Studies have shown that a majority of known human carcinogens also cause cancer in laboratory animals. The converse, however, is not as well established—known animal carcinogens are not equally predictive of human carcinogenicity. A particularly controversial aspect of interspecies extrapolation is application of rodent forestomach tumor data for predicting cancer risk in humans, given that a human counterpart for the rodent forestomach does not exist. Proliferative lesions in the rodent forestomach may result from a combination of factors related to route-specific tissue irritation and/or unnatural dosing regimens and are less likely to be relevant in evaluating human carcinogenic potential, particularly when tumors are exclusive to the forestomach. We review the comparative functional anatomy, physiology, tumor biology, tissue concordance, and historical regulatory practices in the use of forestomach tumors for cancer risk assessment, examining specific chemical examples. We also propose a standardized mode-of-action approach that combines multiple risk characterization criteria, including relevance to human exposure conditions, physiologically based toxicokinetics, genotoxicity, and comparative/mechanistic toxicology. Forestomach tumors associated with chronic irritation of the forestomach epithelium, particularly those induced by repeated oral gavage dosing, should not form the basis for carcinogenic classification or quantitative cancer potency estimates for humans. Genotoxic chemicals and those that cause tumors at multiple sites, at doses at or below the maximum tolerated dose, and in the absence of forestomach irritation, are more likely to be relevant human carcinogens. Cancer risk assessment that utilizes forestomach tumor data should consider relevant human exposures, systemic bioavailability, tissue dosimetry and concordance.