6 Feinberg et al. participants'negative-deviation scores and likelihood of for further examination,see Antisocial Punishment in the being ostracized was explained by reputation strength Supplemental Material). We found that negative- Additional Analyses strength scores predicted whether or not an individual would be ostracized.When both the negative-deviation Earnings and reputation-strength variables were entered as simul ANhouehcopeaioaaesncreaednteenpath produced the greatest ove good.Wher between relative selfishness and ostracism was mediated someone is ostracized,fewer individuals can contribute by the reputation-strength variable,Sobel zs >3.32.ps< to the public good.Thus,ostracism may have been sub whic means the more Ifishly individuals had optima .the more neh from an eamings perspec their upcoming-round partners would ostracize them. the Supplemental Material for further details)and explored differences in earings for each game.As shown Ostracized individuals bebave more coopera tively.In th gossip-w p-wi te t a smalle than half of their points in the finst round 2 However,the gossip-with-ostracism game produced the 1.52)increased substantially thereafter.By Round 6,the highes t individual-level earnings by Round 5,omnibus 2 30,pairwise I S> sen to 547<.0 d 6 to increase their contributions An examination of ostra nibus)>ming pairwise (159)s 2.41.ps .05.ds >0.38.Further.an cized individuals'contributions in the round before and examination of eamings in the gossip-with-ostracism game the round after being ostracized revealed that these pa rom secon heir contrib y an averag amings level h the game concluded. an average of 0.26 points over the same period (see Ostracized Individuals Versus Nonostracized Individual Pro.social gossi de Mate rch has ued that Further when participants returned to plaving afte In line with being ostracized,their contribution amounts were often his arch logistic regression analyses showed that in gossip-wi exclusion in Round 2:83)=2.63..01 n.the more likely they were to engage in gossip. 0.39;Round 4 after exclusion in Round 3:/(178)= 56. 0.26,ps<05,exp(b)s>1.29.Given that it was through p =.58,d 0.08;Round 5 after exclusion in Round 4: gossip tha reputationa was .09, =025: soc instigated and maintained much of the cooperation participants to behave less selfishly and thereby conform that occurred during the gossip-with-ostracism game. to the more cooperative behavior of the rest of the se who contributed the mos General Discussion a group id tha Overall.these results advance the understanding of how (Herrmann.Thoni,&Gachter.2008:Parks stone.2010: gossip promotes prosocial behavior.Although past 3.2016 Feinberg et al. participants’ negative-deviation scores and likelihood of being ostracized was explained by reputation strength. We found that negative-deviation scores significantly pre￾dicted reputation strength in all six rounds, βs > 0.35, ps < .001, R2 s > .12. As reported earlier, reputation strength scores predicted whether or not an individual would be ostracized. When both the negative-deviation and reputation-strength variables were entered as simul￾taneous predictors in a logistic regression analysis, the results revealed that the relationship across rounds between relative selfishness and ostracism was mediated by the reputation-strength variable, Sobel zs > 3.32, ps < .001, which means the more selfishly individuals had behaved in a previous round, the more negative their reputation would be, resulting in a greater chance that their upcoming-round partners would ostracize them. Ostracized individuals behave more coopera￾tively. In the gossip-with-ostracism game, contribution levels among participants who had contributed fewer than half of their points in the first round (M = 2.67, SD = 1.52) increased substantially thereafter. By Round 6, the average number of points these participants contributed had risen to 7.84 (SD = 2.77). Ostracism played a signifi￾cant role in compelling these more egoistic participants to increase their contributions. An examination of ostra￾cized individuals’ contributions in the round before and the round after being ostracized revealed that these par￾ticipants increased their contribution by an average of 2.86 points, Fs > 5.00, ps < .05, η2 s > .27, whereas those who were not ostracized increased their contribution by an average of 0.26 points over the same period (see Ostracized Individuals Versus Nonostracized Individuals Round-to-Round Changes in the Supplemental Material for additional details, including degrees of freedom for F-test results). Further, when participants returned to playing after being ostracized, their contribution amounts were often not significantly different from those of participants who had not been excluded in the previous round—Round 3 after exclusion in Round 2: t(183) = 2.63, p < .01, d = 0.39; Round 4 after exclusion in Round 3: t(178) = .56, p = .58, d = 0.08; Round 5 after exclusion in Round 4: t(180) = 1.71, p = .09, d = 0.25; Round 6 after exclusion in Round 5: t(192) = 1.94, p = .053, d = 0.28. Such a result suggests that ostracism compelled these uncooperative participants to behave less selfishly and thereby conform to the more cooperative behavior of the rest of the group. Of further note, those who contributed the most in a group were never ostracized, which suggests that participants did not engage in antisocial punishment (Herrmann, Thöni, & Gächter, 2008; Parks & Stone, 2010; for further examination, see Antisocial Punishment in the Supplemental Material). Additional Analyses Earnings Although cooperation rates increased in the gossip-with￾ostracism game, this does not necessarily mean that the game produced the greatest overall public good. When someone is ostracized, fewer individuals can contribute to the public good. Thus, ostracism may have been sub￾optimal from an earnings perspective. To examine this possibility, we calculated earnings both at the group level and the individual level (see Calculating Earnings Data in the Supplemental Material for further details) and explored differences in earnings for each game. As shown in Figure 3, earnings suffered early in the gossip-with￾ostracism game because ostracism resulted in a smaller resource pool and smaller multiplier. However, the gossip-with-ostracism game produced the highest individual-level earnings by Round 5, omnibus F(2, 159) = 45.47, p < .001, η2 = .36, pairwise t(159)s > 4.36, ps < .001, ds > 0.69, and the highest group-level earnings by Round 6, omnibus F(2, 159) > 22.71, p < .001, η2 = .22, pairwise t(159)s > 2.41, ps < .05, ds > 0.38. Further, an examination of earnings in the gossip-with-ostracism game from the second round onward revealed a significant upward linear trend, Fs(1, 53) > 42.32, ps < .001, η2 s > .44, which shows that earnings levels were increasing when the game concluded. Prosocial gossip Recent research has argued that gossiping about selfish individuals is a prosocial act that more prosocial indi￾viduals engage in (Feinberg et al., 2012). In line with this research, logistic regression analyses showed that in five of the six rounds of the gossip-with-ostracism game, the more individuals positively deviated from their group mean, the more likely they were to engage in gossip, bs > 0.26, ps < .05, exp(b)s > 1.29. Given that it was through gossip that reputational information was transferred, leading to the exclusion of those who behaved more self￾ishly, this result suggests that those who were more pro￾social instigated and maintained much of the cooperation that occurred during the gossip-with-ostracism game. General Discussion Overall, these results advance the understanding of how gossip promotes prosocial behavior. Although past Downloaded from pss.sagepub.com by Cai Xing on February 13, 2014