There is growing recognition that carcinogenic risk extrapolation to low doses (and standard setting) should consider the mode of action of a given chemical. So far, there is agreement on distinguishing between genotoxic and nongenotoxic chemicals; yet, further differentiations seem appropriate. For genotoxic carcinogens, case studies of chemicals point to many possibilities for assessing carcinogenic risk. There are numerous, apparently genotoxic carcinogens where practical thresholds are a matter of discussion. For instance, positive data of chromosomal effects only, in the absence of mutagenicity, may support the characterization of a compound that produces carcinogenic effects only at high, toxic doses. There is a wide consensus that for non-DNA-reactive genotoxicants, such as aneugens, thresholds should be defined. Specific mechanisms of clastogenicity have been repeatedly addressed as also having thresholds, such as topoisomerase II poisons, or mechanisms based on reactive oxygen. These and other arguments together lead to the distinction of four groups of carcinogens, which have been introduced (C. Streffer et al., 2004, Springer-Verlag). There are nonthreshold genotoxic carcinogens (for low-dose risk assessment, the linear nonthreshold [LNT] model appears appropriate); genotoxic carcinogens, for which the existence of a threshold cannot be sufficiently supported (in these cases the LNT model is used as a default assumption, based on the scientific uncertainty and backed by the precautionary principle); genotoxic carcinogens for which a practical threshold is supported; and nongenotoxic carcinogens and non-DNA-reactive carcinogens (for these compounds a true [perfect] threshold is associated with a clearly founded no-observed-adverse-effect level).