In the evaluation of long-term rodent carcinogenicity studies, many different tumor sites and types are evaluated, which may increase the likelihood of a statistical false positive. To deal with this issue, a number of statistical decision rules have been proposed that take into account multiple comparisons. This paper discusses the various types of decision rules and evaluates the factors that may lead to different interpretations of experimental results. These concepts are illustrated by examining the statistical decision procedures used by three analysts to evaluate the results of 25 long-term rodent carcinogenicity studies carried out by the National Cancer Institute. Agreement among these decision rules is shown to be greater than originally reported. It is also concluded that while the application of statistical decision rules may be of value in some instances to guard against statistical false positives, the final interpretation of the data should be based on biological as well as statistical considerations. Thus, statistical decision rules should not be employed as a substitute for sound scientific judgment in the overall evaluation of these experiments.