To investigate whether medical housestaff report race information differently during case presentations of black patients and white patients, a prospective observational study was performed. Without informing housestaff, a chief resident recorded data during consecutive case presentations over two months. For each presentation, the data included: 1) whether, where, and how often race was identified; 2) whether certain prospectively selected, "possibly unflattering characteristics" were mentioned; and 3) whether any "justifying" diagnoses were considered during presentation or subsequent discussion. Justifying diagnoses were those in which a patient's race was important in considering the likelihood of possible diagnoses. Twenty-three house officers presented 18 black and 35 white patients. A single East Indian patient was excluded from analysis. Race was specified more often during presentations of black than of white patients (16 of 18 for blacks vs. 19 of 36 for whites; p less than 0.01). For two black patients, a justifying diagnosis was considered, but excluding these patients did not change the results. Two other differences did not achieve statistical significance. Race was more often specified prominently and repeatedly during presentations of black patients. Among patients to whom "possibly unflattering" characteristics were attributed, race was more likely to be specified for blacks (10 of 10) than for whites (4 of 9). These case presentations appeared to show a subtle bias.