We examined the overall results of 124 consecutive rodent carcinogenesis assays carried out at the maximum tolerated dose on 37 chemicals reported recently by the Toxicology Program of the United States. In 31 experiments each in male and female F-344 rats and in male and female B6C3F1 mice, tumor increases and decreases occurred in 41 and 46% of the experiments, respectively. In 22 experiments both increases and decreases in tumor incidence were reported. Of the experiments with decreases in tumor incidence, about 70% were associated with lower body weights of the treated animals. However, of the 30 chemicals producing some tumor decreases, 12 showed decreases in some experiments without any association with bodyweight. Ten chemicals that were Salmonella positive produced increases and decreases in tumor incidences and three produced only decreases in tumor incidence. If it is considered that the bioassay provides information relevant to the carcinogenic potential of a chemical, then logically it must also be considered that information about the cancer-preventive potential of a chemical is provided. When a chemical causes increases and decreases in tumors, several questions follow. First, which are more relevant to the health of an exposed individual: tumor increases or tumor decreases? Second, should such a chemical be stigmatized as a "carcinogen," in view of all the legal and economic implications that ensue from such a label? Third, how should one define a carcinogen?