ABSTRACT Statistical tests of the neutrality hypothesis that are based on the sampling theory of Ewens (1972) require the assumption of panmixia. It is proposed that for a population comprising numerous local populations with weak gene flow among them, tests based on Ewens' theory can be applied separately to samples from each local population. At low levels of gene flow, migration acts primarily like mutation, introducing new alleles to each local population. It is shown with simulation results that, at low levels of migration, correlations in allele frequencies among demes are sufficiently small that the results from the application of Ewens' theory to each deme are statistically independent. It is also shown that, by combining the results of the tests in different demes, some statistical power to detect deviations from neutrality is gained. The method is illustrated with the application to data on a salamander species. At low levels of gene flow, population subdivision must be taken account of in testing neutrality and the proposed test provides one way to do so.