Phylogenetic relationships among the salamander families have been difficult to resolve, largely because the window of time in which major lineages diverged was very short relative to the subsequently long evolutionary history of each family. We present seven new complete mitochondrial genomes representing five salamander families that have no or few mitogenome records in GenBank in order to assess the phylogenetic relationships of all salamander families from a mitogenomic perspective. Phylogenetic analyses of two data sets-one combining the entire mitogenome sequence except for the D-loop, and the other combining the deduced amino acid sequences of all 13 mitochondrial protein-coding genes-produce nearly identical well-resolved topologies. The monophyly of each family is supported, including the controversial Proteidae. The internally fertilizing salamanders are demonstrated to be a clade, concordant with recent results using nuclear genes. The internally fertilizing salamanders include two well-supported clades: one is composed of Ambystomatidae, Dicamptodontidae, and Salamandridae, the other Proteidae, Rhyacotritonidae, Amphiumidae, and Plethodontidae. In contrast to results from nuclear loci, our results support the conventional morphological hypothesis that Sirenidae is the sister-group to all other salamanders and they statistically reject the hypothesis from nuclear genes that the suborder Cryptobranchoidea (Cryptobranchidae+Hynobiidae) branched earlier than the Sirenidae. Using recently recommended fossil calibration points and a "soft bound" calibration strategy, we recalculated evolutionary timescales for tetrapods with an emphasis on living salamanders, under a Bayesian framework with and without a rate-autocorrelation assumption. Our dating results indicate: (i) the widely used rate-autocorrelation assumption in relaxed clock analyses is problematic and the accuracy of molecular dating for early lissamphibian evolution is questionable; (ii) the initial diversification of living amphibians occurred later than recent estimates would suggest, from the Late Carboniferous to the Early Permian (approximately 294 MYA); (iii) living salamanders originated during the Early Jurassic (approximately 183 MYA), and (iv) most salamander families had diverged from each other by Late Cretaceous. A likelihood-based ancestral area reconstruction analysis favors a distribution throughout Laurasia in the Early Jurassic for the common ancestor of all living salamanders.