Attempts to relate the quality of physician performance to early factors in the physician's training. Data were collected on students from a sample of representative U.S. medical schools. The information included demographic data, premedical school characteristics, medical school performance, and type of graduate medical education. The American Board of Internal Medicine (ABIM) supplied results of all specialty and subspecialty board examinations administered by them to 1955, 1960, 1965, and 1969 graduates of 9 medical schools, through the year 1973. Regression analyses were performed to assess the relative importance of a number of individual characteristics and "school effects" in predicting ABIM exam scores. The ABIM score correlates well with other evaluation criteria. Using data for individuals who retook the test, the measurement error of the ABIM test is estimated to be relatively small.