In the BUPA Study, a prospective study of 22,000 men attending a screening centre in London, serum samples were collected and stored. The concentration of beta-carotene was measured in the stored serum samples from 271 men who were subsequently notified as having cancer and from 533 unaffected controls, matched for age, smoking history and duration of storage of the serum samples. The mean beta-carotene level of the cancer subjects was significantly lower than that of their matched controls (198 and 221 micrograms l-1 respectively, P = 0.007). The difference was apparent in subjects from whom blood was collected several years before the diagnosis of the cancer, indicating that the low beta-carotene levels in the cancer subjects were unlikely to have been simply a consequence of pre-clinical disease. Men in the top two quintiles of serum beta-carotene had only about 60% of the risk of developing cancer compared with men in the bottom quintile. The study was not large enough to be able to indicate with confidence the sites of cancer for which the inverse association between serum beta-carotene and risk of cancer applied, though the association was strongest for lung cancer. The association may be due to beta-carotene affecting the risk directly or it may reflect an indirect association of cancer risk with some other component of vegetables or with a nonvegetable component of diet that is itself related to vegetable consumption.