It is known that carcinogens designated on the basis of longterm animal test results are extremely diverse in character, both in terms of potencies and the mechanism of action, which leads to complexity in their assessment for cancer risk to humans. The classification of carcin0ogens into two categories, namely, genotoxic and non-genotoxic varieties has been proposed to give a longical foundation on which cancer risk assessment can be reasonably based. The term "genotoxic carcinogen" indicates a chemical capable of producing cancer by directly altering the genetic material of target cells, while "non-genotoxic carcinogen" represents a chemical capable of producing cancer by some secondary mechanism not related to direct gene damage. This classification has contributed to the exclusion of various rodent-specific carcinogens from the group of chemicals with potential cancer risk to humans. However, the term, "nongenotoxic carcinogen" tends to give the mistaken impression that carcinogens shown to be negative for mutagenicity in a series of test systems might be harmless to humans. It should be realized that clear-cut criteria for this classification have not been established because of insufficiencies in the available information concerning mechanisms of action of non-genotoxic carcinogens. Future scientific advances leading to elucidation of the subcellular mechanisms of carcinogenesis are necessary for establishment of the unified, more realistic and mechanism-based approach to cancer risk estimation form exposure to chemicals.