Pleistocene glacial cycles play a major role in diversification and speciation, although the relative importance of isolation and expansion in driving diversification remains debated. We analysed mitochondrial DNA sequence data from 15 great reed warbler (Acrocephalus arundinaceus) populations distributed over the vast Eurasian breeding range of the species, and revealed unexpected postglacial expansion patterns from two glacial refugia. There were 58 different haplotypes forming two major clades, A and B. Clade A dominated in Western Europe with declining frequencies towards Eastern Europe and the Middle East, but showed a surprising increase in frequency in Western and Central Asia. Clade B dominated in the Middle East, with declining frequencies towards north in Central and Eastern Europe and was absent from Western Europe and Central Asia. A parsimonious explanation for these patterns is independent postglacial expansions from two isolated refugia, and mismatch distribution analyses confirmed this suggestion. Gene flow analyses showed that clade A colonised both Europe and Asia from a refugium in Europe, and that clade B expanded much later and colonised parts of Europe from a refugium in the Middle East. Great reed warblers in the eastern parts of the range have slightly paler plumage than western birds (sometimes treated as separate subspecies; A. a. zarudnyi and A. a. arundinaceus, respectively) and our results suggest that the plumage diversification took place during the easterly expansion of clade A. This supports the postglacial expansion hypothesis proposing that postglacial expansions drive diversification in comparatively short time periods. However, there is no indication of any (strong) reproductive isolation between clades and our data show that the refugia populations became separated during the last glaciation. This is in line with the Pleistocene speciation hypothesis invoking that much longer periods of time in isolation are needed for speciation to occur.