Abstract A common situation observed in fragmented habitats is that species densities diminish within smaller fragments. Some species, however, do show an opposite tendency. We argue that release of competition between a strong competitor and a weak one is a plausible explanation for these different sensitivities to area reduction. We provide a quantitative model for competitive release caused by habitat limitation, solely based on the balance of diffusion and growth of the species. We show that in small habitat patches the stronger competitor has its density diminished, as opposite to the weaker competitor who, in a certain area range, has its density increased. We examine field data from an ecological experiment in Amazonia (BDFFP) which measured densities of two Amazonian rodents who showed opposite sensitivity to area reduction and we argue that our model explains the observations accurately. This implies that (i) area reduction is a strong factor determining densities of species in patches of habitat, regardless of considerations on degradation or edge effects and (ii) that species interactions have to be taken into account to explain sensitivity to size of the patches in ecological communities. We also discuss alternative explanations, such as predator release and effects due to imperfect isolation. Moreover, we stress the conceptual and mathematical simplicity of our model, which, nevertheless explains a phenomenon not yet well understood.