Unique aspects of human behavior account for special features of human evolution. For instance, the best data on the mating patterns of undisturbed hunting-gathering populations are those of Norman Tinsdale for Australian Aborigines. Among some 574 linguistically distinct tribes, he collected information on 755 marriages from the period prior to significant contact with Europeans. One can extrapolate from these data on 1510 individuals. In an average tribe of some 500 individuals, 62-65 of them would have had one parent who was a member of a different tribe and 7 or 8 of them would have been offspring of a parent from a distant tribe, not an adjacent one. Such rates of intergroup marriage, generation after generation, would have produced considerable gene flow, a pattern that likely has occurred during most of human prehistory. The pattern of descent is trellis-like, not one of successive fissioning. Some genetic diversity between tribes must have existed, but most genetic variation would have been within rather than between tribal groups. One implication of this pattern is that ethnically defined groups are not a suitable basis for studying human genetic diversity. Geographically stratified random sampling of the species would be more likely to ensure an unbiased estimate of genome variability.