Document Type



Previous attempts to classify South American Indian tribes according to genetic characteristics have failed to yield a hierarchical system of relationships. This can be explained by the facts that (1) tribal populations did not evolve through sequential fissions but through frequent fusions of groups with diverse histories and (2) allele frequencies have been held at nearly common values by intertribal migration or balancing selection. A valid model must allow for fusion and mixed populations as well as for fission; factor analysis or newer methods of fuzzy mathematics permit this. The effects of migration and balancing can be made more manageable by partitioning them according to the limited time periods recorded by haplotypes. An initial attempt using factor analysis and HLA haplotype data on 19 rain forest tribes revealed two overlapping clusters that are largely but not neatly separated by the lower Amazon River. Several tribes, especially in the west, were excluded from these clusters.