Notwithstanding its biologic plausibility, the association between physical activity and endometrial cancer has been analyzed in only a few epidemiological studies. Retrospective assessment of exposure and small sample size often hampers interpretation of published data. We studied risk for endometrial cancer in relation to physical activity at work in a large cohort of Swedish women identified in the nationwide censuses in 1960 and 1970, with jobs that could be consistently classified into one of 4 levels of physical demands. Follow-up from 1971 through 1989 was accomplished through record linkages. Multivariate Poisson regression models were used to estimate relative risk. The risk for endometrial cancer increased regularly with decreasing level of occupational physical activity (p for trend < 0.001), and was associated more strongly with activity in 1970 than in 1960. In multivariate analyses, adjusted for age at follow-up, place of residence, calendar year of follow-up, and social class, the relative risk among women with the same physical activity level in 1960 and in 1970 was 30% higher for sedentary as compared with high/very high activity level; (p for trend=0.04). The protective effect of physical activity appeared to be confined to women aged 50 to 69, among whom sedentary work was associated with a 60% higher risk than that observed among women estimated to be physically most active. The excess seemed to disappear within 10 years after a change in physical activity level. Although confounding cannot be ruled out in our data, occupational physical activity appears to reduce the risk for endometrial cancer.