A case-control study probing the role of diet on the incidence of colorectal cancer was undertaken in Athens, Greece, in a population characterized by ethnic homogeneity but substantial heterogeneity with respect to dietary habits. The case series consisted of 100 consecutive patients with histologically confirmed colorectal cancer admitted to two large hospitals of Athens during a 16-month period; the control series consisted of orthopaedic patients, admitted to the same hospitals during the same time period, individually matched to the index cases by age and sex. Dietary histories concerning the frequency of consumption (per month or per week) of about 80 food items were obtained by the same interviewer. Cases reported significantly less frequent consumption of vegetables (particularly beets, spinach, lettuce and cabbage) and, independently, significantly more frequent consumption of meat (notably lamb and beef). Between the two extremes (high-vegetable, low-meat diet versus high-meat, low-vegetable diet) a risk ratio of about 8 appears to exist, sufficient (in size and direction) to explain a substantial part of the international variation in the incidence of colorectal cancer. Significant associations were not found with beer or other alcoholic beverages, and significant interactions were not noted with respect to age, sex and anatomic localization (colon vs. rectum).