If species interactions affect species distributions, then species occurrence frequencies (nu(i)), defined as the fraction of an archipelago's islands that species i inhabits, should vary with species pool size. A "natural experiment" approximating this test is provided by the Bismarck, Solomon, and New Hebrides archipelagoes, whose bird species pools decrease in that order, the species of each archipelago being mostly a subset of those of the next richer archipelago. The average nu for an archipelago's species decreases with archipelago pool size. In the archipelago with the largest pool, most species are on few islands and few species are on most islands, whereas the reverse is true in the archipelago with the smallest pool. For species shared between two or more archipelagoes, nu(i) decreases with pool size or number of species in the same guild. These interarchipelagal differences in nu(i) or average nu reflect differences in level of interspecific competition, which reduces nus in species-rich archipelagoes in two ways: usually, by reducing a species' incidence on small islands and restricting the species to larger islands; less often (for so-called supertramps), by restricting a species to small islands.