274 HEERDINK.VAN KLEEF.HOMAN.AND FISCHER happy conditions.on the other hand,was not significant.B peers.rather than the difference in self-reported expressed happi Mediation analvsis. The next step was to test whether the he findings regarding the relative influence of a group membe part influence in the with a confo H be exp ed by the extent to which the ipants had felt cject .the indirect the ignincanceofthisindirecteiecl was checked using bootstrap may have dy 2, 0,00 )O r own by I -0.24.95%BG 0.02L 04 rto that of the group,which we defined nity.bu n the m as one normally would for fe 05g (Cialini).Thus.we believ indings indicate that the feelin of rejection that we oled to ed by the lo e their outcomes( group performan and the Discussion by the underlying proces in the General On a m In this study,we replicated the basic findings from Studies 1-3 general level,these data again show that emotion ed b ed o d with ceptance a with ange th more reeco ants who paradigm to resolve this limitation. ngry reactions to conform more hat sho Study 5 should be noted that the n th sho individual to feel re ed.which inun leads him or her reported that they had shown more hanniness tha aim was to exten articipant hes ount of happiness.This nal experi the instr ipulat the much as the participants who had been instructed to show happ emotion expre ed by the majority (anger ther each other in such a situation feg Hess Bourseoi 01 dev d to show cmotion failed to er is seen a ss their tha il parti who had been having been instructed emotions s are safely bound within essed happinessis daries (Pickett Brewer.2005)and with regulating thei troduc and majority happy cond 007).Be e con sence of a difference in thea ount of happiness perceived by group member (cf. 960:Klein et al. happy conditions, on the other hand, was not significant,
0.04, t
0.26, p
.797. Mediation analysis. The next step was to test whether the focal participants’ lower influence in the majority angry condition, relative to the majority happy and nonemotional conditions, could be explained by the extent to which these participants had felt rejected. For this purpose, the indirect effect of the manipulation on the influence ratio through felt rejection was estimated, and the significance of this indirect effect was checked using bootstrap￾ping (see also Study 2, R
10,000 resamples). Our hypothesis— that the focal participants’ lower relative influence in the majority angry condition compared with the majority happy and nonemo￾tional conditions could be explained by felt rejection—was sup￾ported, B
0.24, 95% BCa CI [upper limit: 0.02], p
.04, one-tailed. The difference in the influence ratio between the ma￾jority happy and nonemotional conditions could not be explained by felt rejection, B
.02, 95% BCa CI [0.19; 0.05], p
.59. These findings indicate that the feelings of rejection that were experienced by the focal participants in the majority angry condi￾tion led to reduced influence in the group task, compared with focal participants in the other conditions. Discussion In this study, we replicated the basic findings from Studies 1–3 using an experimental group study involving real interaction. As in previous studies, group members who were faced with a majority that reacted with anger to their ideas felt more rejected than participants who were faced with a majority that did not react with anger. Feeling rejected, in turn, led participants who received angry reactions to conform more in a cooperative group task, compared with group members who were faced with a majority that showed happiness or no emotion. It should be noted that the majority happy condition produced mixed results. Participants who had received an instruction to show happiness reported that they had shown more happiness than participants who had not been instructed to show an emotion, whereas peer-reports indicated that these participants showed the same amount of happiness. This suggests that despite intending to follow the instructions, participants who received an instruction to express no emotions may have been unintentionally smiling as much as the participants who had been instructed to show happi￾ness. Because it can be expected that three students of similar age, who do not know each other in advance, tend to affiliate rather than reject each other in such a situation (e.g., Hess & Bourgeois, 2010; Hinsz & Tomhave, 1991), this is not an unlikely pattern of results. Thus, it may be the case that participants who had been instructed to show no emotion failed to suppress their happiness, rather than that participants who had been instructed to show happiness failed to do so. This tendency to smile for affiliative reasons, despite having been instructed to express no emotions, points to the difficulty of having a neutral or no emotion reference condition in naturalistic settings. Nevertheless, we still see these findings as being in line with the idea that expressed happiness is interpreted as a signal of affiliation (see the Introduction), as the absence of a difference in felt acceptance between the majority nonemotional and majority happy conditions paralleled the ab￾sence of a difference in the amount of happiness perceived by peers, rather than the difference in self-reported expressed happi￾ness. The findings regarding the relative influence of a group member who is facing an angry majority are consistent with a conformity interpretation. However, the design of the study, and especially our operationalization of influence, leaves some room for alternative explanations. For instance, an anonymous reviewer pointed to the possibility that focal participants in the majority angry condition may have also withdrawn from the group decision-making process. Because our participants were not able to physically withdraw from the situation (as in Study 2), they may have simply abided by the majority’s ideas instead of pushing their own ideas. Indeed, this passive strategy is different from the active matching of one’s behavior to that of the group, which we defined as conformity, but the failure to behave as one normally would for fear of negative reactions may also be seen as a special case of conformity: con￾formity by omission (Cialdini & Trost, 1998). Thus, we believe that the type of social influence that the majority’s expression of anger produces is that a deviant group member allows the majority to influence their outcomes (i.e., group performance and their chance of winning a prize), either through active or passive con￾formity. We return to the different types of social influence elicited by expressed emotions and the underlying processes in the General Discussion. On a more general level, these data again show that emotions expressed by a majority can influence the behavior of another group member and that feeling accepted or rejected is a likely explanation. Because we measured felt rejection/acceptance after the group interaction, this chain of causality could not be estab￾lished beyond doubt. In Study 5, we used a different experimental paradigm to resolve this limitation. Study 5 In the previous studies, we showed that a majority’s anger leads a deviant individual to feel rejected, which in turn leads him or her to feel pressure to conform. In Study 5, our aim was to extend these findings by directly testing the causal model using a para￾digm that afforded maximal experimental control and allowed a direct measurement of conformity. For this purpose, we set up an engaging computer-mediated cooperative interaction, in which we manipulated the emotion expressed by the majority (anger vs. happiness). Furthermore, we tested the moderating influence of yet another social-contextual factor that determines the extent to which conformity is meaningful behavior if one feels rejected due to expressions of anger by the majority: the prototypicality of the deviant individual. Prototypicality reflects the extent to which a member is seen as possessing features that are distinct characteristics of group mem￾berships (Hogg, 2005). Prototypical group members can be said to occupy central positions in the group, while their counterparts, peripheral members, are closer to the group boundaries. Prototyp￾ical members are safely bound within the group, whereas periph￾eral members are more concerned with the location of group boundaries (Pickett & Brewer, 2005) and with regulating their behavior to protect their group membership (e.g., Jetten et al., 2006; van Kleef, Steinel, van Knippenberg, Hogg, & Svensson, 2007). Because conformity is a way of showing that one is a good group member (cf. Hollander, 1960; Klein et al., 2007), we may This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 274 HEERDINK, VAN KLEEF, HOMAN, AND FISCHER