In an experiment on simulated juries subjects were asked to record their impressions of the major participants in the trial. 134 subjects heard a taped reconstruction of a case of theft, based on the court transcript, 224 subjects heard a case of rape. Their impressions, recorded on an adjectival scale, were factor analysed. There was considerable stability in the factors that emerged across trials and across target figures. One factor only correlated with verdicts in the case of theft, namely seeing the defendant as untrustworthy correlated with convicting him. In the rape case impressions of the defendants were largely uncorrelated with verdict, although impressions of the victim were highly correlated with verdict. The results are interpreted in terms of the simultaneous operation of the principles of consistency, salience and post‐decision dissonance.