Because of the economic importance of maize and its scientific importance as a model system for studies of domestication, its evolutionary history is of general interest. We analyzed the population genetic structure of maize races by genotyping 964 individual plants, representing almost the entire set of ∼350 races native to the Americas, with 96 microsatellites. Using Bayesian clustering, we detected four main clusters consisting of highland Mexican, northern United States (US), tropical lowland, and Andean races. Phylogenetic analysis indicated that the southwestern US was an intermediary stepping stone between Mexico and the northern US. Furthermore, southeastern US races appear to be of mixed northern flint and tropical lowland ancestry, while lowland middle South American races are of mixed Andean and tropical lowland ancestry. Several cases of post-Columbian movement of races were detected, most notably from the US to South America. Of the four main clusters, the highest genetic diversity occurs in highland Mexican races, while diversity is lowest in the Andes and northern US. Isolation by distance appears to be the main factor underlying the historical diversification of maize. We identify highland Mexico and the Andes as potential sources of genetic diversity underrepresented among elite lines used in maize breeding programs.