Deep phylogenetic relationships of the largest salamander family Plethodontidae have been difficult to resolve, probably reflecting a rapid diversification early in their evolutionary history. Here, data from 50 independent nuclear markers (total 48,582 bp) are used to reconstruct the phylogeny and divergence times for plethodontid salamanders, using both concatenation and coalescence-based species tree analyses. Our results robustly resolve the position of the enigmatic eastern North American four-toed salamander (Hemidactylium) as the sister taxon of Batrachoseps + Tribe Bolitoglossini, thus settling a long-standing question. Furthermore, we statistically reject sister taxon status of Karsenia and Hydromantes, the only plethodontids to occur outside the Americas, leading us to new biogeographic hypotheses. Contrary to previous long-standing arguments that plethodontid salamanders are an old lineage originating in the Cretaceous (more than 90 Ma), our analyses lead to the hypothesis that these salamanders are much younger, arising close to the K-T boundary (~66 Ma). These time estimates are highly stable using alternative calibration schemes and dating methods. Our data simulation highlights the potential risk of making strong arguments about phylogenetic timing based on inferences from a handful of nuclear genes, a common practice. Based on the newly obtained timetree and ancestral area reconstruction results, we argue that (i) the classic "Out of Appalachia" hypothesis of plethodontid origins is problematic; (ii) the common ancestor of extant plethodontids may have originated in northwestern North America in the early Paleocene; (iii) origins of Eurasian plethodontids likely result from two separate dispersal events from western North America via Beringia in the late Eocene (~42 Ma) and the early Miocene (~23 Ma), respectively.