The potential role of physical activity in the primary prevention of non-insulin-dependent diabetes mellitus (NIDDM) is largely unknown. We examined the association between regular vigorous exercise and the subsequent incidence of NIDDM in a prospective cohort of 87,253 US women aged 34-59 years and free of diagnosed diabetes, cardiovascular disease, and cancer in 1980. During 8 years of follow-up, we confirmed 1303 cases of NIDDM. Women who engaged in vigorous exercise at least once per week had an age-adjusted relative risk (RR) of NIDDM of 0.67 (p less than 0.0001) compared with women who did not exercise weekly. After adjustment for body-mass index, the reduction in risk was attenuated but remained statistically significant (RR = 0.84, p = 0.005). When analysis was restricted to the first 2 years after ascertainment of physical activity level and to symptomatic NIDDM as the outcome, age-adjusted RR of those who exercised was 0.5, and age and body-mass index adjusted RR was 0.69. Among women who exercised at least once per week, there was no clear dose-response gradient according to frequency of exercise. Family history of diabetes did not modify the effect of exercise, and risk reduction with exercise was evident among both obese and nonobese women. Multivariate adjustments for age, body-mass index, family history of diabetes, and other variables did not alter the reduced risk found with exercise. Our results indicate that physical activity may be a promising approach to the primary prevention of NIDDM.