Much of the recent debate over race, genetics, and health has focused on the extent to which typological notions of race have biological meaning. Less attention, however, has been paid to the assumptions about the nature of "populations" that both inform contemporary biological and medical research and that underlie the concept of race. Focusing specifically on Africa in the 1930s and 1940s, this paper explores the history of how fluid societies were transformed into bounded units amenable to scientific analysis. In the so-called "Golden Age of Ethnography," university-trained social anthropologists, primarily from Britain and South Africa, took to the field to systematically study, organize, and order the world's diverse peoples. Intent on creating a scientific methodology of neutral observation, they replaced amateur travelers, traders, colonial administrators, and missionaries as authoritative knowledge producers about the customs, beliefs, and languages of indigenous peoples. At the same time, linguists were engaged in unifying African languages and mapping language onto primordial "tribal" territories. We argue that the notion of populations or "tribes" as discrete units suitable for scientific sampling and classification emerged in the 1930s and 1940s with the ethnographic turn in social anthropology and the professionalization and institutionalization of linguistics in Western and South African universities. Once named and entered into international atlases and databases by anthropologists in the U.S., the existence of populations as bounded entities became self-evident, thus setting the stage for their use in large-scale population genetic studies and the contemporary reinvigoration of broad claims of difference based on population identification.