Many scientific and statistical issues are involved in generalizing the results of randomized clinical trials. This paper examines the issues of heterogeneity of risk and heterogeneity of effect, as well as quantitative vs. qualitative heterogeneity. Generalization of quantitative results to populations at different risk is shown to be problematic, independent of any other considerations. Issues of heterogeneity as they relate to target population and choice of therapy are discussed. A few proposals are offered regarding the design and report of trials.