm and a threat to their group’s identity. Participants justified their disapproval of these deviants by focusing on the need for group cohesion and loyalty, while they signified their approval by spotlighting the need for autonomy. Our findings suggest that children’s intergroup attitudes are influenced by how the behavior of their peers matches different levels of group norms.


Research Article
Extensive research in social and cognitive psychology has shown that social groups are readily categorized as either "us" or "them"-and individuals typically favor "us" over "them." This pervasive tendency, known as in-group bias, may be related to prejudice (Dovidio, Hewstone, Glick, & Estes, 2010;Rutland, Killen, & Abrams, 2010). In-group bias is highly common among children but not universal. For example, low-status or socially disadvantaged children often do not exhibit explicit in-group bias (e.g., Griffiths & Nesdale, 2006;Shutts, Kinzler, Katz, Tredoux, & Spelke, 2011). Thus, the ubiquitous assumption that individuals are driven by ethnocentric motives to support their in-group may not be the whole story. In particular, the conditions under which children prefer out-group members are not well understood.
Psychological theories have explained out-group favoritism by proposing that stereotypes or value consensus across group boundaries can reinforce and legitimate existing social systems or hierarchies ( Jost & Banaji, 1994;Lee, Pratto, & Johnson, 2011). For example, system-justification theory posits that holding favorable attitudes toward the existing social system and status quo may override individuals' tendency to show in-group favoritism ( Jost, Banaji, & Nosek, 2004).
To date, the occurrence of out-group favoritism in childhood has been relatively neglected. We argue that an overlooked factor that may determine in-group bias in childhood has to do with group norms and how children's desire to uphold these norms to augment group identity can engender both in-group bias and out-group favoritism (see relevant research in adults by Marques, Abrams, Paez, & Hogg, 2001). Recent research has suggested that even preschool children conform to peergroup norms (e.g., Corriveau & Harris, 2010), and peer groups become increasingly salient in late childhood and adolescence. By adolescence, individuals are especially concerned about being socially excluded by peers (e.g., Brown, Clasen, & Eicher, 1986;Gieling, Thijs, & Verkuyten, 2010). This aligns with recent developmental neuroscience research into brain development and social exclusion in adolescence, during which individuals become increasingly sensitive to other's emotions and to being rejected from their peer group (see Somerville, 2013).
Research on the importance of group norms and sensitivity to peer rejection indicates that children and adolescents are concerned about adhering to group norms as well as favoring members of their own group and that this concern is motivated by a desire for positive group identity (see Killen, Rutland, Abrams, Mulvey, & Hitti, 2013;Rutland et al., 2010). Indeed, research on peergroup dynamics has revealed that by middle childhood, individuals prefer a member of the out-group who conforms to their in-group's norm (i.e., an out-group deviant) over a member of their own group who dissents from their in-group's norm (i.e., an in-group deviant; e.g., Abrams, Rutland, & Cameron, 2003). From middle childhood into adolescence, individuals' understanding of group dynamics continues to develop as they begin to simultaneously consider multiple norms, conventional and moral, when making judgments and reasoning about social relations within and between groups Killen et al., 2013;. Therefore, given this developmentally sensitive period, we used a sample of 9-to 13-year-old participants in the present study.
Social-psychological research, however, has shown that young adults favor an out-group deviant over an ingroup deviant only when the latter is a full member (i.e., has been accepted by the group after a period of socialization) and the in-group lacks a sense of "groupness" (Lewis & Sherman, 2010;Pinto, Marques, Levine, & Abrams, 2010). This is because full members have a special role in validating the group's identity, and if they deviate from the group's norm, they become a meaningful threat to the group's identity (Levine & Moreland, 2002). This research with adult samples has demonstrated a context in which perception of a threat to the group identity may influence the manifestation of in-group preference.
What has not been studied is how different levels of group norms are related to out-group preference in childhood. Generic norms (defined as societal-level expectations) are different from group-level norms (defined as expectations held by a specific peer group). Individuals may attend to generic norms that align with shared expectations of a specific group more than to those norms that are solely formulated by a particular peer group.
We argue that the out-group preference that has been documented in previous studies is most likely when the out-group deviance aligns with larger generic norms, or, in other words, when deviance actually conforms to broad societal expectations. This is a fundamental distinction, because in this situation, the out-group deviance provides two bases for attracting favorable responses from an ingroup member-by supporting the in-group norm and supporting the generic (societal-level) norm. Understanding this distinction can illuminate how group norms bear on the wider issue of in-group bias and out-group preference.
Studying this phenomenon in childhood provides novel data that has the potential to contribute to our understanding of the formation of intergroup attitudes. First, understanding the origins of prejudice and bias is important for interventions. By adulthood, intergroup biases are deeply entrenched; therefore, childhood is the time for intervention (see Killen, Rutland, & Ruck, 2011). Second, extensive research has shown the powerful influence of specific-peer-group norms on children's development (e.g., Nesdale & Lawson, 2011). Third, children's daily lives, unlike the lives of adults, are more constant and organized by both large (e.g., school) and local (e.g., clubs) groups that are dominated by peers and authority sanctioned. Therefore, we investigated the role of different levels of group norms in the formation of outgroup preferences in the current experiment.
Specifically, the present study systematically examined the role that both specific-peer-group norms and larger societal generic norms play in the manifestation of outgroup and in-group preference from middle childhood into adolescence (i.e., among 9-to 16-year-olds). We examined these preferences within the context of schoolgroup membership because this group identity is central to adolescents' lives (Eccles & Roeser, 2013). Further, previous research using a school context revealed that children who judged an out-group deviant were more favorable than children who judged an in-group deviant (e.g., Abrams et al., 2003).
We used a within-participants design to test whether this preference persisted when individuals judged both an in-group and an out-group deviant. We tested the hypothesis that this favoritism depends on whether the individual who deviates from a group norm simultaneously conforms or does not conform to a generic normone that applies to both the in-group and the out-group. Thus, the current study fills a key gap in our current understanding of the role that group norms play in the manifestation of in-group bias.
Uniquely, in the present study, we varied the type of deviance shown by an in-group and an out-group deviant peer. The deviants opposed their group's norm in favor of the other groups' norm and concurrently either conformed or did not conform to a generic norm. Specifically, we told participants that it was a tradition (i.e., a larger societal expectation or generic norm) that children wear shirts for school-affiliated clubs to community-wide events for all the clubs. Critically, though, some groups adhered to this generic norm whereas some violated it. Thus, we presented participants with in-and outgroup deviants who were either "traditional" (wearing the club shirt even though other in-group members did not) or "nontraditional" (not wearing the club shirt even though other in-group members did).
We expected that participants would show a preference for out-group over in-group deviance only when the deviant conformed to the traditional norm. This deviance from tradition by the out-group member would align not only with the in-group norm but also with the generic norm, whereas such deviance by an in-group member would be a fundamental threat to the group's identity because it would not only violate the group norm but also support a generic norm shared with the out-group. Endorsement of this norm would represent a challenge to the distinctiveness of the in-group relative to the out-group (Tajfel, 1978). In contrast, favoritism toward out-group over in-group deviance should not be present when the deviant also dissents from the traditional norm because this type of behavior infers low status upon all individuals within schools (Eccles & Roeser, 2013;Turiel, 1983). Therefore, in this case, both in-group and outgroup deviants should be equally disliked.
The present study also examined the social reasoning individuals use to justify deviance. Social reasoning developmental theory asserts that reasoning about deviance provides important information about why deviance within groups is condemned or tolerated (see Rutland et al., 2010). Similar to Kuhn's (1991) research on argumentation and Saxe's neuroscience research on individuals' interpretations of social interactions (Saxe, Whitfield-Gabrieli, Scholz, & Pelphrey, 2009), our research involved using reasoning analyses to examine the conceptual interpretations of events posed to participants regarding third-party behavior (in contrast to examining their post hoc justifications of their own behavior). Research has indicated that from childhood into adolescence, adherence to social-conventional traditions is viewed as a matter of autonomy (Horn, 2003) or as a matter of group loyalty or group functioning (Horn, 2003;Killen et al., 2013). We therefore expected our participants to use autonomy reasoning to justify their tolerance of deviance, and reasoning based on group functioning (i.e., maintaining group cohesion) or group loyalty (i.e., being faithful to others) to denounce deviance.
Finally, we were interested in whether children's sensitivity to group norms and the pressure to conform (Nesdale, Maass, Durkin, & Griffiths, 2005;Rutland, Cameron, Milne, & McGeorge, 2005) becomes stronger into adolescence. We therefore investigated age-related differences, from middle childhood into adolescence, in the relationship between individuals' own evaluations of deviance and their expectations about how their in-group peers would evaluate deviance. We anticipated that the importance of group norms and concerns about peergroup rejection would only increase into adolescence (Killen et al., 2013;Somerville, 2013). We expected that, when out-group deviance was preferred to in-group deviance, the relationship between the perceived ingroup norm and participants' own evaluations of the deviant act would be significantly stronger with age.

Design and procedure
The study used a mixed Age (4th grade, 8th grade, 10th grade) × Group (in-group, out-group) × Type of Deviance (traditional, nontraditional; within-subjects) design. To establish group membership, we told participants that they belonged to a mixed-gender peer club at their actual school (i.e., the in-group). They were also shown an illustration (see Fig. S1 in the Supplemental Material available online) of a mixed-gender group of children in a peer club from a familiar school in the participant's school district that often served as a competitive context for sports and teams (i.e., the out-group).
Next, all participants were told that the school district provided club shirts that students were expected to wear at district special events (e.g., assemblies) so that club members could be identified. This established a generic norm (i.e., a tradition), shared by the in-group and the out-group, of wearing a club shirt to special events. Then participants were introduced to their in-group norm and the out-group norm, which either matched (i.e., wearing a club shirt) or were counter to (i.e., not wearing club shirt) the tradition.
Two deviants from the in-group and the out-group club, respectively, were then portrayed as challenging their group norm and following the norm of the other group. The type of deviance shown by these two peers differed. The traditional deviant was the group member who conformed to the traditional norm by wearing the club shirt but deviated from the group norm of not wearing the club shirt to a district event. The nontraditional deviant, by not wearing the club shirt to the event, did not conform to the traditional norm or the in-group norm. Table 1 displays the four types of deviant behaviors within our design.

Measures and analysis
For each deviant behavior, participants completed four dependent measures. For the first item, they evaluated the deviant act as "okay" or "not okay" (i.e., "Do you think X (deviant peer) was okay or not okay to do what he/she did?"). For the second item, they rated the deviant act (i.e., "How okay or not okay do you think X was for doing what he/she did?"), using a 6-point scale (1 = really not okay; 6 = really okay). The third item was a measure of participants' reasoning behind (i.e., justification for) their evaluation of the deviant act (i.e., "Why?"). For the fourth item, participants indicated the perceived in-group norm about evaluating the deviant peer by rating the group's favorability toward that peer (i.e., "How do you think the group feels about having X in the group?"), using a 6-point scale (1 = very bad; 6 = very good).
Responses to the reasoning question were coded using coding categories drawn from social domain theory (Smetana, 2013;Turiel, 2006), previous research (e.g. Killen et al., 2013), and a content review of pilot data. As in previous research on social exclusion, participants referenced social-conventional reasoning involving notions of group loyalty and group functioning, as well as personal reasoning related to individual choice and autonomy (Killen et al., 2013). The coding system assigned each response to one of three categories: group loyalty (e.g., "He didn't show commitment to us"), group functioning (e.g., "He will upset things because he's going against what the group wants"), or autonomy (e.g., "It's okay for him to be different; he can do what he wants to do").
Fewer than 5% of the participants used two codes. Justification responses were analyzed using an established data-analytic procedure (see Killen et al., 2013): Each justification was coded as 1 for full use of the category, .5 for partial use of the category, or 0 for no use of the category. Coding was conducted by three coders blind to the hypotheses of the study. An assessment of 25% of the interviews (n = 116) demonstrated appropriate interrater reliability, Cohen's κ = .86. Analyses of variance (ANOVAs) were used to analyze proportions, given that they are robust to the problem of empty cells compared to other data analytic procedures (e.g., log-linear models; see Wainryb, Shaw, Laupa, & Smith, 2001).

Results
Initial analyses examined gender differences, but because no results were significant, gender was dropped from the analyses.

Rating of the deviant act
A 3 (grade: 4th, 8th, 10th) × 2 (group: in-group, outgroup) × 2 (type of deviance: traditional, nontraditional) ANOVA with repeated measures on the last factor revealed main effects of group, F(1, 455) = 4.21, p = .041, η p 2 = .01, and type of deviance, F(1, 455) = 125.66, p = .001, η p 2 = .22. There was also, as anticipated, a significant interaction between group and type of deviance, F(1, 455) = 5.66, p = .018, η p 2 = .01. A simple main-effects analysis showed that participants rated the traditional deviant act (i.e., wearing the shirt when the group did not) by the out-group peer (M = 4.60, SD = 1.49) as more acceptable than the same act by the in-group peer (M = 4.19, SD = 1.48), t(461) = 2.92, p < .01, d = 0.28. In contrast, participants rated the nontraditional deviant act (i.e., not wearing the shirt when the group did) by the in-group peer (M = 3.24, SD = 1.42) and the out-group peer (M = 3.29, SD = 1.47) as equally unacceptable, t(460) = −.37, p = .71. Participants thought an out-group deviant act was more tolerable than an in-group deviant act only when this act was in line with the generic school norm (i.e., traditional). We also conducted t tests against the midpoint value of 3.5. These results revealed that both the out-group and in-group traditional deviant acts were significantly above the midpoint-in-group: t(227) = Does not wear the club shirt when the ingroup does 7.230, p < .001; out-group: t(234) = 11.237, p < .001-and that both the out-group and in-group nontraditional deviant acts were significantly below the midpoint-in-group: t(233) = −2.758, p = .006; out-group: t(227) = −2.120, p = .035. These findings are shown in Figure 1.

Relationship between perceived group norm and rating of the deviant act
We expected that when the deviant conformed to the traditional norm, the relationship between the perceived in-group norm and the participant's own rating of the deviant act would be significantly stronger among the older compared to younger individuals. To test our hypothesis, in IBM SPSS Statistics (Version19; IBM, Armonk, NY) using bootstrapping, we entered the centered continuous variables for age and perceived group norm together with their interaction term hierarchically to predict the participants' ratings of the traditional and nontraditional deviant acts.
Perceived group norm was a significant predictor of ratings of the traditional deviant act (β = 0.28, t = 6.47, p = .001), R 2 = .10, F(3, 462) = 16.83, p < .001. Participants' rating of the act was more positive the more they perceived their group to be favorable toward the traditional-deviant peer. As expected, there was also a significant interaction between age and perceived group norm (β = 0.11, t = 2.49, p = .013). Simple-slopes analyses revealed that the relationship between participants' perception of their group's norm and their own rating of the traditional deviant act was significantly stronger among the older (t = 6.25, p = .001) compared to the younger (t = 2.36, p = .032) participants. This effect is shown in Figure 3. Perceived group norm was also a significant predictor of ratings of the nontraditional deviant act (β = 0.33, t = 7.51, p < .001), R 2 = .10, F(3, 461) = 19.35, p < .001, but, as expected, there was no significant interaction between age and perceived group norm (β = 0.02, t = 0.48, p = .63).

Discussion
In this study we showed, for the first time, how different levels of group norms are related to the expression of out-group preference among children and adolescents.
In the same experiment, we showed that the presence or absence of an out-group over in-group preference depended on how the deviance matched the group and generic norm. These findings reveal how deviation from group norms can engender either in-group bias or outgroup favoritism, helping to better our understanding of the manifestation of out-group preference.
Overall, the participants rated the deviant peers who conformed to the generic norm more positively than those who did not support this norm. However, as expected, a preference for the out-group deviant over the in-group deviant was evident only when both deviants conformed to the generic norm. This type of deviance was unique in that the out-group member acted in alignment with the participants' own group norm and with the generic (i.e., . For older and younger participants and higher and lower group-norm scores, we substituted values 1 standard deviation above and below the means, respectively. traditional) norm. Conversely, by rejecting the participants' own group norm and supporting the generic norm (i.e., to follow tradition), the in-group deviant threatened the distinctiveness of the in-group and its group identity (Tajfel, 1978). These findings are also compatible with psychological theories that emphasize the importance of conventions and stereotypes in perpetuating the status quo (e.g., Jost et al., 2004;Rutland et al., 2010;Turiel, 1983). Conventions and traditions typically have high status because they are culturally sanctioned behavior and are indicators of social hierarchy. Our participants were arguably signaling their broad alignment with convention and the status quo (the "system") by favoring the out-group deviant only when that deviant acted in line with tradition and supported the high-status form of behavior.
Our findings concerning reasoning support this argument, because participants mostly used group-functioning reasoning (i.e., relating to the need to maintain group cohesion and effectiveness) to justify their disapproval of deviants from tradition who conformed to the generic norm. In contrast, participants who justified their support for these same deviants typically used autonomy reasoning (i.e., relating to the need for personal autonomy and choice). A minority of participants used group-functioning reasoning to signify their approval for the traditional deviant. In the case of the out-group traditional deviant, participants did not perceive this deviant as an out-group member but rather as a member of a common in-group that conformed to the generic tradition. There was also a developmental trend whereby older participants increasingly used reasoning based on group loyalty (i.e., being faithful to others) rather than reasoning based on group functioning to disapprove of deviants who dissented from the generic norm.
Finally, with age, individuals increasingly used their beliefs about how their own group would evaluate deviants who behaved traditionally (i.e., the perceived ingroup norm) to inform their evaluations of those deviants. Consistent with previous research in middle to late childhood (FitzRoy & Rutland, 2010;Nesdale & Lawson, 2011;Rutland, 2013;Rutland et al., 2005), this study suggests that sensitivity to group norms and concerns about social exclusion get stronger into adolescence. This developing sensitivity to group norms is likely due to the increasing role of the peer group in adolescence and the increasing importance of ensuring group functioning with age (Brechwald & Prinstein, 2011;Horn, 2003).
Making children more aware of group dynamics, the pressures to conform, and the legitimacy of resisting conformity can help reduce the potential development of prejudice driven by in-group bias. This cannot be achieved simply through legislation or political dictate; rather, it requires educational interventions that focus on changing social relationships (i.e., child-child or childadult interactions) and children's social-cognitive skills (i.e., social perspective taking and moral reasoning; see . We have demonstrated a context in which the in-group is not preferred, with a deviant from the out-group who conformed to a generic norm being preferred to an equivalent in-group deviant. We do not suggest that in-group bias is absent in childhood; indeed, research has shown that in-group bias is a general phenomenon in childhood (Dunham, Chen, & Banaji, 2013). On the contrary, we propose that an often overlooked factor that determines the emergence of in-group bias has to do with group norms and, specifically, how the motivation to conform to these norms can contribute to displays of in-group bias or out-group favoritism. The current study provides novel data on the early ontogenesis in childhood of sensitivity to conditions that lead to out-group preference. Given that prejudice is the opposite of out-group preference, determining the contexts in which individuals display outgroup favoritism provides a window into opportunities for reducing prejudice in adulthood.

Author Contributions
A. Rutland, M. Killen, and D. Abrams developed the study concept. All authors contributed to the study design. Testing and data collection were performed by K. L. Mulvey and A. Hitti. A. Rutland, K. L. Mulvey, A. Hitti, and M. Killen performed the data analysis and interpretation. A. Rutland drafted the manuscript, and all authors provided critical revisions. All authors approved the final version of the manuscript for submission.