Diet is believed to modulate cancer risk and this relationship has been widely studied in the gastrointestinal tract. Observational epidemiologic studies have provided most of the evidence about the effects of diet on cancer risk because clinical trials to determine nutritional exposures are often impossible, impractical, or unaffordable. Although a few foods or nutrients are thought to protect against specific types of cancer, it seems clear that the strength and even direction of dietary associations (increasing or decreasing risk) is organ-site- and even histology-specific, along the gastrointestinal tract. Although some hypotheses are supported by a substantial body of observational data (drinking hot maté [an infusion of the herb Ilex Paraguarensis] contributes to esophageal cancer), there are not much data to support others. We discuss some highly touted hypotheses and draw interim conclusions about what is known and what could be done to improve the level of evidence. The complex nature of diet and its associations can be productively investigated with disease-specific studies. However, public health recommendations for normal-risk individuals regarding diet and gastrointestinal cancer should probably emphasize the importance of eating for overall health rather than eating specific foods to reduce risk for specific cancers.