Research and theory emphasizing the role of cue diagnosticity in judgment (e.g. Skowronski and Carlston, 1987, 1989) suggests that under the proper conditions: (a) negativity effects should be observed in judgments of honesty/dishonesty; (b) positivity effects should be observed in judgments of intelligence/unintelligence, and (c) intelligence-implicative and dishonesty-implicative cues should be increasingly difficult to contradict as those cues become more extreme. Two experiments yielded data consistent with these predictions. In addition, two other important findings emerged from these studies. The results of Experiments 1 and 2 indicated that subjects do not respond as if highly diagnostic cues are sufficient for category membership, suggesting that the representational format of trait categories does not correspond to the format suggested by the ‘classical model’ of categorization. The results of Experiment 2 also indicated that negativity and positivity effects are not substantially altered by a role-playing manipulation designed to increase subjects' involvement in the judgment task.