Procedures to reveal ‘immunotoxic’ potentials of chemicals in animal experiments (mostly in rodents) have been recommended, but the selection of test systems is rather arbitrary. The predictive power of extrapolations to the possible situation in humans is unknown because human studies to confirm or to reject clues from animal data are largely lacking. End points selected in animal studies and those expected to be relevant in humans are not identical. Results of animal experiments are based on doses, generally ignoring the important species differences in pharmacokinetics. This unfavorable situation is especially pronounced when attempting to evaluate ‘environmental chemicals’. Because much more information is available on many medicinal drugs, exposures can be defined and pharmacokinetic data are available or obtainable. The situation is even more complicated when attempting to assess possible adverse effects on the developing immune system: in addition to the problems mentioned, numerous different developmental periods with varying susceptibilities must be considered, and species differences in the immune response are superimposed with large differences in pre-, early post-, and later postnatal development. Simultaneously, the kinetic variables are continuously changing with time (with additional variability among species). Different results, even between rats and mice, are bound to occur. Extrapolation to the situation possibly relevant for human exposure will be almost impossible, especially from rodent data. The majority of such effects induced peri-or early postnatally may be expected to be reversible. It must also be assessed whether qualitatively different adverse effects are likely to be induced during ‘development’, which cannot be revealed (accepting quantitative differences) by more easily performed tests on the adult organism. Considering the intrinsic difficulties, the most promising approach would be to directly obtain data from human trials. This is feasible for medicinal drugs. Alternatively, data on nonhuman primates, the species phylogenetically closest to man, may provide useful information. The status quo for such a strategy and the possible pitfalls are discussed in this overview.