The objectives of this study are to compare the rural location of international medical graduates (IMGs) and U.S. medical graduates (USMGs) by specialty (primary care vs. specialty care) according to geographical measures of need. This study utilized a cross-sectional survey using the 1997 American Medical Association Physician Masterfile for all active post-resident allopathic physicians and the Area Resource File (ARF) (Bureau of Health Professions, 1996) for all active post-resident osteopathic physicians in 1995 in the rural U.S. physician work force (N = 69,065). Allopathic physician ZIP code location was matched to county data using the ARF. The key measure was the difference in proportions between USMGs and IMGs in each state's rural counties characterized by need: high infant mortality, low socioeconomic status, high proportion of nonwhite population, high proportion of population 65 years and older, and low physician-to-population ratio. Primary care and specialty care rural physicians were studied separately. A disproportion of IMGs were located in needy rural counties of more states than were USMGs. Further, IMG disproportions were generally larger than USMG disproportions when they existed. Disproportions of IMGs tended to be located more often in the central and south census regions. Disproportions of specialty care IMGs were more frequent and of greater magnitude than those of primary care IMGs. Variations in the relative and absolute numbers of IMGs and USMGs among the states was wide. Services delivered by active post-resident primary care and specialty care IMGs appeared to be disproportionate to their overall number compared with USMGs in numerous needy rural counties. The extent of the IMG "safety net" presence differed, however, by the criteria used. Still, proposed limits on IMG entry into U.S. residency training may create long-term problems of access to rural physician services absent policies to induce USMGs or midlevel practitioners to locate in such areas. State-by-state assessments of the potential impact of IMG restrictions are called for because of the wide state-level variation that existed in comparative IMG-USMG distributions.