PURPOSE: To describe trends in specialty choice and to identify predictors of primary care specialty choices among graduates of U.S. MD-granting medical schools. METHOD: A longitudinal study evaluated 1997-2006 medical school graduates who completed the Association of American Medical Colleges Matriculating Student Questionnaire and Graduation Questionnaire. Multivariate logistic regression identified significant predictors of graduates choice of primary care specialty (general internal medicine, general pediatrics, internal medicine subspecialties, pediatrics subspecialties, family medicine, and obstetrics-gynecology) or "no-board-certification specialty," compared with all other specialties (reference). RESULTS: The sample included 102,673 graduates (64.9% of all 1997-2006 graduates). General internal medicine, family medicine, general pediatrics, and obstetrics-gynecology choice decreased, whereas internal medicine subspecialties, pediatrics subspecialties, and no-board-certification specialty choice increased over time (each: P < .001). Female graduates and those who planned to practice in underserved communities, espoused more-altruistic beliefs about health care, and ascribed greater importance to social responsibility in their choice of medicine at matriculation were more likely to choose general internal medicine, general pediatrics, family medicine, or obstetrics-gynecology, whereas graduates who had a physician parent and who planned full-time academic medicine careers were less likely to do so (each: P < .01). Graduates with higher debt were less likely to choose internal medicine and pediatrics specialties (each: P < .001) and more likely to choose obstetrics-gynecology (P = .001). CONCLUSIONS: Generalist-primary care specialty choices declined since 1997, whereas primary care subspecialty and no-board-certification specialty choices increased. Associations between primary care specialty choices and demographic, attitudinal, and career intention variables can inform the design of interventions to address expected primary care workforce shortages.