The association of physical activity level with the risk of death was analysed for a cohort of 636 healthy Finnish men aged 45-64 years followed up for 20 years. 39% of the cohort were classed as highly active physically at baseline in 1964. Up to 1984 there were 287 deaths, 106 of them due to coronary heart disease (CHD). During the first-two thirds of the follow-up, men with high physical activity had a lower risk of death than did men with low physical activity. During the last third, the survival curves of the men with high and low physical activity gradually converged. Of the men who died, those with high physical activity lived 2.1 years longer (p = 0.002) than those with low physical activity, after adjustment for age, smoking, blood pressure, serum cholesterol, and body mass index. This difference was due mainly to fewer CHD deaths among the highly active group. Low physical activity was clearly weaker than smoking as a predictor of risk of death. High physical activity may thus independently prevent premature death among middle-aged men, but it probably does not prolong the maximum achievable life-span.