B[a]P (benzo[a]pyrene) has been used as a prototype carcinogenic PAH since its isolation from coal tar in the 1930's. One of its diol epoxides, BPDE-2, is considered its ultimate carcinogen on the basis of its binding to DNA, mutagenicity and extreme pulmonary carcinogenicity in newborn mice. However, BPDE-1 has a similar binding to DNA and mutagenicity but it is not carcinogenic. In addition, BPDE-2 is a weak carcinogen relative to B[a]P when repeatedly applied to mouse skin, the conventional assay site. Its carcinogenicity is increased when applied once as an initiator followed repeatedly by a promoter. This indicates a major role for promotion in carcinogenesis by PAHs. Promotion itself is a 2-stage process, the second of which is selective propagation of the initiated cells. Persistent hyperplasia underlies selection by promoters. The non-carcinogenicity of BPDE-1 has yet to be resolved. PAHs have long been considered the main carcinogens of cigarette smoke but their concentration in the condensate is far too low to account by themselves for the production of skin tumors. The phenolic fraction does however have strong promotional activity when repeatedly applied to initiated mouse skin. Several constituents of cigarette smoke are co-carcinogenic when applied simultaneously with repeated applications of PAHs. Catechol is co-carcinogenic at concentrations found in the condensate. Since cigarette smoking involves protracted exposure to all the smoke constituents, co-carcinogenesis simulates its effects. Both procedures, however, indicate a major role for selection in carcinogenesis by cigarette smoke. That selection may operate on endogenous mutations as well as those induced by PAHs. There are indications that the nicotine-derived NNK which is a specific pulmonary carcinogen in animals contributes to smoking-induced lung cancer in man. Lung adenoma development by inhalation has been induced in mice by the gas phase of cigarette smoke. The role of selection has not been evaluated in either of these cases.