Educational decision makers (n = 224) participated in a computer-simulated decision-making experience designed to ascertain the extent to which referral information biased classification decisions. Subjects were randomly assigned to 16 different conditions that varied on the basis of the referred child's sex, socioeconomic status, physical attractiveness, and nature of referral problem. Subjects accessed test data, all of which reported performance in the normal range. Only the nature of the referral problem was found to be influential in the simulated decisions. A referred child was more likely to be diagnosed as emotionally disturbed when the referral statement of the problem was listed as behavioral rather than academic. Subjects ignored standardized test information indicative of average performance, and retained the stereotype created by the referral information. The results are discussed with regard to implications for assessment of children and future research.