There has been a clear trend towards decreased reliance upon animal studies and increased emphasis upon experiments with human-derived tissues. Nonetheless, we continue to need investigations of interspecies differences for two principal reasons: (1) to prospectively design experiments so that the animal species most similar to humans can be chosen, on a case-by-case basis, for each drug; (2) to properly evaluate and interpret data obtained from the experiments ("risk assessment"). Four core examples derived from the work in our FDA laboratory are used to illustrate these points. For paclitaxel, different metabolites were formed in humans and rats, which makes metabolic drug-drug interaction studies in rats irrelevant. For zidovudine (AZT), rapid glucuronidation in humans produced a much shorter half-life than expected from studies in animals, which have negligible glucuronidation. The toxicology and efficacy of both parent drug and metabolite need to be assessed in cases such as iododeoxydoxorubicin, in which the parent molecule is the dominant circulating species in mice, but patients have more than 10-fold greater exposure to the metabolite compared with the parent. While rats have highly-active arylamine N-acetyltransferases, dogs totally lack this enzyme family, and humans have intermediate amounts. For some situations, we've suggested that it can be desirable to inhibit NAT to make the human exposure more similar to dogs. In conclusion, although the ratio of animal:human data is decreasing, our ability to use animal data effectively for drug development has actually increased. Continued focus should be placed upon the application of comparative interspecies data for prospective design of animal experiments and retrospective interpretation of animal findings in terms of the potential for human risk and benefit.