Use of laboratory animal data to make quantitative predictions of the risks of toxic effects in humans assumes that a relationship exists between the potencies in animals and humans and that its parameters can be estimated adequately. Such "scaling rules" have been used to predict the risks of carcinogenicity or other effects. A survey of the literature yielded only a modest number of papers devoted to the validity of these interspecies risk extrapolations, of which approximately 25 attempt quantitative comparisons for either radiation or chemical hazards. Some authors have investigated relatively large data sets in an attempt to identify the scaling rule that provides the best correlation of risks in two or more species. Others have selected a scaling rule and investigated whether its predictions from data in laboratory species match the risks found in humans. Opinion is divided on the validity of specific extrapolation rules and the utility of animal experiments for quantitative risk assessment. Correlations exist among risk levels in various species, but many factors appear to influence toxicity that are not captured in a simple scaling rule such as dose per unit weight or per unit surface area. Although scaling rules are useful, better projections will be made if case-specific factors such as pharmacokinetics can be considered. Further careful comparisons of quantitative risk estimates are needed.