A prospective cohort study of London civil servants was used to examine the relation of physical activity to various causes of death. 6,702 men aged 40-64 y who participated in a baseline examination between 1969 and 1970 were followed up for 25 y during which time there were 2859 deaths. The association of two measures of physical activity (leisure time activity and usual walking pace) with cause-specific mortality was examined. Walking pace demonstrated inverse relations with mortality from all-causes, coronary heart disease (CHD), other cardiovascular disease (CVD), all cancers, respiratory disease, colorectal cancer and haematopoietic cancer following adjustment for risk factors which included age, employment grade, smoking, body mass index, and forced expiratory volume (P [trend]<0.05 for all). In analyses restricted to men without disease at entry, walking pace retained inverse associations with all-cause, CHD, other cardiovascular disease, and haematopoietic cancer mortality (P [trend]<0.05 for all). Leisure time activity was also inversely associated with mortality from all-causes, CHD, other CVD, and all-cancers following adjustment for risk factors (P [trend]<0.05 for all). Eliminating deaths in the first 5 and 10 y of follow-up did not greatly alter these associations. It is concluded that physical activity may confer protection against death due to some cancers, in addition to reducing cardiovascular disease risk.