This review attempts an analysis of the major components which make it extremely difficult to extrapolate toxicological data obtained with chemicals from animals to man. A first problem concerns the use of doses to express the unit of comparison across animal species; the dose is a parameter exogenous to the body and when a chemical enters the body concentrations should be utilized. There is in fact evidence that for several chemicals equal doses in different animal species do not mean equal concentrations in blood or tissues. Concentrations of chemicals should be measured for extrapolation purposes as close as possible to the site of the toxic effect. A second problem regards the fact that several chemical are transformed in the body into other chemical species--sometimes few and sometimes many--and some of these species (active metabolites) display biological activity in some cases higher than different from or antagonistic to those of the parent compounds. Some of these metabolites are highly reactive and therefore bind to body components, particularly macromolecules such as proteins and nucleic acids. There is evidence that metabolism is quantitatively and/or qualitatively different in various animal species. A third problem concerns the difference in various animal species in the biological substrates on which chemicals exert their toxic effects. Equal concentrations of chemicals and their metabolites do not mean equal toxic effects across animal species because endogenous metabolic processes, cell permeability, enzymes, and receptors are not necessarily the same in animals and man. To overcome these difficulties there is a need for closer integration of different disciplines in the toxicological evaluation of chemicals. A scientific rather than a routine approach in toxicology is emphasized.