The association between dietary intake of vitamin A and lung cancer risk was examined in a population-based, case-control study of 447 patients and 759 control subjects in New Mexico. A food frequency interview was used to measure usual consumption of total vitamin A retinol, preformed vitamin A, and carotene. With all respondents combined, the odds ratios for lung cancer increased as intakes of total vitamin A and carotene declined but did not vary with intake of preformed vitamin A. When the subjects were stratified by ethnic group, Hispanic or non-Hispanic (Anglo) white, significant effects of vitamin A consumption were limited to the Anglos. In the Anglos, the protective effects of total vitamin A and carotene consumption were present in males and females, but varied strongly with cigarette smoking habits. In Anglo smokers, significant increases in the odds ratios with declining intake were observed in former but not in current smokers. Among the former smokers, significant effects of total vitamin A and carotene consumption were present only in those who had stopped smoking for 6 to 15 yr. Limitation of the protective effect of vitamin A and carotene consumption to past smokers has important implications for the design of clinical trials and for cancer control strategies.