Dietary data relating in 406 patients with gastrointestinal cancers (cases) and 812 controls have been analysed to test the hypotheses that dietary vitamin A (retinol) and its precursor, carotene, may be protective agents in human carcinogenesis. There was no deficit among cases in the intake of retinol-containing food items whereas several of the main carotene-containing fruits and vegetables were eaten less often among cases than among controls. However, when total daily intake was estimated, there was no protective effect of carotene, perhaps implying that some other constituent of the fruits and vegetables is protective.