1. Methods of interspecies extrapolation using physiological models and allometric scaling have been reviewed with their possible application to drug development, both for candidate drug selection and the interpretation of toxicokinetic data. 2. Physiological models offer a mechanistic approach to extrapolation from one species to another, examining individual components which interrelate to produce the characteristics of the whole system. Tissues of interest are arranged in anatomical order based on blood circulation, and the disposition of a drug can be simulated with knowledge of tissue size (volume), tissue perfusion (blood flow), drug permeability, binding of the drug between the tissue and blood (partition), as well as elimination. Using this approach the behaviour of the drug under different conditions, such as dose route, disease state or animal species, can be predicted. 3. The alternative approach of allometric scaling is an empirical examination of relationships between size, time and its consequences. A regression of the logarithm of the pharmacokinetic parameter and the logarithm of the body weight of the animal species produces a linear relationship which enables the value of pharmacokinetic parameters in any animal species to be calculated from the product of an allometric coefficient and the body weight to a power function. 4. Whilst this technique gives acceptable predictions for the pharmacokinetics of those drugs eliminated renally, or which are blood flow-dependent, there is poor prediction for humans for low clearance drugs primarily eliminated by the mixed-function oxidase system. This appears to be a result of differences in maturation, and can be corrected for by including a brain weight or maximum life-span potential term into the allometric equation. 5. Of the two approaches described for extrapolation of pharmacokinetics between animal species, physiological models tend to be resource-demanding and costly, with more frequent failures, but can be invaluable for examining target organ exposure and for the targeting of drugs as in cancer chemotherapy. For routine drug development, however, allometric scaling is potentially more useful since it uses data which are routinely obtained and the calculations are relatively simple. 6. The problems of intraspecies scaling from high-dose data to low-dose predictions are discussed with respect to current models of dose levels. A new approach is proposed using a modified Hill equation based on drug exposure, which should allow for a more meaningful determination of the toxicity of a compound with different drug exposures.(ABSTRACT TRUNCATED AT 400 WORDS)