Group membership is a social dimension that distinguishes oneself from others, including friendship, race, class, nationality, and even mere membership triggered by artificial cues (Lieberman and Linke, 2007; Chierchia et al., 2020; Zhang et al., 2021, 2022). Researchers have examined the effect of group membership on TPP, but their findings have been inconsistent. Two competing hypotheses, the Mere Preferences Hypothesis and the Norms Focused Hypothesis (McAuliffe and Dunham, 2016; Zhang et al., 2020), were developed to explain the contradictory results. The Mere Preferences Hypothesis suggests that individuals’ positive evaluation toward the ingroup would reduce TPP for ingroup perpetrators, supported by the majority of evidence based on adults (Yudkin et al., 2016; McAuliffe et al., 2017; Guo et al., 2020, 2022; Yang et al., 2023) and children (Jordan et al., 2014), which supports the ingroup favoritism phenomenon (IGF). The Norms Focused Hypothesis emphasizes that individuals’ concern for maintaining norms of group cooperation would enhance TPP for ingroup violators, as demonstrated by some evidence based on adults (Mendoza et al., 2014; Delton and Krasnow, 2017) and children (Gonzalez-Gadea et al., 2022), known as the black sheep effect (BSE). Even though they differ in the direction of the effect, IGF and BSE are two ways for people to maintain the group identity and cohesion (Zhang et al., 2020). However, various systematic reviews and meta-analyses have found that children, adolescents, and adults are more likely to punish outgroup offenders than ingroup criminals (McAuliffe and Dunham, 2016; Lazić et al., 2021). Therefore, group membership can influence adolescents’ TPP, showing that youth punish outgroup members more harshly than in-group members (Hypothesis 1).

Perceived unfairness is one potential explanation for the proposed effect of group membership on TPP (Lu and McKeown, 2018). Fairness preference theory suggests that humans have a strong disgust for inequality and are willing to consume resources to punish offenders when they experience or witness injustices (Fehr and Schmidt, 1999). Some studies have shown that adolescents have strong equity concerns and a high willingness to sacrifice their personal interests to uphold fairness norms; thus, perceived unfairness has become an important driving force for implementing punishment (Güth and Kocher, 2014; Lu and McKeown, 2018). Nevertheless, the perception of injustice is not invariable, and it will depend on the group relationship of both sides. Firstly, individuals’ perception of injustice is less prominent when unfair proposals are made by ingroups than by outgroups (Lu and McKeown, 2018). In addition, perceived unfairness is associated with TPP, such that the greater the perceived unfairness, the more motivated people are to punish (Fehr and Gächter, 2002). Finally, self-reported justice perception mediates the relationship between partners’ social distance (i.e., human vs. computer partner) and rejection behavior among healthy adults and patients with major depressive disorders (Wang and Li, 2013; Jin et al., 2022). Therefore, we proposed that group membership may be associated with more TPP for outgroup members via increased perceived unfairness (Hypothesis 2).

According to negative emotion theory, perceiving negative emotions such as anger, frustration, and disgust that arise from behavior violations can form the basis for punishing behaviors, triggering a desire to punish others in response to real-life immorality (Hartsough et al., 2020). Self-reported anger has been suggested as a possible motivation for TPP in some studies (Fehr and Fischbacher, 2004; Gummerum et al., 2016) and could mediate the association between unfair offers and TPP in adults (Gummerum et al., 2020). Additionally, the harshness of TPP increased significantly when anger was induced but decreased when anger was inhibited (Gummerum et al., 2022). More importantly, the experience of anger caused by injustice differs depending on the peer group to which one belongs. Bicskei et al. (2016) found that the same unkind behavior by outgroups was associated with greater anger-like emotions than that of ingroups, and punishment behavior was strongly influenced by anger-like emotions. Finally, Wang and Li (2013) found that self-reported feelings of anger could mediate the association between social relations (i.e., friend, teacher, and stranger) and adults’ punishment in the ultimatum game. Hence, we proposed that anger played a crucial role in the relationship between group membership and TPP (Hypothesis 3).

An evaluation-emotional-behavioral model was employed by Seip et al. (2014) to explain the mechanism underlying costly punishments of unfairness: evaluating an action or event as unjust leads to anger toward the offender, which can then drive people to punish those who violate social norms, even if punishment comes at a price. Several studies have indicated that violations of fairness can lead to perceived unfairness, which leads to anger and, ultimately motivates punishment by second and third parties (Singer and Steinbeis, 2009; Mendoza et al., 2014). For example, Mendoza et al. (2014) demonstrated that increasingly unfair offers predicted lower perceived fairness, thereby resulting in a strengthened level of anger and ultimately prompting individuals to reject the unfair offers in ultimatum game. Accordingly, we proposed that perceived unfairness and anger can exert a chain-mediating effect between group membership and TPP (Hypothesis 4).

Social role theory emphasizes that different societal stereotypes are assigned to boys and men as compared to girls and women, with girls expected to be more communal and caring and boys are expected to be agentic and dominant (Eagly, 2009). These gender role beliefs greatly influence boys’ and girls’ perceptions, emotional experiences, and behavioral responses to norm-violating behavior (Chawla et al., 2020). Laboratory and field studies suggest that boys are more likely to judge private behavior negatively, experience greater anger, and punish offenders more severely when they experience normal transgressions than girls (Kromer and Bahçekapili, 2010; Bonini et al., 2011; Balafoutas and Nikiforakis, 2012; Rodriguez-Ruiz et al., 2019). As a result of this socialization and other forces, boys and men tend to be more socially dominant than girls and women (Pratto et al., 1994; Du et al., 2021), and norms for masculinity are more rigid than norms for femininity (Koenig, 2018). Accordingly, boys show greater ingroup favoritism both cognitively, emotionally and behaviorally than girls during intergroup interactions. For example, boys exhibit greater ingroup favoritism than girls when responding to unfair distributions (Wu and Gao, 2018). Thus, we proposed that the chain mediation of perceived unfairness and anger was more pronounced in boys than in girls (Hypothesis 5).

The current study adopted a complete between-subject design of 2 (group membership: classmate vs. stranger) × 2 (gender: boy, girl). Based on an a priori power analysis, the sample size was estimated using G*Power 3.1 (Faul et al., 2007). F tests and ANOVA (fixed effects, special effects, main effects, and interactions) in G*Power (version 3.1.9.7) were selected. To detect a medium effect (f² = 0.25), N = 128 participants (32 participants per group) with 0.80 power and 0.05 Types I error rates were needed. Experimental data were collected from two junior high schools in Henan Province, China. The distribution and collection of situational questionnaires were conducted by a trained research assistant with standardized processes for completing the questionnaires. During the study, eight classes of seventh- and eighth-graders were randomly selected. Four classes were randomly assigned to the classmate condition, and the other four classes were assigned to the stranger condition.

A total of 350 questionnaires were distributed in the form of class tests. After removing missing values or other ineffective responses, the final data set consisted of 306 questionnaires, with a minimum of 49 respondents for each condition. The sample included 175 boy students (57.19%) and 131 girl students (42.81%) between the ages of 12 and 15. Their average age was 13.46 ± 0.75 years, with 69.99% in seventh grade and 33.01% in eighth grade. All subjects self-reported no mental or psychological disorders and gave their oral informed consent. Ethics committee approval was obtained from the Faculty of Education at Henan Normal University, and protocol adherence to the Declaration of Helsinki was ensured.

Students were instructed to complete a pen-and-paper test on TPP in the classroom as a class. There were four main sections of the assessment, including basic personal information, third-party punishment tasks, group membership manipulation and check, and self-report assessment. These four sections were always administered in the same order as below.

A 2 (group membership: classmate vs. stranger) × 2 (gender: boy vs. girl) ANOVA on the IOS scale scores showed that only the main effect of group membership was significant, F(1,302) = 588.52, p < 0.01, partial η² = 0.66. Identification with a classmate was larger (M = 3.50, SE = 0.07) than identification with stranger (M = 1.42, SE = 0.06), see Figure 1A. This finding indicated that the manipulation of group membership was successful.

A 2 (group membership: classmate vs. stranger) × 2 (gender: boy vs. girl) MANOVA on perceived unfairness, anger, and TPP found that the main effects of group membership were significant, Fs(1,302) > 19.75, ps < 0.01, partial η²s > 0.06. As compared to strangers’ selfish behavior, participants perceived unfair allocation from classmates as less unfair, experienced less anger, and punished less severely. Moreover, the interactions by group membership and gender were also significant, Fs(1,302) > 6.13, ps < 0.05, partial η²s > 0.02. Further analysis shown that girls’ TPP was influenced by group membership, F(1,302) = 7.67, p < 0.01, but not by their perceptions of unfairness and anger, Fs(1,302) < 1.82, ps > 0.05. In particular, girls punished classmates (M = 2.08, SE = 0.12) less severely than stranger (M = 2.51, SE = 0.10). In contrast, boys’ perceptions of unfairness, anger, and TPP were affected by group membership, Fs(1,302) > 29.23, ps < 0.01, showing that boys perceive classmate’ transgressions as less unfair, experience less anger, and impose softer punishments comparing to stranger’s transgressions (see Figures 1B–D). The main effects of gender were not significant, Fs(1,302) < 2.69, ps > 0.05.

The descriptive statistics and correlations of the variables are reported in Table 1. Dummy codes were used for group membership, with ingroup coded as 0 and outgroup coded as 1. Group membership was significantly positively associated with perceived unfairness, anger, and TPP (r = 0.28, 0.28, 0.38, ps < 0.01), supporting Hypothesis 1. Unfair perception was significantly positively associated with anger and TPP (r = 0.68, 0.67, ps < 0.01). Anger was significantly positively correlated with TPP (r = 0.67, p < 0.01).

A moderated chain mediation model was conducted by using Model 85 in the Process 4.0 macro of SPSS 26.0. Dummy codes were used for gender and group membership, with girl and ingroup coded as 0 while boy and outgroup coded as 1. Confounding effects were reduced by including age and grade as control variables. The results showed that group membership could significantly positively predict unfair perception, anger and TPP (β = 0.27, 0.09, 0.18, ps < 0.01); unfair perception could significantly positively predict anger and TPP (β = 0.67, 0.36, ps < 0.01); and anger could significantly positively predict TPP (β = 0.37, p < 0.01). Thus, Hypothesis 2, 3 and 4 were supported. The interaction between group membership and gender had a significant effect on perceived unfairness (β = 0.14, p < 0.05). In contrast, the interaction between group membership and gender had no effect on anger (β = 0.04, p > 0.05) and TPP (β = 0.03, p > 0.05). Thus, Hypothesis 5 was partially supported (see Table 2).

A slope test was conducted to clarify the mechanisms by which group membership and gender interact with perceived unfairness. The result showed that group membership significantly positively predict the boys’ perceived unfairness (simple slope = 0.39, t = 5.47, p < 0.01), but cannot predict girls’ perceived unfairness (simple slope = 0.11, t = 1.23, p > 0.05) (see Figure 2A). The figures of the chain mediation model separately for boys and girls were shown in Figures 2B,C.