Pleistocene fragmentation of the Amazonian rainforest has been hypothesized to be a major cause of Neotropical speciation and diversity. However, the role and even the reality of Pleistocene forest refugia have attracted much scepticism. In Amazonia, previous phylogeographical studies have focused mostly on organisms found in the forests themselves, and generally found speciation events to have predated the Pleistocene. However, molecular studies of open-formation taxa found both north and south of the Amazonian forests, probably because of vicariance resulting from expansion of the rainforests, may provide novel insights into the age of continuous forest cover across the Amazon basin. Here, we analyse three mitochondrial genes to infer the phylogeography of one such trans-Amazonian vicariant, the Neotropical rattlesnake (Crotalus durissus), which occupies primarily seasonal formations from Mexico to Argentina, but avoids the rainforests of Central and tropical South America. The phylogeographical pattern is consistent with gradual dispersal along the Central American Isthmus, followed by more rapid dispersal into and across South America after the uplift of the Isthmus of Panamá. Low sequence divergence between populations from north and south of the Amazon rainforest is consistent with mid-Pleistocene divergence, approximately 1.1 million years ago (Ma). This suggests that the Amazonian rainforests must have become fragmented or at least shrunk considerably during that period, lending support to the Pleistocene refugia theory as an important cause of distribution patterns, if not necessarily speciation, in Amazonian forest organisms. These results highlight the potential of nonforest species to contribute to an understanding of the history of the Amazonian rainforests themselves.