In this study, we evaluate diet diversity, diet composition, and risk of colon cancer in an incident population-based study of 1,993 cases and 2,410 controls in the Kaiser Permanente Medical Care Program of Northern California, eight counties in Utah, and the Twin Cities area of Minnesota (United States). Ninety-one and one-half percent of the population were non-Hispanic White. Dietary intake was obtained using an adaptation of the CARDIA diet-history questionnaire. Diet diversity was defined as the number of unique food items reported; diversity also was explored within six major food groups. Composition of the diet was described by estimating the proportion of total number of food items contributed by major food groups. Younger individuals, higher educated individuals, and those who lived in larger households reported eating the most diverse diet. Total diet diversity was not associated with colon cancer. However, eating a diet with greater diversity of meats, poultry, fish, and eggs, was associated with a 50 percent increase in risk among all men (95 percent confidence interval [CI] = 1.1-2.0; P trend = 0.01), with slightly stronger associations for younger men and men with distal tumors. A diet with a greater number of refined grain products also was associated with increased risk among men (odds ratio [OR] = 1.7, CI = 1.3-2.3). Women who ate a diet with a more diverse pattern of vegetables were at approximately a 20 percent lower risk than women who had the least diverse diet in vegetables. Assessment of diet composition showed that men who consumed a large proportion of their food items from meat, fish, poultry, and eggs were at an increased risk, with the most marked association being for distal tumors (OR = 1.7, CI = 1.2-2.5). Women who consumed the largest percentage of their food items in the form of plant foods (fruits, vegetables, or whole grains) were at a reduced risk of developing colon cancer (OR = 0.7, CI = 0.5-1.0).