There are a variety of ways in which we learn about the relationship of diet to cancer prevention or cancer etiology in humans. Epidemiologic studies provide the basis for many of our hypotheses, while animal studies provide opportunities to try out more specific models. However, only human clinical trials can adequately test these concepts and identify which food compounds are related to cancer prevention. Further, clinical trials can provide opportunities to examine underlying mechanisms. Clinical trials in the area of diet and cancer prevention have largely been of two types: single (or dual) nutrients (purified, in high-dose pills) tested in double-blind placebo-controlled trials, and randomized dietary interventions, using education of the subjects to achieve broad scale dietary change. This paper argues for an intermediate form of clinical trial, using a food-based supplement in a randomized double-blind placebo-controlled design. This model is particularly appropriate for studying food components such as fiber for which the activity may rely on higher order structure.