Asian box turtles (genus Cuora, family Geoemydidae) comprise a clade of 12 aquatic and semiaquatic nominate species distributed across southern China and Southeast Asia. Over the last two decades, turtles throughout Asia have been harvested at an unsustainable rate to satisfy demands for food, traditional Chinese medicine, and the pet trade. Consequently, all species of Cuora were recently placed on the IUCN Red List, nine are currently listed as critically endangered by the IUCN, and all species are listed in Appendix II of CITES. We compiled a 67-specimen mitochondrial (~1,650 base pairs (bp) from two mitochondrial genes) and a 40-specimen nuclear-plus-mitochondrial (~3,900 bp total, three nuclear introns plus an additional~860 bp mitochondrial gene fragment) DNA data set to reconstruct the phylogeny of Cuora species and to assess genetic diversity and species boundaries for several of the most problematic taxa. Our sampling included 23 C. trifasciata, 17 C. zhoui and 1-4 individuals of the remaining 10 species of Cuora. Maximum likelihood, maximum parsimony and Bayesian analyses all recovered similar, well resolved trees. Within the Cuora clade, mitochondrial and nuclear sequence data indicated that both C. zhoui and C. mccordi represent old lineages with little or no history of interspecific gene flow, suggesting that they are good genealogical species.Based on mtDNA, Cuora pani was paraphyletic and C. trifasciata was composed of two highly divergent lineages that were not each other's closest relatives; both cases of non-monophyly were due to a mtDNA sequence that was widespread and identical in C. aurocapitata, C. pani and C. trifasciata. However, when combined with nuclear DNA results, our data indicate that C. trifasciata is a single, monophyletic taxon, and that mitochondrial introgression and nuclear-mitochondrial pseudogenes have led to a complex pattern of mitochondrial gene relationships that does not reflect species-level histories. Our results imply that captive ''assurance colonies'' of both C. trifasciata and C. pani should be genotyped to identify pure, non-hybrid members of both taxa, and we recommend that introgressed and pure taxa be managed as independent entities until the full evolutionary histories of these critically endangered turtles are resolved.