LAUSR

Abstract Patch size and isolation are traditionally considered as main determinants of species richness in fragmented landscapes, grounded on Island Biogeography Theory (IBT). The Habitat Amount Hypothesis (HAH) is the more recent alternative: species richness could be predicted exclusively by the total amount of habitat surrounding sampling sites. However, tests may be biased towards HAH by the use of poor metrics of patch isolation, and because landscape variables are measured only within the scale of effect for habitat amount. Here we compare the HAH, IBT, and patch isolation as predictors of species richness of forest-dependent small mammals in an Atlantic Forest fragmented landscape using two measures of patch isolation: considering all (overall) or only the nearest three (restrict) forest remnants within the scale of effect for each variable. The model with habitat amount had more support than models with patch size and isolation (representing IBT), or patch size alone, but the model with overall patch isolation was equally plausible. Had we used only restricted patch isolation, we would have found support only for the HAH, disregarding patch isolation. The appropriate metric of patch isolation is critical for robust tests of the HAH, which should be considered in future studies to avoid biased results in favour of the HAH. Our results provide strong evidence for either HAH or overall patch isolation over IBT, and both may offer simplicity to decision-making.