This paper argues that racial classification and generalization may sometimes be justified in clinical treatment and research, in part to ensure better outcomes for the individual patients subject to such classification and generalization; in part to enable medicine to eliminate the need for racial categories. But race must be used carefully and sparingly because of the risk to the individual patient of overgeneralization, and the risk to society of reinforcing a false understanding of race as a biological category. Even if the use of racial categories in biomedical research subverts rather than reinforces those categories, that research may well lead to the recognition of non-racial, genetically based groups, which will be susceptible to harmful, if less invidious, stereotyping. Law, Probability and Risk (9:215−226 2010)