Data for relationships between in vivo doses inferred from levels of hemoglobin (Hb) or DNA adducts and administered (by inhalation or injection) doses of ethylene oxide (EO) in mice, rats and humans are reviewed. At low absorbed doses or dose rates these relationships appear to be linear, whereas at higher dose rates deviations from linearity due to saturation kinetics of detoxification and of DNA repair as well as certain toxic effects have to be allowed for. If these factors are taken into consideration, a rather consistent picture is obtained for animal studies, with a variation by less than a factor 2 between estimates of adduct level increments or in vivo dose increments per unit of administered dose. Although the value for in vivo dose per unit of exposure dose (ppm-hour) in humans is uncertain because of unreliable data for the time-weighted average exposure level, the most likely value for this relationship, supported by data for ethene, agrees with data for the rodents. In the animal species testis doses are approximately one-half of the blood doses inferred from Hb adducts.