Physical activity may help preserve cognitive function and decrease dementia risk, but epidemiologic findings are inconsistent. The authors conducted a prospective study to determine the association between physical activity and risk of dementia, Alzheimer's disease, and vascular dementia. The US study population comprised 3,375 men and women aged 65 years or older, free of dementia at baseline, who participated in the Cardiovascular Health Cognition Study in 1992-2000. Leisure-time energy expenditure and an activity index reflecting number of different physical activities were calculated. Analyses were based on Cox proportional hazards models. There were 480 incident cases of dementia over an average of 5.4 years of follow-up. After multivariate adjustment, participants in the highest quartile of physical energy expenditure had a relative risk of dementia of 0.85 (95% confidence interval: 0.61, 1.19) compared with those in the lowest quartile, and participants engaging in >or=4 activities had a relative risk of dementia of 0.51 (95% confidence interval: 0.33, 0.79) compared with those engaging in 0-1 activity. These associations were more marked in apolipoprotein E genotype (APOE) epsilon4 allele noncarriers but were absent in carriers. A similar pattern was observed for Alzheimer's disease and vascular dementia. Mechanisms to explain the observed relations deserve further study.