This study examines the association between type of internal medicine training and satisfaction ratings among 509 patients who visited the clinic of an urban teaching hospital over a 3-month period in 1994. When controlling for patient, health-system, and other resident factors, primary care training was significantly associated with higher satisfaction ratings (cumulative odds ratio 1.53; 95% confidence interval 1.04, 2.25; p = .031) than categorical training. Using satisfaction ratings to rank the residents without adjusting for patient and health-system factors would have correctly classified only 27% of the residents in the lowest quartile. These findings have implications for both the education and potential employment of internists.