The Toxicology and Microbiology Division of the US EPA, Health Effects Research Laboratory has initiated a research program to develop a matrix of short-term tests to distinguish carcinogens from non-carcinogens among genotoxic substances and to develop methods for predicting relative carcinogenic potency based on the results of short-term in vivo bioassays. The paper includes a basic description of the project and outlines the proposed statistical methods for evaluating the Carcinogenesis Testing Matrix (CTM). Successful development, testing and implementation of the test matrix system is very dependent on statistical design, analysis and prediction procedures. Since this is an extension of the usual application of a single screen to predict a qualitative effect, considerable statistical development and testing is needed. Three short-term in vivo bioassays have been proposed for the CTM, the sencar mouse skin bioassay, the mouse lung adenoma bioassay and the liver foci bioassay. Appropriate statistical methods in terms of statistical power and robustness for analyzing the data from each assay are being developed and evaluated.