We studied the relation between self-reported physical activity and cancer in the first National Health and Nutrition Examination Survey (NHANES I) cohort, originally examined between 1971-75, and followed prospectively through the Epidemiologic Follow-up Study (NHEFS), conducted between 1982-84. Among 5,138 men and 7,407 women 25-74 years old, for nonrecreational activity we observed increased risk of cancer among inactive individuals compared to very active persons (for men, relative risk [RR] 1.8, 95% confidence interval [CI] = 1.4, 2.4; for women RR 1.3, 95% CI = 1.0, 1.8). These findings were unchanged after adjustment for cigarette smoking, body mass index (BMI), and other potential confounders. Sites which demonstrated stronger inactivity-cancer associations included colorectum (RR 1.6, 95% CI = 0.7, 3.5) and lung (RR 1.6; 95% CI = 1.2, 3.5) among men, and breast (post-menopausal) (RR 1.7; 95% CI = 0.8, 2.9) and cervix (RR 5.2; 95% CI = 1.4, 14.5) among women, although these findings for women were based on relatively few cases. The association between inactivity and cancer was greater among persons of moderate (or lower) BMI, those cases occurring three or more years after baseline, and, in women, those more than 60 years old. In contrast, recreational exercise showed little relation to cancer, with the exception of prostate cancer. The results suggest that inactive individuals are at increased risk of cancer.