Background/Question/Methods We modify MacArthur and Wilson's theory of island biogeography, which implicitly assumes ‘passive' dispersal and population dynamics, by adding choice-based immigration and emigration functions that both vary with isolation. We fit the models to a large number of existing biogeographic datasets. Results/Conclusions We show that in many cases, the modified models are preferable to the original model even given their increased complexity. For many datasets involving birds, there is evidence for a strong tendency to leave those patches that are closest to larger patches — a tendency that overwhelms the original model's assumption of a higher (passive) colonization rate and results in expectedly impoverished species richness. If real, this phenomenon has disturbing implications for conservation-oriented landscape planning. Habitat corridors, in particular, might in some circumstances have the opposite effect from the one intended, leaving connected fragments worse off than before.