We present a scheme for classifying chemical carcinogens according to the weight of the evidence that each substance poses a human cancer hazard. The approach represents a logical extension of and builds upon those previously developed by the International Agency for Research on Cancer, the U.S. Environmental Protection Agency, and the so-called Tripartite Group of industrial scientists. It takes into account new scientific knowledge about chemical carcinogenesis and animal models. Eight categories are presented: known human carcinogen (Category 1), carcinogenic activity in animals, probable human carcinogen (Category 2), possible human carcinogen (Category 3), equivocal evidence for carcinogenic activity (Category 4), evidence inadequate for classification (Category 5), carcinogenic activity in animals; probably not a human cancer hazard (Category 6), carcinogenic activity in animals; considered not a human cancer hazard (Category 7), evidence of noncarcinogenicity (Category 8). Evidence useful for categorization includes human studies, animal bioassays, corroborative evidence from bioassays, and mechanistic studies relevant to determining the predictivity of animal responses for human hazard. Weighing this evidence to derive a conclusion about classification is a process that requires expert judgment; it cannot now be reduced to a simple set of decision rules. However, we identify the kinds of information that can be useful in this process, and indicate how each might most appropriately be used.