KRIZAN AND JOHAR There was no evidence for a three-way interaction between spicy e hot ve and conditi en togcther with the fact that t the pres vulnerability increased aggression equally against the original ulne nd vu ender did not interact with any the cation (B 73 0: main or inte d in E: that w the ggressive reta .and to toward Simplesop d that those high on vulnerable na eported inc 001)thandid B=43 29.p .02) ive re to pro tion both against the initial pro a hierarchica a third party Accord山ngly. blocks)on hot ility r ray interaction entered in suc dp四 205. 05)in addit effect of provocation (B -2.74.1 8.44 <001 no othe a effects were significant he ature of this interaction ontrol and the ons is indica d fo s stan ru dev and the ntrol and the are indicate igh on vulnerabl cived one and a half times les tha among those low on vulner bility only did th B 2.20.SE 44.p<.001) 13SE=1.05. Dep F vith the nare tic rage h on narc istic vulnerability increa ed by 43%a lepressi nly those high on v arc m responded with mor did not 51).0 e,the same ar alysis cond ted on the cific iter evealed act the sonal (i.e..di reacted to the cation with suggest they took it personally.Taken together,thes od d on:all ke signatures of narciss tic rage Contro Provocation ulation igure 2 choices (all ps>17).anger (all ps58).trustwo and high-v able respectively. aggression was examined as retaliation for a relatively minormaking sauce choices (which always indicated a strong dislike of spicy foods). In addition, those who chose the hot version of the sauce expected the coparticipant to dislike their choice even more, r  .27, p .001. Taken together with the fact that the presumed coparticipant was not initially provoked by the participant, and had a free choice between administering a mild (tea) and noxious (bitter gourd juice) food, these observations confirm that partici￾pants’ choice of hot sauce reflected hostile aggression. Note that gender did not interact with any experimental manipulations to shape key dependent variables and including it in the analyses reported next did not alter their outcome. Narcissism and aggression. Did narcissism augment these hostile and aggressive responses to provocation? Recall that we expected vulnerable (but not grandiose) narcissism to intensify aggressive retaliation and rage reactions, and to do so even toward an “innocent” target. In the context of our experimental design, this implied an interaction between narcissistic vulnerability and the provocation condition, but no three-way interaction between these factors and target condition (as vulnerability should augment aggressive responses to provocation both against the initial pro￾vocateur and a third party). Accordingly, a logistic regression analysis (with main effects, two-way interactions, and the three￾way interaction entered in successive blocks) on hot-sauce choice revealed a significant interaction between vulnerable narcissism and provocation (B  1.76, SE  .821, Wald  4.62, p .05), in addition to replicating the main effect of provocation described earlier (B  2.90, SE  1.43, Wald  4.13, p .05; no other effects were significant). The nature of this interaction is illustrated in Figure 2, in which predicted proportions of choosing hot over mild sauce across the control and the provocation conditions are indicated for individuals standing one standard deviation above, and one below, on the measure of vulnerable narcissism. Whereas the aggressive choice among those low on vulnerability only in￾creased by 17% as a function of provocation (simple slope B  1.13, SE  1.05, p  .28), the same choice among those high on narcissistic vulnerability increased by 43% as a function of provocation (simple slope B  4.67, SE  2.08, p .05), indicating a much stronger aggressive response to provocation. There was no evidence for a three-way interaction between narcissistic vulnerability, provocation, and target conditions (B  1.37, SE  1.06, Wald  1.69, p  .19), indicating that vulnerability increased aggression equally against the original provocateur and the annoying third party. Anger. Similarly, a hierarchical regression analysis examin￾ing anger as a function of experimental conditions and narcissistic vulnerability revealed an interaction between provocation and vul￾nerability (B  .30, t  2.37, p .05), in addition to replicating the experimental effect of provocation (B  .73, t  5.19, p .001; no other main effects or interactions were significant, all ps .29). These results are illustrated in Figure 3, in which anger reported across the control and the provocation conditions is indicated for individuals standing one standard deviation above and below on vulnerable narcissism. Simple slope analyses con￾firmed that those high on vulnerable narcissism reported increases in anger that were more than twice as large (B  1.04, t  5.40, p .001) than did those low on vulnerable narcissism (B  .43, t  2.29, p  .02). Trustworthiness. Furthermore, a hierarchical aggression analysis examining perceived trustworthiness as a function of experimental conditions and narcissistic vulnerability revealed an interaction between provocation and vulnerability (B  .47, t  2.05, p .05), in addition to replicating the experimental effect of provocation (B  2.74, t  8.44, p .001; no other main effects or interactions were significant, all ps .16). These results are illustrated in Figure 4, in which trustworthiness reported across the control and the provocation conditions is indicated for individuals standing one standard deviation above and below on vulnerable narcissism. Simple slope analyses confirmed that those high on vulnerable narcissism perceived one and a half times less trustworthiness (B  3.30, SE  .44, p .001) as a function of provocation than did those low on vulnerable narcissism (B  2.20, SE  .44, p .001). Depression. Finally, in accord with the narcissistic rage hy￾pothesis, vulnerability interacted with provocation in predicting depressive responses (B  .16, t  2.33, p .05, with no other effects reaching significance, all ps .10). This interaction is illustrated in Figure 5, and simple slope analyses confirmed that only those high on vulnerable narcissism responded with more depressive reactions (B  .42, SE  .13, p .01), whereas those low on vulnerable narcissism did not (B  .08, SE  .13, p  .51). Of note, the same analysis conducted on the specific item “ashamed” revealed a similar interaction pattern, although not reaching the conventional level of significance (B  .10, p  .18). Despite the fact the provocation was not overtly personal (i.e., did not resemble a typical “ego threat”), the fact that narcissistic participants reacted to the provocation with depression, even shame, does suggest they took it personally. Taken together, these results implicate narcissistic vulnerability as a potent predictor of aggression, anger, mistrust, and depression in the face of provo￾cation; all key signatures of narcissistic rage. Narcissistic grandiosity. Critically, narcissistic grandiosity did not interact with any experimental manipulations in shaping aggressive choices (all ps .17), anger (all ps .58), trustwor￾thiness (all ps .27), or depression (all ps .33). In contrast to prior findings on the role of grandiosity in aggression, when aggression was examined as retaliation for a relatively minor, 0 0.1 0.2 0.3 0.4 0.5 Control Provocaon Proporon Choosing Hot Version of Sauce Low Vulnerable Narcissism High Vulnerable Narcissism Figure 2. Aggressive choices as a function of provocation and vulnerable narcissism in Study 4. Low- and high-vulnerable narcissism represents individuals one standard deviation below and above the mean on vulner￾able narcissism, respectively. 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. 794 KRIZAN AND JOHAR