Results of experimental trials with antiviral agents in humans have varied from encouraging to controversial to negative, usually as a result of the difficulty in defining the true therapeutic index (ratio of efficacy to toxicity) of toxic drugs for the treatment of diseases that are potentially severely debilitating or lethal. Reasons for current difficulties relate mainly to inadequacies of preclinical studies and the lack of appropriate controls. The inadequacies include poor definition of the effect of drugs or viruses on cellular metabolism, incomplete pharmacologic studies in animals or humans, and, because of the latter, inappropriate animal models. In human trials, historical data have often been used instead of true controls because of the presumed severity of candidate diseases. Use of such data led to a false impression of drug efficacy, an impression later refuted when proper control studies demonstrated that the range of disease was much greater than had been previously supposed. Data bearing on these points for the most commonly employed experimental compounds (cytosine arabinoside, adenine arabinoside, and 5-iodo-2'-deoxyuridine) are contrasted to highlight difficulties as well as to provide perspectives for antiviral chemotherapy of herpesvirus infections.