Mock juries of 5–7 jurors viewed one of three video-recorded versions of a rape victim's testimony, role-played by a professional actress. The statement was given in a free-recall manner with one of three kinds of emotions displayed, termed congruent, neutral and incongruent emotional expressions. The juries were requested to reach a decision on items in a short questionnaire, probing the perceived credibility of the witness and judgements of the probability of a guilty verdict. The jurors were then asked to complete the questionnaire a second time, individually and anonymously. A control group filled out the questionnaire individually without preceding jury deliberations. When participants judged credibility and guilt independently, without a preceding jury discussion, the displayed emotions strongly influenced the judgements. However, discussions in the context of the jury strongly attenuated the effect of displayed emotion, with judgements converging on the credibility of a neutral emotional expression as judged by independent participants, and the attenuating effect outlasted the jury-situation. The results are consistent with research within social psychology showing that social stereotypes and prejudices are often neutralised by group discussions. Copyright © 2006 John Wiley & Sons, Ltd.