The association between serum alpha-tocopherol levels and the subsequent incidence of cancer was investigated in a longitudinal study of 21,172 men initially aged 15-99 years in six geographic areas in Finland. The baseline examination was conducted in 1968-1972, and during the follow-up of 6-10 years, 453 cancers were diagnosed. The serum alpha-tocopherol levels were measured from stored serum samples from these men and from 841 male controls, matched for municipality and age, who did not develop cancer during the follow-up. The mean levels of serum alpha-tocopherol among the cancer cases and controls were 8.02 and 8.28 mg/liter, respectively. A high serum alpha-tocopherol level was associated with a reduced risk of cancer. The relative risk of cancer in persons in the two highest (threshold 8.70 mg/liter) quintiles of serum alpha-tocopherol was 0.64 (95 per cent confidence interval = 0.49-0.85) in comparison with those in the three lowest quintiles. The association was strongest for the combined group of cancers unrelated to smoking and varied between subgroups of the study population as well as between different cancers. The association persisted when adjusted for serum cholesterol, serum vitamin A, serum selenium, and various confounding factors. It also persisted when subjects with possible signs of cancer at the time when the blood samples were drawn or with cancers diagnosed during the first two years of follow-up were excluded. These findings agree with the hypothesis that high vitamin E intake protects against cancer.