This paper hypothesizes that beta-carotene mediates the association between low serum cholesterol and increased risk of lung cancer, predicts that the association should be greater in population strata with low intake of beta-carotene than in those with high intake if the hypothesis is correct, and investigates this prediction with data from a 24-year cohort study of 1,960 middle-aged employed men. In the total cohort, serum cholesterol was not related to risk of lung cancer. The relative risk associated with a difference of -1.0 mmol per liter in serum cholesterol was 1.01 (95% confidence interval of 0.80-1.27) after adjustment for cigarette smoking, age, and intake of beta-carotene. In contrast, however, when the study group was restricted to men with intake of beta-carotene less than 5,000 (N = 929) or less than 3,000 IU per day (N = 272), comparable relative risks were 1.10 and 1.21, respectively. Although the 95% confidence intervals for these relative risks were broad and included unity, the result is consistent with expectation. We conclude that the hypothesis warrants further investigation.