Since many antiviral substances with potential for use in humans are in various phases of evaluation, criteria must be developed for selection of those compounds with the greatest probability of efficacy and least toxicity. We lack background experience in evaluation of antivirals to permit extrapolation from in vitro tests to use in humans; it is of critical importance, therefore, to develop animal models for evaluation of antiviral substances before trials in humans and to establish guidelines for the relative predictive reliability of in vitro screening and evaluation in animal models. The complexity of drug-host and virus-host interaction and other factors may limit the predictive value of some or all experimental systems. Although the use of animal models is an important phase in the evaluation of antiviral chemotherapeutic agents, the models must be carefully studied and the interaction of drug and virus in the experimental animal specifically defined if optimal guidelines for the predictive value of model systems are to be developed. These guidelines must then be modified as experience is gained with antiviral substances that reach human trials.