Federal support of graduate medical education (GME) has been accepted as an intrinsic ingredient of the Medicare program since that program's inception. Streams of clinical income generated by teaching hospitals, medical faculty practice plans, Medicaid, and other state and federal sources have also made important contributions to GME. Although it is difficult to ascribe legislative intent precisely, Medicare funding seems to have been based on a 2-fold assumption: that GME was socially beneficial and that there were legitimate costs to teaching hospitals associated with their educational missions, even though such costs were hard to identify specifically. The benefits and costs include the high quality generally ascribed to a teaching hospital environment; extra services and teaching costs; active, unsponsored research; a higher proportion of complex medical conditions and care; and technology development and introduction. The argument for Medicare support of GME is thus partly based on better service to all Medicare beneficiaries, and partly on a broader social investment in education and teaching hospitals, with benefits accruing to both present and future Medicare participants. Teaching hospitals and their young physicians in training are also important in providing care to underserved poor populations.