This paper presents an extension of Freidson's typology concerning the four medical practice settings which physicians may enter. First, Freidson's typology is modified to contain only three medical practice settings: 1) solo practices; 2) small group practices (partnerships or associations consisting of two to seven physicians); and, 3) large group practices (having eight or more physicians). Then, it is argued that the most interesting sociological difference between these three medical practice settings is the differential probability for effective peer regulation, with that probability lowest in solo practices, highest in large group practices, and lying somewhere between these extremes in small group practices. Finally, it is argued that because physicians recognize these differential probabilities for peer regulation, they seek out those niches (i.e., medical practice settings) which most closely reflect their preferences. This extended version of Freidsonian theory is then incorporated into an analytic model using the sociodemographic, environmental, and attitudinal characteristics of physicians to predict their practice choices. Data from a 1979 national survey of approximately 4,500 physicians are used to assess the model empirically. The results obtained from these analyses conform quite well with both our general and specific expectations. The implications of these findings for the future configuration of the American health care delivery system are discussed with special reference to health maintenance organizations.