Background/Question/Methods The effects of land cover surrounding forest fragments (termed the “matrix”) on biodiversity remain poorly understood. This landscape component, however, can greatly influence species patterns by altering movement of individuals among patches and by providing differential resources. In this study, we examined how matrix habitats affect bird communities in Jamaica across four landscape types (agriculture, bauxite mining, residential development, and forested). Our approach is novel in isolating effects of matrix composition by controlling for the spatial pattern of primary habitat (i.e., number and distance of forest patches), as well as by selecting matrix types that varied systematically in both matrix connectivity and resource availability in real landscapes. Results/Conclusions Over 400 point count surveys were conducted across 20 1-km2 landscapes. Bird communities differed across all four landscape types. Variance in bird community composition was not strongly correlated with primary habitat quality or spatial structure but rather with matrix conditions. Response varied by trophic guild, with nectarivores, omnivores and frugivores least sensitive and insectivores most sensitive to fragmentation and landscape composition. To explore potential explanatory mechanisms, we considered species-area and isolation relationships – and the extent to which matrix quality mediates these relationships. All fragmented landscapes displayed a significant richness-area relationship. Agricultural and bauxite landscapes exhibited the strongest richness-area relationship, with residential landscapes the weakest, attributed to the fact that residential areas provide additional matrix resources (i.e., gardens). In contrast, only bauxite landscapes displayed a significant richness-isolation relationship, indicating that corridors and garden plots may aid dispersal in agricultural and residential landscapes, respectively. These findings lend crucial empirical support to the longstanding but effectively untested matrix hypothesis that landcover surrounding forest fragments influence species within-patch persistence.