Breast cancer, the most frequent spontaneous malignancy diagnosed in women in the western world, is continuously increasing in incidence in industrialized nations. Although breast cancer develops in women as the result of a combination of external and endogenous factors such as exposure to ionizing radiation, diet, socioeconomic status, and endocrinologic, familial, or genetic factors, no specific etiologic agent(s) or the mechanisms responsible of the disease has been identified as yet. Thus, experimental models that exhibit the same complex interactions are needed for testing various mechanisms and for assessing the carcinogenic potential of given chemicals. Rodent mammary carcinomas represent such a model to a great extent because, in these species, mammary cancer is a multistep complex process that can be induced by either chemicals, radiation, viruses, or genetic factors. Long-term studies in rodent models have been particularly useful for dissecting the initiation, promotion, and progression steps of carcinogenesis. The susceptibility of the rodent mammary gland to develop neoplasms has made this organ a unique target for testing the carcinogenic potential of specific genotoxic chemicals and environmental agents. Mammary tumors induced by indirect- or direct-acting carcinogens such as 7, 12-dimethlbenz(a)anthracene or N-methyl-N-nitrosourea are, in general, hormone dependent adenocarcinomas whose incidence, number of tumors per animal, tumor latency, and tumor type are influenced by the age, reproductive history, and endocarinologic milieu of the host at the time of carcinogen exposure. Rodent models are informative in the absence of human data. They have provided valuable information on the dose and route of administration to be used and optimal host conditions for eliciting maximal tumorigenic response. Studies of the influence of normal gland development on the pathogenesis of chemically induced mammary carcinomas have clarified the role of differentiation in cancer initiation. Comparative studies with the development of the human breast and the pathogenesis of breast cancer have contributed to validate rodent-to-human extrapolations. However, it has not been definitively established what type of information is necessary for human risk assessment, whether currently toxicity testing methodologies are sufficient for fulfilling those needs, or whether treatment-induced tumorigenic responses in rodents are predictive of potential human risk. An alternative to the traditional bioassays are mechanism-based toxicology and molecular and cellular approaches, combined with comparative in vitro systems. These approaches might allow the rapid screen of chemicals for setting priorities for further studies to determine the dose-response relationship for chemical effects at low doses, to assess effects other than mutagenesis and/or tumorigenesis, or to establish qualitative and quantitative relationships of biomarkers to toxic effects. Until there is enough information on the predictive value of mechanism-based toxicology for risk assessment, this approach should be used in conjunction with and validated by the traditional in vivo long-term bioassays.