The authors evaluated and reviewed the literature on the effects of medical school curricula, faculty role models, and federal biomedical research support on the specialty choices of U.S. medical students. All 275 articles on these subjects published from 1984 through 1993 were considered. An instrument was developed to assess the quality of the articles. A total of 85 articles met study criteria and were reviewed. The mean score achieved was 42.7% of the total possible points. Major educational reforms emphasizing primary care have resulted in significant increases in the percentages of graduates choosing generalist careers. Except for required clinical training in family practice, individual curriculum components have generally not been successful. Students and physicians often stated that faculty role models influenced specialty choices, and there is some evidence that faculty composition is related to students' career choices. There was a consistent inverse correlation between the amount of federal biomedical research support received and the percentage of a school's graduates choosing generalist careers. It is unknown whether this relationship is causative and, if so, how research funds affect specialty choices. The best strategies to enlarge the proportion of medical students choosing generalist careers include institutional reform to emphasize generalist training, increasing the size of generalist faculty, and requiring clinical training in family practice. The relationship of federal biomedical research support to the specialty choices of medical students needs to be studied further. Research on specialty choice could be improved by including a larger number of schools and students, studying trends over several years, and using validated measures and outcomes, control groups, and multivariate analyses.(ABSTRACT TRUNCATED AT 250 WORDS)