As we consider the epidemiologic evidence on diet-cancer associations, we should keep in mind the geographical differences in cancer incidence that stimulated us to look for a diet-cancer link in the first place. For many cancers, the variation in geographic incidence is in the range of an order of magnitude (see Table 2). Thus, if variation in dietary components are primarily responsible for these huge geographic variations in cancer incidence, then the effect of dietary components will not be subtle. Based on the magnitude of geographic variation, one might be primed to look for large differences in cancer incidence associated with different dietary intakes--relative risks of two or greater. This is not to say that a true relative risk of 1.2 or 0.8 for cancer associated with some dietary habit is not potentially important, but it will not get us very far along the way of explaining the geographic variation. The recommendation of the National Research Council is the consumption of five servings of fruit and vegetables daily. This level of consumption would place one in the "high consumption" group in most of the epidemiological studies showing a strong association between fruit and vegetable intake and cancer. Unfortunately, only 10% of the U.S. population actually consumes the recommended amount. Conversely, 20% to 35% of the U.S. population consume levels of fruits and vegetables that put them in the "low consumption" groups. Although causality cannot be established, the evidence is overwhelming that some constituent(s) of fruits and vegetables is (are) protective; low intake carries a greater than two-fold risk of cancer. The risk of cancer associated with low fruit or vegetable consumption may only be exceeded by that of smoking. In the typical diet in the United States, the main source of antioxidant nutrients vitamin C, carotenoids, and vitamin E comes from fruit and vegetable consumption. Without a complete understanding of the mechanism for the effect of fruits and vegetables or antioxidants, the epidemiologic evidence strongly supports the clinical benefits that can be achieved by promotion of consumption (Table 5).