Aim: This study aims to reconstruct the evolutionary history of the Crocidura poensis species complex and to identify factors driving diversification within it. We tested whether: (a) there is a pattern of allopatric differentiation coincident with the location of hypothesized Pleistocene forest refugia, (b) sister taxa are separated by broad rivers, (c) sister taxa occupy adjacent but distinct habitat.Location: Sub‐Saharan African forests and adjacent savanna.Taxon: Shrews.Methods: Analyses were based on 247 specimens collected from across the distribution of the species complex. We reconstructed the phylogeny (Bayesian and maximum likelihood methods) and assessed historical biogeography of this taxonomic group using a combination of mitochondrial and nuclear markers. We mapped the genetic diversity and estimated the divergence times by a relaxed clock model. Informed by multilocus species delimitation methods, we discussed possible taxonomic implications.Results: This complex is composed of nine major genetic lineages (proposed species). The earliest split within this complex occurred after 2.0–2.4 Ma, which corresponds to a period of increased aridity and/or extreme environmental variability. Most other divergence events occurred after the Early‐Middle Pleistocene Transition (1.2–0.8 Ma). Divergent selection across ecological gradients could explain diversification within the West African lineage. In Central Africa, the observed phylogeographic pattern fits the Pleistocene refuge hypothesis and supports the existence of multiple small rather than a few large forest refugia during glacial maxima. Large rivers, like the Congo and Sanaga Rivers, are important barriers to gene flow for several lineages but probably were not the primary cause of differentiation.Main conclusions: Both geographic isolation in distinct forest refugia and divergent selection along ecological gradients could explain Pleistocene diversification within this complex.