Analogously to an earlier study using methylnitrosourea (MNU) the prenatal-toxic risk of low doses of ethylmethanesulfonate (EMS) was estimated using different procedures, comparatively. First, the risk of low doses was estimated using linear extrapolation to zero. When using the variable "all gross structural abnormalities" the lowest effective dose in the experiment was 150 mg/kg body wt (5.6% incidence), the additional risk over background was calculated to be 5.0%, and the hypothetical incidence 0.1% was associated with the dose 3 mg/kg EMS. When evaluating "gross structural limb abnormalities," which are not observed in controls, the dose associated with the hypothetical incidence 0.1% was 17.4 mg/kg EMS. Furthermore, derived from a dose-response study of teratogenicity extrapolation to the possible risk of low doses was performed using nonlinear mathematical models. In this case, the results obtained are dependent on the dose response variable as well as from the statistical approach which was chosen. As an example, the values obtained from one evaluation are given: all gross structural abnormalities, Weibull transformation, jackknife approach: ED0.1% = 72 mg/kg EMS. For comparison a "virtually safe dose" was calculated by use of the no-observed-adverse-effect-level (NOAEL) risk factor approach. The NOAEL under our experimental conditions was 100 mg per kg body wt. By using an arbitrarily chosen risk factor of 100 a "safe dose" of 1 mg EMS per kg body wt was obtained. In addition, molecular dosimetry of the DNA adduct rate of O6-ethylguanine in the 11-day-old embryos was used. Based on the assumption that a linear correlation exists between this specific adduct rate and the incidence of teratogenic effects, the hypothetical incidence of 0.1% was associated with a dose of 99 mg/kg EMS. This value is quite similar to that obtained by extrapolation using probit analysis which is in contrast to the results obtained with MNU.