In Hubbell's model of zero-sum ecological drift, loss of species from communities by stochastic change in population size is balanced by the production of new species by processes analogous to mutation or to the fission of populations. Large regional metacommunities contain so many individuals that species are lost slowly and infrequent speciation events can maintain high diversity. However, validation of Hubbell's model requires that community size, diversity, and species life spans match up, and this is shown to be a problem with both the mutation and fission mechanisms of speciation. In the first case, most species are rare and ephemeral and would fail to be recognized by traditional taxonomic practices. In the second case, species life spans are so long that diversity builds to unrealistic levels. Thus, the problem confronting community drift probably is not the maintenance of diversity but rather its control, requiring such mechanisms as environmental change or occasional evolution of competitively superior species that sweep the metacommunity. Testing the community drift model will require close attention to community size and species life spans.