This study examined physician test-ordering behavior at a large urban Veterans Administration Medical Center staffed by two medical schools. Patients are assigned randomly to the two schools' wards, which operate independently of each other. The medical records of 119 patients, admitted during a 5-month period, were abstracted for data on the process and outcomes of medical care. Only one statistically significant difference in diagnostic testing practices was revealed by the observation that 12% of patients on one service received another abdominal film, while no patients on the other service did. The implications of studies such as this for small-area analysis are discussed. If variation in the utilization of health services is observed in a geographic region, then a study of physician behavior at a single facility may help to identify the reason for the variation.