HIV-1 is one of the fastest evolving entities known. Given that census population sizes of HIV-1 within individuals are much greater than the inverse mutation rate, every possible single point mutation in the viral genome occurs each generation. This enormous capability to generate genetic variation allows for escape from immune surveillance and antiviral therapy. However, compared to this potential, populations of HIV-1 within individuals exhibit little genetic variation. This discrepancy between the known mutation rate of HIV-1 and the average level of genetic variation in the env gene observed in vivo is reflected in comparisons of the actual numbers of productively infected cells, estimated as 107, and the effective population size, estimated as 103. Using approximate Bayesian computation, we evaluated several hypotheses based on a variety of selective and demographic processes to explain the low effective population size of HIV-1. Of the models we examined, the metapopulation model, in which...