Inverse association between physical activity and colon cancer is well established, at least in men. We investigated the association of occupational physical activity with subsite-specific colorectal cancer risk. On the basis of occupational titles from the Swedish nationwide censuses in 1960 and 1970, we defined a cohort of women and men with the same work-related physical demands in 1960 and 1970. Incidence of colon and rectum cancer during 1971-1989 was ascertained through linkages to the Cancer Register. Relative risks (RRs) were estimated through Poisson regression. The risk for colon cancer increased with decreasing occupational physical activity. RR among sedentary women and men was 1.2 and 1.3 (P for trend=0.08 and <0.001). For men, the risks for proximal and distal colon cancer increased by 20 and 40% (P for trend=0.005 and <0.001). Inactivity seemed to be particularly associated with descending colon cancer (RR =2.4, P for trend<0.001). In women, the inverse association with activity was concentrated to proximal parts of colon; RR for cancer in the proximal and transverse colon among sedentary women was 1.4 and 2.0 (P for trend <0.07 and <0.01). Cancer of the rectum was not associated with activity in either sex. We confirmed the well-known inverse relationship between activity and risk of colon cancer but not rectal cancer in both sexes. Data suggest that the physical activity-related variation in risk among women is greatest in the proximal and middle parts of the colon, whereas the corresponding peak in men seems to be more distal. Sex-specific anatomic and motility differences of the colon might contribute to this subsite difference.