The forest refuge hypothesis (FRH) has long been a paradigm for explaining the extreme biological diversity of tropical forests. According to this hypothesis, forest retraction and fragmentation during glacial periods would have promoted reproductive isolation and consequently speciation in forest patches (ecological refuges) surrounded by open habitats. The recent use of paleoclimatic models of species and habitat distributions revitalized the FRH, not by considering refuges as the main drivers of allopatric speciation, but instead by suggesting that high contemporary diversity is associated with historically stable forest areas. However, the role of the emerged continental shelf on the Atlantic Forest biodiversity hotspot of eastern South America during glacial periods has been ignored in the literature. Here, we combined results of species distribution models with coalescent simulations based on DNA sequences to explore the congruence between scenarios of forest dynamics through time and the genetic structure of mammal species cooccurring in the central region of the Atlantic Forest. Contrary to the FRH predictions, we found more fragmentation of suitable habitats during the last interglacial (LIG) and the present than in the last glacial maximum (LGM), probably due to topography. We also detected expansion of suitable climatic conditions onto the emerged continental shelf during the LGM, which would have allowed forests and forest-adapted species to expand. The interplay of sea level and land distribution must have been crucial in the biogeographic history of the Atlantic Forest, and forest refuges played only a minor role, if any, in this biodiversity hotspot during glacial periods.T he extreme biological diversity of tropical forests has inspired and puzzled naturalists and scientists for centuries, and the forest refuge hypothesis (FRH) has long been one of the major paradigms to explain it. According to the FRH, forest retraction and fragmentation during glacial periods would have promoted isolation and consequently allopatric speciation in forest patches, or ecological refuges, surrounded by open habitats in the Amazon (1). Although originally based on climate fluctuations in the Pleistocene, the FRH was subsequently invoked for climate changes irrespective of the time period (2). The FRH was also applied to South America's Atlantic Forest (3), one of the topfive biodiversity hotspots on Earth (4). The FRH gained broad acceptance during the 1980s when empirical paleoecological data from neotropical rainforests were still lacking. Nevertheless, heavy criticism came upon the FRH because some paleobotanical data showed that forests had persisted throughout glacial cycles (5). As paleoclimatic models of species and habitats became widely used, recent studies revitalized the FRH, not by considering refuges as the main drivers of allopatric speciation, but instead by suggesting that high contemporary diversity and endemism are associated with historically stable Atlantic Forest areas (6).This hypothesis is ba...