We supplement a previously published mitochondrial DNA data set of grass snake sequences (ND1, ND2, ND4, cyt b, in total 3,806 bp) with sequences of Corso-Sardinian and Tuscan specimens and infer their phylogeny using Bayesian, maximum likelihood and maximum parsimony methods. In addition, we estimate divergence times of grass snake clades using a relaxed molecular clock calibrated with fossil evidence, and, in a second approach, the post-Messinian reopening of the Strait of Gibraltar. Recently it was suggested that Corso-Sardinian grass snakes represent a distinct species: Natrix cetti. All tree-building methods revealed well-supported branching patterns and deep divergences among grass snakes. However, sequences of N. natrix were consistently paraphyletic with respect to Corso-Sardinian sequences. The sister group of Corso-Sardinian grass snakes is a clade embracing N. n. helvetica and N. n. lanzai. Extensive gene flow between N. n. helvetica and a more distantly related subspecies (N. n. natrix) is well known, which is why we conclude that the status of Corso-Sardinian grass snakes as subspecies of N. natrix should be reinstated. Many currently recognized grass snake subspecies conflict with mitochondrial clades, suggestive of inappropriate morphological taxon delineation and mitochondrial introgression. Divergences among grass snakes are old, and the results of the two independent dating approaches are largely congruent. Accordingly, the Alpine orogenesis seems to have caused the origin of the oldest clade, corresponding to Iberian N. n. astreptophora. The formation of Corso-Sardinian grass snakes was dated to the Early Pliocene and could result from post-Messinian flooding of the Mediterranean Basin. Another deeply divergent clade of approximately the same age, endemic in central and northern Europe, suggests the Pleistocene survival of grass snakes north of the Alps. At least one glacial refuge in which old lineages survived Pleistocene cold periods was located on each of the three major southern European peninsulas and in Anatolia. Due to pronounced sequence divergences among Italian and southern Swiss grass snakes, we hypothesize multiple refugia south of the Alps and in the Apennine Peninsula, and there is evidence for two refuges on the Balkan Peninsula.