In order to test the results of a previous study of didelphid marsupial phylogeny based on IRBP nuclear gene sequences (Jansa and Voss, 2000. Phylogenetic studies on didelphid marsupials I. Introduction and preliminary results from nuclear IRBP gene sequences. Journal of Mammalian Evolution 7: 43–77), we surveyed external, cranial, dental, and karyotypic characters among a more densely taxon-sampled didelphine ingroup. Separate maximum-parsimony analyses of these nonmolecular data and of a new (taxon-dense) IRBP matrix yielded superficially dissimilar strict-consensus topologies. However, no didelphine clade that was even moderately well supported by either separate analysis was contradicted by any equivalently well-supported clade in the other. Instead, all examples of taxonomic incongruence involved weak nodal support from one or both datasets. A maximum-likelihood analysis of the IRBP data produced a consensus topology that was completely congruent with, although slightly more resolved than, the maximum-parsimony consensus. A combined (simultaneous) maximum-parsimony analysis of both datasets (nonmolecular + IRBP) produced a consensus topology that closely resembled the results of analyzing IRBP separately. Most of the didelphine relationships previously reported by Jansa and Voss (op. cit.) are supported by these analytic exercises, with some notable exceptions. The taxon currently known as Marmosa canescens is conspicuously divergent from congeneric species and variously clusters with three different groups (“other Marmosa” + Micoureus, Monodelphis, or higher didelphines [= clade H of Jansa and Voss, op. cit.] ) in several parsimony-equivalent resolutions of a fourfold basal polytomy in the IRBP and combined-data consensus topologies. Even without canescens, however, the genus Marmosa is not demonstrably monophyletic. The nomenclatural consequences of these results are discussed, and a new genus is described for “Marmosa” canescens. Future analyses should test the monophyly of other speciose didelphine genera, but new sources of character data will be needed to offset the loss of resolution and decreased nodal support that are often caused by denser taxon sampling.