Objective: To assess the educational outcomes of training junior medical students in an eight-week combined ambulatory/inpatient rotation compared with those of training exclusively on inpatient services. Design: Participants were randomly assigned to one of three groups: those who volunteered for the ambulatory/inpatient rotation and were randomly selected; those who volunteered and were not selected; and those who did not volunteer. Setting: University-based internal medicine (IM) inpatient services and community-based clinics and private practices. Participants: Seventy-five third-year medical students taking the required eight-week IM clerkship. Intervention: Assessment included both pre- and posttest measurement of students' knowledge of general internal medicine and a profile of the types of patient problems seen by students in ambulatory settings. Results: While students' general medicine knowledge scores increased significantly from pre- to posttest (p < 0.001), there was no significant difference in scores between the ambulatory/inpatient and exclusive inpatient groups. Patient log data indicated notable differences in the diagnostic compositions of students' patient loads. For instance, 38% of the ambulatory diagnoses were infectious disease, neurologic, endocrine, rheumatologic, or dermatologic problems, and another 15% were non-IM (e.g., obstetric/gynecologic; ear, eye, nose, and throat) problems. Only 24% of the inpatient diagnoses reflected these specialty areas. Conclusions: Ambulatory training did not significantly affect students' knowledge gain compared with that for exclusive inpatient training, but student evaluations of the rotation plus patient log data suggested that ambulatory training can provide a more complete view of general medicine practice than can exclusive inpatient training.