There are several approaches to evaluate the genetic effects of low doses of ionizing radiation in a human population. Based on experiments on mice using high doses, the genetic risk for humans has to be derived by extrapolation to low doses, assuming mouse and man respond in a similar way. Microdosimetric considerations indicate a linear relationship; studies on the molecular mechanism of radiation-induced mutagenesis lead to the conclusion that, in addition to a constitutive mutagenic process, an inducible compound exists which may cause linear-quadratic dose-effect curves. Linear extrapolation seems thus to be justified; however, the genetic risk of low doses may be slightly overestimated.