PURPOSE. To examine the medical school applications of physicians who are now established in their careers to see whether any of the applicants' data might have had predictive value for the admission process of a medical school wishing to increase its production of generalist physicians. METHOD. Three members of the Committee on Admissions of the University of North Carolina at Chapel Hill School of Medicine followed the same procedure they use when reviewing current medical school applicants in reading the applications of 148 graduates of the class of 1985. The readers recorded data from all parts of the American Medical College Application Service form–face sheet, personal statement, and the record of course work–and used alumni records and published physician locators to determine these class members' whereabouts and the specialties (and subspecialties) they were practicing 13 or 14 years after applying to medical school. RESULTS. Thirty-four percent of the class had elected generalist medical careers (the practice of family medicine, general internal medicine, or general pediatrics). A high service index (reflective of a demonstrated orientation toward community service prior to medical school matriculation) predicted strongly the choice of a generalist medical career. Conversely, the absence of any clear evidence of a service orientation predicted still more strongly a non-generalist career. Less strong predictors of a generalist practice included the selection of a generous number of non-science-content courses as an undergraduate, lower socioeconomic family origin, and a record of leadership in one or more extracurricular activities during college. CONCLUSION. If an admission committee informs itself of “what finally happens” to those it admits, its decisions can contribute to achieving whatever policy its medical school adopts with respect to the mix of physicians it wishes to produce.