The evidence suggests that the present genetic affinities among human populations are the result both of a complex settlement history and of a series of historical events which punctuated that history. Human populations seem to have separated when much of the current genetic diversity was already present. Many alleles are not restricted to one or few populations: they occur in wide geographical areas. Recent processes of gene flow have also helped to spread alleles far from their origins. With the possible exception of societies in which different immigrating groups have not substantially mixed, markers such as stature and skin colour are not useful for the kind of classification that would be needed to identify individuals at increased risk for certain diseases. Although genetic diversity among populations is limited, its pattern reflects both geographical proximity and genealogical relationships between populations; the latter may sometimes be inferred from their current linguistic affinities. Departures from the pattern of similarity between genetic, geographical, and linguistic distances have allowed us to identify, and sometimes to reconstruct, some major colonization and expansion processes. The details may be hard to infer from genetic data, but some agreement between gene diversity measures and archaeological and palaeontological findings is undeniable. Molecular data have allowed us to refine the results obtained from allele-frequency studies. They have confirmed the known relationships and extent of differences between groups of populations defined on linguistic grounds, and in some cases they have allowed us to trace specific mutations along migration pathways, as in the Pacific Ocean or in America. A major achievement has been the demonstration of a recent common origin of all human populations, even though the geographical origin of the ancestral population and its exact genetic composition cannot yet be reconstructed very precisely: East Africa and the gene pool of East Africans are likely candidates. The hypothesis of multiregional evolution is clearly refuted by mitochondrial and Y-chromosome data, as much older molecular ancestors would have been found if more than one branch of Homo erectus gene pools had been transmitted to us. The genetic study of human populations is done through a discrete sampling process, but genetic differentiation is a continuous process in both time and space. Even if we recognize distinct groups and clusters of populations in our analyses, smoother transitions are expected in the field. Our present classification and clustering systems will certainly need to be refined when additional populations are included in the analyses.