An essay addressing two questions: First, is it possible to perform analysis--cost benefit analysis--that will provide answers to the questions of how much and where the federal government should make its investment in biomedical research? Second, if outright answers are not possible, is it possible to perform analysis that will lead to better informed judgments as to how much, where, and in what manner the government should invest in biomedical research? The answer to the first question is, "No." Much of the information needed for cost-benefit analysis is not known; some of it will never be known. The answer to the second question is, "Yes." However, there is no single analytic approach that leads to confident biomedical research policy decisions. Instead, there are many questions that can be addressed by many forms of analysis. This study provides a structure for analysis in four broad areas: (1) the predictability of scientific advances, (2) the links between federal program expenditures and scientific advances, (3) the relationship of scientific advances to improved health, and (4) assessments of the value that society places on improved health and longevity.