During years, it has been widely admitted in the scientific community that there was no threshold in mutagenesis: a compound was or not a mutagen. The meaning of such a proposition was that a risk existed at all exposure level, because, at least theoretically, one molecule is sufficient to cause the formation of a DNA adduct which is able to induce a mutation. However, works carried out in the last few years have shown that in the case of some specific mechanisms of mutagenesis, a threshold could be demonstrated essentially in the case of compounds that do not react directly with DNA. Several types of thresholds exist, and the simple statistical threshold is not sufficient in terms of risk assessment. A biological threshold that is consistent with a mechanism of action of the mutagen should be established. Amongst these mechanisms, we can mention some mechanism with a demonstrated threshold: effects of aneugens, effects of topoisomerases inhibitors, effects of DNA polymerases inhibitors, effects of compounds with a different metabolism at high doses compared to low doses. On the contrary, for some mechanisms, the demonstration of the mechanism is suspected, but not totally demonstrated. It is the case of compounds which induce nucleotides pool imbalance or compounds which are DNA repair inhibitors. In some cases, when a redundancy exists in the repair of damages, like oxidative DNA damage, a threshold is suspected. Some authors even recently proposed the possibility of a threshold in the case of alkylating agents. The majority of mutagenic thresholds were demonstrated in vitro, however some mechanisms were demonstrated in vivo, for example in the case of micronucleus induction by hypo or hyperthermia in rodents bone marrow. The use of threshold in risk assessment requires the use of the most sensitive endpoint for example, non disjunction in the case of aneugens, confusing factors like apoptosis should be eliminated and species sensitivities should be taken into account. A very important point to consider is to demonstrate that the mechanism with threshold was really thee only one involved in the mutagenic effect. The demonstration of such thresholds is of particular interest for human risk assessment in the case of mutagens and of genotoxic carcinogens.