9629876 Holyoak How spatial structure affects populations dynamics, and how habitat fragmentation affects biodiversity are two of the most important questions in modern ecology. In metapopulations of single species, or sets of species that do not interact, subdivision of a habitat is expected to cause collapse to global ("metapopulation") extinction, such that species with the lowest dispersal rates will go extinct soonest. In contrast, metapopulation models of interacting species predict that an intermediate degree of subdivision can promote coexistence of otherwise unstably interacting species. The present study offers a unique opportunity to investigate how habitat subdivision influences the structure and persistence of species assemblages. This work will be conducted in aquatic microcosms containing an assemblage of protozoans in which the pair-wise and multispecies interactions are known. Two alternative hypotheses will be tested: (1) Increased subdivision always decreases the persistence of all species, with the poorest dispersers becoming extinct soonest, and (2) Subdivision promotes the persistence of some or all species, at least up to some intermediate degree of subdivision; this would imply that interactions between species are important. These hypotheses will be tested by measuring the persistence of a variety of species combinations in microcosms with different degrees of subdivision. Dispersal rates will also be measured, so that the sequence in which species go extinct can be recorded. The experiments will show whether the response of assemblages to habitat subdivision can be predicted from the dynamics of the particular species involved, versus the type and details of between-species interactions. This study will demonstrate whether habitat fragmentation (subdivision) promotes biodiversity. In particular, the experiments will test the parts of ecological theory that are most directly relevant to predicting the response of endangered species living in patchy habitats to varying degrees of fragmentation.