Previous investigations regarding the influence of physical activity on cancer risk generally have utilized a single assessment of activity, thus failing to account for changes over time. Additionally, one assessment may be somewhat imprecise. We attempted to overcome these problems in a study of 17,607 men (aged 30-79 yr), followed from 1962 or 1966 (1962/1966) through 1988. We assessed physical activity (based on self-reported stair climbing, walking and participation in sports or recreational activities) twice: in 1962/1966 and again in 1977. A total of 280 colon, 53 rectal, 454 prostatic, 262 lung, and 88 pancreatic cancers developed during follow-up. Among alumni with Quetelet's index > or = 26 units, those highly active (energy expenditure > or = 2,500 kcal [> or = 10,460 kJ].wk-1) had 0.19 (95% confidence interval, 0.02-1.52) to 0.56 (0.29-1.09) times the colon cancer risk of those inactive (< 1,000 kcal [< 4,184 kJ].wk-1). Colon cancer risk was unrelated to level of activity among alumni with Quetelet's index < 26 units. Highly active alumni also had 0.39 (95% confidence interval, 0.18-0.85) to 0.62 (0.45-0.85) times the lung cancer risks of their inactive colleagues. Physical activity was not significantly associated with risks of rectal, prostatic, or pancreatic cancers. The biologic basis for a protective effect of increased activity on cancer risk appears plausible.