During debriefing, 4 participants expressed suspicion regarding the authenticity of the feedback. We analyzed data with and without these participants, and no significant differences were observed between the full sample and the reduced sample. We present the data below with the suspicious participants excluded.
Various time points from the physiological data were unscorable because of faulty sensors, loss of signal, or noisy signals. Thus, the physiological data have variable degrees of freedom. Videotapes of 7 additional participants were impossible to code, either because of the quality of the video or because the video was cut off during the interaction.
CV Data: Baseline and Speech Reactivity
Mean VC, CO, HR, TPR, and MAP values were calculated for each minute within each rest and task period. We began by examining the last minute of the baseline for differences based on participants’ race, evaluators’ race, and/or feedback manipulation. Though random assignment was successful and there were no baseline differences in CV responses due to feedback or evaluator’s race, there were differences in VC between White participants and Black participants, F(1, 106) = 12.16, p < .0007. Black participants exhibited a shorter preejection period (M = 102.4, SD = 12.27) than did White participants (M = 111.7, SD = 15.00). Because of these baseline differences, we conducted all subsequent analyses examining VC reactivity by controlling for participants’ baseline VC.
We then examined CV responses during the speech that occurred prior to the feedback manipulation. There were no differences in responses by evaluator’s race or participant’s race. It is important to note that the speech was not a face-to-face speech, which appears to more reliably lead to differences in CV reactivity as a result of the situational context (Mendes, Blascovich, Hunter, Lickel, & Jost, 2007
Hypotheses testing began with an examination of the various dependent variables (CV reactivity, self-reports, attributions, performance, and observers’ ratings) and the three independent variables (participant’s race, evaluator’s race, and feedback) in a series of 2 × 2 × 2 analyses of variance (ANOVAs). We first examined whether the three-way interactions were significant. We obtained significant three-way interactions for most of the critical dependent variables: attributions to discrimination, self-reported external negative emotion (i.e., anger), CV responses (CO and TPR), performance, and observers’ coding indexing vigilance and positive emotion. A complete summary of the full model ANOVAs are presented in .
Summary of Effects From ANOVAs Testing the Participant’s Race × Evaluator’s Race × Feedback
We then decomposed the significant three-way interactions by examining the effects of participant’s race and evaluator’s race separately for the rejection conditions and the acceptance conditions. To conduct these simple effects tests, we used the overall mean square error and degrees of freedom found in . Significant two-way interactions were then further examined by simple effects tests, within participant’s race, to determine whether the effect for evaluator’s race was significant. Finally, we tested our a priori predictions regarding differences between Black participants and White participants paired with different-race partners, which we tested using planned contrasts. Specifically, we examined whether Black participants paired with White evaluators differed from White participants paired with Black evaluators. Nonsignificant differences are characterized as symmetrical effects and significant differences are characterized as asymmetrical effects. We also examined participant race effects in same-race pairings, and we never observed any significant race effects. That is, within both feedback conditions, Black participants paired with Black evaluators yielded patterns of findings similar to White participants paired with White evaluators. Means and standard deviations are found in .
Summary of Means, Standard Deviations, and Effects From Primary Dependent Variables
Not surprisingly, participants were more likely to attribute the cause of feedback to discrimination after social rejection than they were to attribute the cause of feedback to discrimination after social acceptance, F(1, 110) = 29.59, p < .001. Furthermore, participants, in general, were more likely to attribute feedback to discrimination when the evaluators were White than when the evaluators were Black, F(1, 110) = 6.76, p < .01. However, the predicted three-way interaction was significant, F(1, 110) = 7.35, p < .008. To decompose this interaction, we conducted simple effects tests within the social rejection condition and within the social acceptance condition.
Within the social rejection condition, the Participant × Evaluator race interaction was significant, F(1, 110) = 9.71, p < .01. As expected, Black participants were more likely to attribute rejection to discrimination when paired with White evaluators than when paired with Black evaluators, F(1, 110) = 14.18, p < .001. Among White participants, the direction of the means was consistent with predictions—more attributions to discrimination when the evaluator was Black than when the evaluator was White—but this simple effect was not significant, F(1, 110) = 1.01, ns. Comparisons of responses within the different-race dyads revealed that Black participants paired with White evaluators were more likely to attribute rejection to discrimination (M = 2.9, SD = 0.98) than were White participants paired with Black evaluators (M = 2.5, SD = 0.87), although this difference was not significant, F(1, 110) = 1.75, ns. In sum, Black participants were more likely to attribute rejection to discrimination in a different-race interaction than in a same-race interaction, but among the different-race dyads, there were no significant differences between Black participants and White participants.
As the main effect for feedback revealed, attributions to discrimination were low following social acceptance. None of the simple effects and none of the planned contrasts were significant (Fs < 1).
We then examined the self-reported emotion ratings that participants completed prior to the cooperative task but following the feedback. As expected, there were main effects for feedback. Participants who received social acceptance feedback reported more positive emotion than those who received social rejection feedback, F(1, 110) = 7.72, p < .007. Similarly, participants reported more external negative emotion when they received social rejection feedback than when they received social acceptance, F(1, 110) = 31.66, p < .001. Consistent with our predictions, however, the three-way interaction for external negative emotions was significant, F(1, 110) = 4.58, p < .04.
Again, we decomposed the interaction by first examining the effects of the interaction of participant’s race and evaluator’s race separately for social rejection and social acceptance conditions. Among those who received social rejection feedback, the Participant × Evaluator race interaction was significant, F(1, 110) = 4.20, p < .05. White participants rejected by a Black evaluator reported more external negative emotions than did White participants rejected by a White evaluator, F(1, 110) = 3.85, p < .052. Similarly, Black participants rejected by a White evaluator reported more anger than did those rejected by a Black evaluator, though this effect was not significant, F(1, 110) = 2.70, p < .10, Cohen’s d = .31. We then compared the means from Black and White participants paired with different-race partners. In support of the symmetrical prediction, Black and White participants did not significantly differ in their reports of external negative emotions (F < 1). We did not hypothesize or observe differences in external negative emotion among the groups who received social acceptance feedback.
To test our predictions for CV reactivity, we focused on the first minute of the task because cardiac habituation can occur quickly, thus differences in cardiac responses related to challenge and threat are most pronounced in the earlier minutes. However, the effects we observed for the first minute persisted for the entire 4 min of the task. The three-way interactions were observed for CO, F(1, 102) = 8.13, p < .006, and for TPR, F(1, 102) = 15.02, p < .001. The three-way interaction for VC, controlling for baseline VC, was not significant, F(1, 102) = 2.31, p < .13.
We hypothesized that social rejection from a same-race partner would result in CV responses consistent with threat (lower CO and higher TPR), whereas rejection from a different-race partner would result in activational CV responses (relatively higher CO and lower TPR). Consistent with predictions, we observed significant Participant × Evaluator race interactions for CO, F(1, 102) = 6.43, p < .05, and for TPR, F(1, 102) = 11.18, p < .05. Decomposing the two-way interactions revealed significant (or near significant) differences by evaluator’s race for Black participants, for CO, F(1, 102) = 3.10, p < .08, and for TPR, F(1, 102) = 6.38, p < .013. Black participants rejected by Black evaluators exhibited significantly greater increases in vascular resistance and exhibited less cardiac efficiency than did those rejected by White evaluators (see and ). Similar patterns of reactivity were observed among White participants for CO, F(1, 102) = 8.33, p < .01, and for TPR, F(1, 102) = 6.12, p < .02. In general, rejection from a same-race partner resulted in increased TPR and decreased CO (i.e., a threat pattern of CV reactivity), whereas rejection from a different-race partner resulted in decreased TPR and increased CO (i.e., an activational pattern of CV reactivity). An examination of the effects within the different-race dyads revealed no effects for participant’s race. Both White and Black participants rejected by different-race partners exhibited activational CV patterns (Fs < 1).
Cardiovascular reactivity from the first minute of the cooperative task following rejection, by participant’s race and evaluator’s race. The error bars indicate the standard error of the means.
Our predictions following social acceptance were that a same-race partner would engender CV reactivity consistent with activational responses (higher CO and lower TPR), whereas positive feedback from a different-race partner might engender CV reactivity threat (lower CO and higher TPR). The Participant × Evaluator race interactions were significant for TPR, F(1, 102) = 4.52, p < .036, and near significant for CO, F(1, 102) = 3.11, p < .08 (). Among Black participants, those paired with White evaluators exhibited significantly lower CO reactivity (M = −0.5, SD = 1.0) and greater TPR reactivity (M = 184.8, SD = 417) than did those paired with Black evaluators (CO: M = 0.8, SD = 1.1; TPR: M = −97.4, SD = 201), for CO, F(1, 102) = 4.89, p < .03, and for TPR, F(1, 102) = 8.22, p < .005. Black participants who received positive social feedback from White evaluators exhibited CV reactivity consistent with threat, whereas Black participants who received positive feedback from Black evaluators exhibited CV reactivity consistent with challenge. The simple effects tests among White participants were not significant (Fs < 1). White participants who received social acceptance feedback exhibited CV responses consistent with challenge states, regardless of the race of the evaluator. Planned contrasts confirmed the asymmetry, for CO, F(1, 102) = 4.84, p < .03, and for TPR, F(1, 102) = 5.21, p < .025. Black participants exhibited threat CV responses in different-race pairings, whereas White participants exhibited challenge responses in different-race pairings.
Cardiovascular reactivity from the first minute of the cooperative task following social acceptance, by participant’s race and evaluator’s race. The error bars indicate the standard error of the means.
We operationalized performance as the number of words identified by the participant during the word-finding task. We observed a main effect for race, F(1, 108) = 5.62, p < .01 (on average, White participants found more words than Black participants); however, it is important to note that White participants were more likely to report that they played a similar word game (44%) in the past than were Black participants (29%). Because past experience with similar games is an important predictor of performance, we controlled for past experience in all analyses. We again observed the three-way interaction, F(1, 108) = 8.26, p < .005.
We predicted that participants who showed activational responses would perform better than those who showed an inhibitional response. Following the confirmation of the predictions with CV reactivity, we anticipated that those who were rejected by a different-race evaluator would perform better than would those rejected by a same-race evaluator. Consistent with the predictions, the Participant × Evaluator race interaction was significant, F(1, 108) = 3.92, p < .05. Among Black participants, those rejected by White evaluators performed better than did those rejected by Black evaluators, though this was short of significance, F(1, 108) = 2.65, p < .10, Cohen’s d = .57. Similarly, among White participants, those rejected by Black evaluators performed better than did those rejected by White evaluators, but again this effect was not significant, F(1, 108) = 2.83, p < .09, Cohen’s d = .61. In general, participants rejected by a different-race partner performed (marginally) better than did those rejected by a same-race partner. We then examined the performance effects within the different-race pairings. Even in light of the participant race main effect described above, the difference between Black participants paired with White evaluators (adjusted M =13.2, SD = 4.5) and White participants paired with Black evaluators (adjusted M = 15.9, SD = 5.6) was not significant, F(1, 108) = 2.15, p < .15.
Among those who received social acceptance feedback, the Participant × Evaluator race interaction was significant, F(1, 108) = 3.99, p < .048. The simple effects test among Black participants was significant, F(1, 108) = 5.61, p < .02. Black participants performed better when paired with Black evaluators who gave positive feedback, compared with Black participants paired with White evaluators. Among White participants, the effect for evaluator’s race was not significant (F < 1). The planned contrast comparing Black participants with White evaluators and White participants with Black evaluators was also significant, F(1, 108) = 5.86, p < .025. Black participants paired with White evaluators performed worse than did White participants paired with Black evaluators.
Our final dependent variables were from the behavioral coding, in which coders (blind to the social feedback condition) rated the extent to which participants appeared vigilant (monitored the other participant) and displayed internal negative emotions (e.g., shame), external negative emotions (e.g., anger), and positive emotions during the word-finding task. The three-way interactions were confirmed for vigilance, F(1, 98) = 3.96, p < .05, and for positive emotions, F(1, 103) = 4.91, p < .03. The interaction for external negative emotions was just short of significance, F(1, 103) = 3.10, p < .08, Cohen’s d = .03.
Among those who received social rejection, the two-way interactions were not significant for vigilance or positive emotions. We predicted that those paired with different-race evaluators who had just given them rejecting social feedback would show more anger than would those rejected by same-race evaluators. Even though the omnibus interaction was just short of significant, because we had a priori predictions regarding this effect, we tested the Participant × Evaluator race interaction, which was significant, F(1, 103) = 6.02, p < .05. Among Black participants, those rejected by White evaluators were perceived as appearing angrier than were those rejected by Black evaluators, F(1, 103) = 4.57, p < .04. Similarly, among White participants, those rejected by Black evaluators appeared angrier than did those rejected by White evaluators, F(1, 103) = 4.04, p < .05. Again, the effects within different-race settings appeared to be symmetrical in that White participants rejected by Black partners were rated as exhibiting similar levels of external negative emotions (M = 1.6, SD = 0.8) as Black participants rejected by White partners (M = 1.7, SD = 0.8; F < 1).
Among those who received social acceptance feedback, we hypothesized that Black participants might be uncertain of or suspicious of social acceptance from White evaluators, resulting in increased vigilance. We tested this hypothesis by examining our behavioral measure of vigilance—how often participants monitored their partner during the cooperative task. We observed a significant two-way interaction, F(1, 98) = 4.55, p < .05. Among Black participants, the evaluator’s race resulted in a significant effect, F(1, 98) = 4.74, p < .05. Black participants interacting with White evaluators monitored them more during the task than Black participants interacting with Black evaluators. Among White participants, the race of the evaluator was not significant (F < 1). Planned contrasts comparing the intergroup pairings yielded a significant effect as well, F(1, 105) = 4.17, p < .05. Black participants paired with White evaluators monitored them more often (M = 3.1, SD = 2.5) than did White participants paired with Black evaluators (M = 1.6, SD = 1.2).
We also hypothesized that positive emotion would be greater following social acceptance from a same-race partner than following social acceptance from a different-race partner. Analysis of coders’ ratings of how much positive emotion the participants exhibited during the cooperative task revealed a significant Participant × Evaluator race interaction, F(1, 103) = 4.56, p < .05. As predicted, among Black participants, those paired with a same-race evaluator displayed more positive emotion than did those assigned to a different-race evaluator, F(1, 103) = 11.06, p < .002. The simple effects test was not significant among White participants. Similar to the other measures, Black participants positively evaluated by White partners were rated as expressing less positive emotion (M = 1.5, SD = 0.3) than White participants positively evaluated by Black evaluators (M = 2.0, SD = 0.5), F(1, 105) = 5.85, p < .05.
Differentiating CV Activational Responses: Positive Emotions Versus Anger
We observed two meta-conditions that resulted in CV reactivity consistent with activational patterns—social acceptance from same-race partners and social rejection from different-race partners.2
However, we expected these two conditions to engender very different emotions. Specifically, we expected social acceptance from same-race partners to engender positive emotions (pride, confidence), and we expected social rejection from different-race partners to engender anger. We found initial confirmation for these predictions with behavioral observation and, to a lesser extent, self-reported emotions.
We then turned to our predictions regarding differentiating anger from positive emotions using cardiac variables, specifically, VC and HR reactivity. We predicted that anger would yield larger cardiac responses than would positive emotions; hence, we expected that VC and HR changes would be higher in the rejection/different-race conditions than in the acceptance/same-race conditions. An ANOVA revealed significant differences in VC (averaged across the task) between conditions in the direction hypothesized, F(1, 54) = 4.34, p < .04. Participants rejected by different-race partners exhibited greater VC (M = 17.15, SD = 10.3) than did those who received positive feedback from same-race partners (M = 11.5, SD = 10.0). The same analysis performed on HR changes revealed significant differences between conditions in the predicted direction as well, F(1, 55) = 8.31, p < .006. Again, the discrimination condition resulted in greater increases in HR (M = 11.73, SD = 8.8) than did social acceptance by a same-race partner (M = 5.69, SD = 6.8).
Finally, we examined correlations between cardiac variables (VC and HR) and observed emotions. We predicted that VC responses would be related to greater anger, but only in conditions in which anger was evoked—the rejection/different-race conditions. Consistent with these predictions, VC changes were positively and significantly related to expressed anger (average r
= .53, p
< .01) as were HR changes (average r
= .36, p
< .05). In the same-race and positive feedback conditions, these relationships were not observed (for VC and expressed anger, average r
= −.07, ns
; for HR, average r
= −.23, ns
). Furthermore, we expected that positive emotions would be related to a dampening of CV responses (see Fredrickson, Mancuso, Branigan, & Tugade, 2000
). The results for this effect were weaker but were in the predicted direction. In the same-race and positive feedback condition, less VC was related to greater observer-rated positive emotion (r
= −.31, p
< .052). HR changes were also negatively related to positive emotion following positive feedback, but the relationship was weak (r
= −.19, ns
). These analyses show some support for the idea that VC and HR responses tend to increase as the intensity of anger increases.
To explore possible mechanisms of the performance effects observed, we examined relationships between CV reactivity and performance. Consistent with the idea that threat responses impair cognitive performance, we observed that the greater the CV threat pattern, the fewer words participants found (for CO, r
= .31, p
< .007; for TPR, r
= −.22, p
< .03). As expected, VC and HR did not predict performance. These relationships lend support to the argument that threat reactivity is part of the profile of physiological responses that may be associated with performance decrements and may operate in stereotype threat situations (Blascovich, Spencer, Quinn, & Steele, 2001
) as well as other threatening or distressing situations.