Carcinogenesis probably proceeds through a succession of cellular events. Understanding of these events may provide a rational basis for the development of anticarcinogenic treatments. These will be designed to reverse or delay the evolution of a tumor before the stage at which invasion develops. The design and conduct of trials of such agents will be easiest if they are aimed at relatively late stages in the carcinogenic process. Recent research on viral and cellular oncogenes, growth factors, and the cellular mechanism of action of phorbol ester tumor promoters raises the hope that each will be understood and related to the others by their effects on different components of a set of central controls for cellular growth and differentiation. This may ultimately provide a means for rational design of anticarcinogenic treatment. Understanding is still very far from complete, however, and we are still a long way from potential clinical application. An immediate alternative to the long-term rational approach is an empirical one based, for example, on the use of retinoids. The design and interpretation of empirically based trials of anticarcinogenic agents requires careful thought.