Species movement and occupancy of habitat patches are dependent on landscape permeability. Some land-use types (e.g., roads) may be barriers to animal movement. Analyses of the effect of landscape structure on patch occupancy usually use circular buffers around focal patches. The main assumption of this methodological approach is that species are affected by a particular landscape element equally in every direction from a given patch. This assumption is likely not to hold if animal movement is restricted by barriers because barriers reduce movement patterns and reshape the ideal circular buffer into a noncircular buffer. We developed a method to determine the effect of landscape variables on the distribution of two amphibian species that explicitly takes dispersal barriers into account. We extracted landscape variables within (i) circular buffers (CB) and (ii) barrier-based buffers (BBB). BBB were produced by reducing the boundaries of CB according to major impassable barriers. The BBB approach almost doubled the explained deviance of multiple regression models in comparison with the CB approach. Moreover, CB and BBB models included different predictor variables. We suggest that the BBB approach is more useful than the traditional CB analyses of species–habitat relationships because ecological barriers are explicitly taken into account.