OBJECTIVES: To determine whether payment of primary care physicians based on capitation, with an additional incentive payment for low hospital-utilization rates, resulted in lower hospital-utilization rates among patients of these physicians than among patients of physicians still paid on a fee-for-service basis. DESIGN: Retrospective cohort study. SETTING: Capitation-based and fee-for-service primary care practices in Ontario. SUBJECTS: Thirty-nine physicians whose method of payment was converted from fee-for-service to capitation during the period from June 1985 to January 1989 and 7 physicians who remained in fee-for-service practice, two of whom were matched with one physician in capitation-based practice on the basis of practice location, type of practice (academic v. community), hours of practice (part-time v. full-time), years since graduation, physician group size, practice size (number of patients), type of group (primary care v. multispecialty), sex, certification in family medicine, country of graduation (Canada v. other) and age. One physician in capitation-based practice was matched with only one physician in fee-for-service practice. OUTCOME MEASURES: Annual hospital-utilization rates (hospital separations or hospital days per 1000 patients in each practice) for the physicians paid on a capitation basis 3 years before, 1 year before and 3 years after they converted from fee-for-service payment and at corresponding periods for the matched physicians still paid on a fee-for-service basis. RESULTS: The mean annual rate of hospital days used, adjusted for the age and sex of patients as well as for their social-program-recipient status, fell from 1085 per 1000 patients (3 years before the conversion date) to 1030 (1 year before conversion) and to 954 (3 years after conversion) in capitation-based practices. For the matched physicians in fee-for-service practice, the rates during the corresponding periods were 1085, 1035 and 956 hospital days per 1000 patients. The pattern was similar for rates of hospital separations, adjusted for patient's age, sex and social-program-recipient status. There were no statistically significant differences between the rates of hospital utilization among patients of physicians in capitation-based practices and the rates among those of physicians in fee-for-service practices during each of the three periods, nor were there significant differences in the changes in rates. CONCLUSION: Capitation payment, with an additional incentive payment to encourage low hospital-utilization rates, did not reduce hospital use. Factors other than the method of physician payment appear to be responsible for the variations in hospital utilization among practices.