Phylogenetic analyses of sequence data from the nuclear gene encoding the interphotoreceptor retinoid–binding protein (Irbp) provide compelling evidence that the recently described Amazonian marsupial genus Hyladelphys is a didelphid and suggest that it occupies an internal branch separating the traditionally recognized subfamilies Caluromyinae and Didelphinae. Although this phylogenetic position also is supported by morphological character data, analyses of sequence data from the dentin matrix protein 1 gene (Dmp1) place Hyladelphys within the didelphine radiation. A parsimony analysis of the combined morphological and molecular data supports the Irbp-only results, but a partitioned Bayesian analysis of the combined gene data does not provide strong support for either placement. The implications of these results are discussed in terms of long-branch attraction, base-compositional bias, and other possibly confounding factors. Whether Hyladelphys is the sister group of Didelphinae or an independent lineage at some basal level within that subfamily, the absence of any close relationship to other Recent taxa is clearly indicated. We redescribe the genus based on new morphological character data and comment on the probable existence of undescribed taxa in addition to H. kalinowskii, the type species.