Descriptions of human genetic variation given thirty years ago have held up well, considering the substantial accrual of DNA sequence data in the interim. Most importantly, estimates of between-group genetic variation have remained relatively low. Despite the low average level of between-group variation, clusters recently inferred from multilocus genetic data coincide closely with groups defined by self-identified race or continental ancestry. This correspondence implies that genetic factors might contribute to unexplained between-group phenotypic variation. Current understanding of the contribution of genes to variation in most complex traits is limited, however. Under these circumstances, assumptions about genetic contributions to goup differences are unfounded. In the absence of detailed understanding, 'racial' and 'ethnic' categories will remain useful in biomedical research. Further, we suggest approaches and guidelines for assessing the contribution of genetic factors to between-group phenotypic differences, including studies of candidate genes and analyses of recently admixed populations.