The effectiveness of specific training in statistics and decision-making principles upon physicians' judgmental skills was assessed by means of problems of intuitive logical reasoning. The responses of 43 statistically sophisticated physicians (SP) were compared to those of 42 practicing physicians (PP), 43 clinical nurses (CN) and 41 hospital laborers (HL). On problems evaluating use of faulty heuristics in judgments of conditional probabilities, the SP group's responses were the most biased. The proportion of subjects displaying consistent use of a particular heuristic in solving the three problems were 0.36 (SP), 0.45 (PP), 0.35 (CN) and 0.41 (HL). On problems assessing use of prevalence rate data in estimating probabilities, SP performed substantially better than the other three groups: 34% of their responses were accurate. However, 37% of their responses reflected ignorance of prevalence information concepts. We conclude that intensive statistical and decision-making training of physicians is likely to be of only limited value for improving clinicians' judgmental skills.