Three studies demonstrated that induced compliance can change socially significant attitudes and that the change generalizes to broader beliefs. White college students wrote an essay endorsing a pro-Black policy that was costly to Whites. In Experiments 1 and 2, attitudes and general beliefs about Blacks became more favorable in both high-and low-choice conditions, provided publicity of the essay was high. Overall, choice and publicity had additive effects on attitude change. Some highchoice Ss wrote only semipositive (semicompliant) essays and did not change their essay attitudes.Yet their beliefs about Blacks still became more favorable. In Experiment 3, racial ambivalence, but not prior attitude, predicted essay compliance (positive or semipositive essay). Ambivalent Ss were more likely to comply than were less ambivalent Ss. Thus, dissonance-mediated change may be facilitated when internal conflict already exists.Can induced compliance be used to reduce prejudice? If so, what are the implications for theories about prejudice and about cognitive dissonance-the process presumed to underlie induced compliance? These questions were the focus of the research reported in this article.

Previous research has suggested that Whites' evaluations of Blacks who are presented positively or negatively tend to be more extreme than evaluations of similar White targets. In Study 1, White subjects rated a Black or White confederate who was responsible for success or failure at a joint task. There was a clear cross-race polarization of evaluations. Study 2 tested two possible explanations of the polarization phenomenon-the authors' ambivalence formulation and Linville and Jones's cognitive complexity hypothesis. As hypothesized, ambivalence was directly related to favorability of impression ratings in the success condition. In the failure condition, the correlation was in the predicted inverse direction but was not significant. The difference between the two correlations, as expected, was significant. No support was obtained for the cognitive complexity hypothesis. Rather, in the failure condition complexity was directly related to the unfavorability of ratings, contrary to the hypothesis.

Students watched a theft video, attempted an identification from a thief-present or thief-absent lineup under unbiased or biased instructions, and rated identification confidence. In Experiment 1, the participants received (bogus) positive, negative, or no pre-identification feedback about a recall test. Biased instructions and positive feedback increased confidence and ratings of eyewitnessing conditions. In Experiment 2, biased instructions increased confidence unless the thief was absent and lineup members were similar, where they decreased confidence. According to the cue-belief model, biased instructions send a positive accuracy cue regarding the most familiar-looking lineup member. If none stands out, instructions conflict with an inclination to reject the lineup. Feedback may create a belief about memory quality that is a cue regarding likely recognition accuracy.

In 2 experiments, college students read a murder-trial transcript that included or did not include court-appointed expert testimony about eyewitness memory. The testimony either preceded or followed the evidence, and the judge's final instructions reminded or did not remind jurors about the expert's testimony. Expert testimony decreased perceptions of guilt and eyewitness believability when it followed the evidence and preceded the judge's reminder. This effect occurred whether the prosecution case was moderately weak or moderately strong. Jurors' need for cognition (NC) was curvilinearly related to convictions in a strong case. Low and high NC jurors convicted less than did moderate NC jurors. Greater scrutiny by high NC jurors may make them more likely to consider evidence for the weaker side.