A critical evaluation of the use of statistical models in carcinogenic risk assessment is made with emphasis on the strengths and weaknesses of current practice. The objective is to bring together information from the fields of toxicology and statistics to develop a sound scientific basis for making risk decisions. It is emphasized that risk assessment is a complex, multifaceted process that is not easily quantified and, at present, must be based on qualitative as well as quantitative information. It is evident that dose-response modeling of tumor incidence data from animal studies is of limited value in estimating human risk associated with low-dose chemical exposures. The models take into account only one part of the complex process and have no accepted basis in biology. The key decision points, qualitative factors, and quantitative considerations are identified and discussed and a risk assessment framework that incorporates these inputs is presented. A model that incorporates the tumorigenic dose-response information as well as the qualitative and quantitative biological factors that affect the estimate of risk is proposed. It is concluded that much work needs to be done before a completely quantitative approach to risk assessment is to be useful; in particular pharmacokinetic modeling should be pursued more aggressively.