If animal cancer tests are reasonably accurate in predicting the probability, sites, and severity of human cancers, then regulation of carcinogens is on firm ground. But if the tests are poor predictors of human cancer, then regulation of exposure to chemicals rests on quicksand.