Epidemiologic studies suggest that low carotene intake and low levels of serum retinol may be associated with an increased risk of cancer. Likewise, in some animal studies vitamin E has been associated with a reduced rate of induced cancers. Therefore, we measured retinol, retinol-binding protein, vitamin E (alpha-tocopherol), and total carotenoids in serum collected in 1973 from 111 participants in the Hypertension Detection and Follow-up Program who were free of cancer at the time but were diagnosed as having cancer during the subsequent five years. These measurements were compared with those in 210 controls who were matched for age, sex, race, and time of blood collection, and who remained free of cancer. Mean values for retinol were similar for cases and controls (67.3 and 68.7 micrograms per deciliter, respectively [95 per cent confidence limits for case-control difference, -6.7 to 3.5]). Values were also similar for retinol-binding protein (6.01 and 5.94 mg per deciliter [-0.42 to 0.56]), and carotenoids (114.5 and 111.6 micrograms per deciliter [-9.1 to 15.9]). The mean base-line retinol level in the 18 subjects with subsequent lung cancer was higher than that in their matched controls (79.0 vs. 71.4 micrograms per deciliter, -4.9 to 19.7). Serum vitamin E levels were somewhat lower in subjects who later had cancer than in controls (1.16 and 1.26 mg per deciliter, -0.22 to 0.02), in part because of the confounding effect of serum cholesterol levels (when adjusted for lipid levels, the case-control difference was -0.05 mg per deciliter; -0.17 to 0.07). These data do not support hypotheses relating intake or serum levels of antioxidant vitamins to a reduced cancer risk.