The authors prospectively examined the influence of increased levels of physical activity on risk of stroke in members of the Framingham Study cohort. Two separate analyses were performed, one during midlife in 1,897 men (mean age = 49.7 years) and 2,299 women (mean age = 49.9 years) and another when the cohort was older (1,361 men (mean age = 63.0) and 1,862 women (mean age = 63.7)). A structured questionnaire administered at two separate examinations was used to estimate the amount of metabolic work done during a typical 24-hour period. Physical activity was categorized into tertiles, and medium and high levels of physical activity were compared with a low level of physical activity, which was used as the referent group. Cox proportional hazards, life table, and time-dependent covariate analyses were used to examine the relation between level of physical activity and stroke risk over a follow-up period of up to 32 years. In men, adjusted analyses revealed that increased levels of physical activity were protective. The strongest effect was obtained from an analysis involving older cohort members in the medium tertile (risk ratio = 0.41, 95% confidence interval 0.24-0.69). High levels of physical activity did not confer an additional benefit over medium levels. Adjusted analyses showed no significant protective effect in women. These results indicate that medium and high levels of physical activity among men are protective against stroke relative to low levels.