9870766 Richman Self-recognition systems including the immune system in vertebrates and the self-incompatibility locus in flowering plants exhibit spectacular genetic polymorphism maintained by selection. The proposed work will test recent models of self-recognition genes which predict that more distantly related alleles will be favored by selection. This prediction will be examined in island populations of both animals and plants which have undergone a dramatic reduction in population size following isolation of islands from the mainland due to post-Pleistocene sea level rise. The premise for studying island populations is that non-random loss of alleles will be most evident when the strength of diversifying selection is reduced, as expected for newly isolated populations which will lose genetic variation over time. If relatively distantly related alleles are favored during subsequent reduction in allelic variation, this will be detected by comparison of allelic genealogies for island and mainland populations. If more divergent alleles are favored by selection this has implications for both models of the evolution of the immune system and recognition of disease antigens, and for evolutionary studies which attempt to infer the history of populations from the diversity of self-recognition genes. For example, preservation of divergent alleles by selection supports models of the immune system that predict more divergent alleles are able to present a wider variety of potential disease antigens to the immune system, conferring enhanced resistance to the host. This result would also have important implications for recent studies of self-recognition genes in humans which conclude that population size has been large over the history of diversification of the primate lineage. If more divergent allelic specificities are preferentially maintained, these studies will tend to overestimate the size of populations. Since these estimates are present ed as lower limits on population effective size, this potential bias may substantially alter the conclusions reached.