DUGUID AND GONCALO we examined the conditional indirect effectsat the 47069,D= d low tatus (M 451.32.SD =280.40,44) 2.40.p condition. 3.39.p<.001 but not in the no-evaluation ss whether loss mediated the inter anc wasnot being d ther and high-status (M=437.09.SD=230.33).(6) -0.34.p 2009 Preacher&Hayes.004).Based mples the ow-status (M41.68. 24.91.F1.45) 95%CI ranged between 16.93 and 59.98.excluding zero. Discussion ANOVA O hat individu th mid and and ically.RT 2.03. -statu no significant diffe e in th indivTeatedwords sen by those with high or 9u such that when y thought the g evaluate anding of why middl when they thoughhe 0.94,SD=0.94 eon tasks that demand creative solutions,occupying the middle ted,middl-statusnd ve tctendtcdourpweiousfining.byinvcstigatiagaakin words thar atus (M =0.92.SD=1.04 rienced by middle be a off i =0.79.SD ns of tas On the on han -430 nce in th relate n their creativity ing 2011:Camach and,this focus may y actually improve perf on tasks,lik 279 72 individua d status part nts,45)=183,p .08 Study 4 est the hypothe s th the lassical 63n 376 However,ther vas an inter status and examining the impact of being in the middle versus the top an evaluated middle-status individuals aster than the RTs of those with high middlle.r betw status ted.the differences in the rTs of individuals with middle high. low status ere of stat nlike status,which is social etween status and evaluation on RT was no longer significant interaction between status and evaluation was significant, F(2, 136) 3.15, p .046, p 0.04. Planned contrasts showed that when performance was being evaluated, middle-status partici￾pants’ RTs (M 270.63, SD 223.89) were faster than those of high-status (M 470.69, SD 234.59), t(45) 2.98, p .005, and low-status (M 451.32, SD 280.40), t(44) 2.40, p .021, participants. There was no significant difference in RTs of high- and low-status participants, t(47) 0.26, p .794. When performance was not being evaluated, the differences in the RTs of middle-status participants (M 463.73, SD 312.50) and high-status (M 437.09, SD 230.33), t(46) 0.34, p .738, and low-status (M 413.68, SD 224.91), F(1, 45) 0.63, p .533, participants were not significant. There was no signifi￾cant difference in the RTs of high- and low-status participants, t(45) 0.35, p .726. Threat of status loss. ANOVA on the composite threat of status loss measure revealed a main effect of status, F(2, 136) 8.15, p  .001. Middle-status individuals (M 1.67, SD 1.74) identified more threat-of-status-loss-related words than high-status (M 1.08, SD 1.04), t(93) 2.03, p .045, or low-status (M 0.79, SD 0.75), t(86) 3.20, p .002, individuals. There was no significant difference in threat-of-status-loss-related words chosen by those with high or low status, t(94) 1.59, p .116. There was also a main effect of being evaluated, F(1, 136) 7.18, p .008, such that when they thought they were being evaluated (M 1.41, SD 1.53), individuals identified more threat-related words than when they thought they were not being evaluated (M 0.94, SD 0.94). Importantly, there was also a significant interaction of status and evaluation, F(2, 136) 12.57, p .001. Planned contrasts showed that when performance was being evaluated, middle-status indi￾viduals (M 2.64, SD 1.87) identified more threat-of-status￾loss-related words than high-status (M 0.92, SD 1.04), t(45) 3.96, p  .001, or low-status (M 0.79, SD 0.83), t(44) 4.39, p .001, individuals. There was no significant difference in threat-of-status-loss-related words identified by high￾and low-status individuals, t(47) 0.48, p .636. When there was no evaluation, the differences on the threat￾related words identified by middle-status individuals (M 1.25, SD 1.03) and high-status (M 0.79, SD 1.02), t(46) 1.55, p .129, and low-status (M 0.78, SD 0.67), t(45) 0.04, p .972, individuals were not significant. There was also no significant difference in threat-related words identified by high￾and low-status participants, t(45) 1.83, p .080. Mediation analysis. To test the hypothesis that threat of status loss accounts for the relationship between status and evaluation on convergent thinking, we conducted a mediation analysis. As es￾tablished above, there was no main effect of status,  0.14, t 1.63, p .106, or evaluation,  0.07, t 0.89, p .376. However, there was an interaction between status and evaluation on convergent thinking,  0.20, t 2.29, p .024, such that when performance was being evaluated, middle-status individuals’ RTs were faster than the RTs of those with high- and low-status participants. However, when performance was not being evalu￾ated, the differences in the RTs of individuals with middle, high, and low status were not significant. There was a significant relationship between threat of status loss and RT, 0.43, t 5.70, p  .001. The interaction between status and evaluation on RT was no longer significant when threat of status loss was entered into the regression, 0.08, t 0.35, p .700. To further interpret these findings, we examined the conditional indirect effects at the levels of status. These effects showed that threat of status loss mediated the effect of evaluation on RT in the evaluation condition, z 3.39, p  .001 but not in the no-evaluation condition, z 1.36, p .295. To assess whether threat of status loss mediated the interac￾tive effects of status and evaluation on RT, we utilized boot￾strap estimates to generate bias-corrected 95% CIs (Hayes, 2009; Preacher & Hayes, 2004). Based on 1,000 samples, the 95% CI ranged between 16.93 and 59.98, excluding zero. Discussion As predicted, the results demonstrate that individuals with mid￾dle status performed better on the Stroop task, which requires convergent thought, than individuals with high and low status when they thought they were being evaluated. Specifically, RTs for correctly identifying the color of the words were significantly shorter for individuals with middle status compared with those with high status and low status. Furthermore, threat of status loss was shown to mediate the relationship between status and evalu￾ation and performance on the Stroop task. These findings provide a deeper understanding of why middle status might stifle creativity. Although we have shown, in Studies 1 and 2, that individuals with middle status will be at a disadvan￾tage on tasks that demand creative solutions, occupying the middle status position in the social hierarchy may not inevitably lead to error, as the early research assumed (Homans, 1961). In Study 3, we extended our previous findings by investigating a task in which performance might be boosted by threat of status loss. The anxiety experienced by middle-status individuals may be a trade-off in terms of task performance. On the one hand, the anxiety middle￾status individuals feel at the prospect of being evaluated negatively and potentially losing status may constrain their creativity by limiting their willingness to explore new solutions (Byron & Khazanchi, 2011; Camacho & Paulus, 1995). Yet, on the other hand, this focus may actually improve performance on tasks, like the Stroop test, that require convergent thought to complete quickly and accurately (Friedman & Förster, 2005; Peterson & Nemeth, 1996). Study 4 Drawing on the classical research on middle-status conformity, we have argued that middle-status individuals should be less creative than individuals with high or low status. Would the same hold true for power? Study 4 builds upon the previous findings by examining the impact of being in the middle versus the top and bottom of the power hierarchy on creativity (Galinsky et al., 2008); previous research focused on high and low power, excluding the middle. Though differences between power and status are rarely delineated, there is recent evidence to suggest that their conse￾quences are not necessarily identical (Blader & Chen, 2012). An important difference between status and power is that, unlike status, which is socially conferred, power is a property of the actor and is less susceptible to the subjective evaluations of 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. 8 DUGUID AND GONCALO