One of the most problematic issues in vertebrate phylogenetics is the disagreement between phenotypic and molecular inferences regarding the relationships among hagfishes, lampreys, and gnathostomes. Phenotypic characters support monophyly of lampreys and gnathostomes, whereas nearly all published analyses of molecular data sets support monophyly of hagfishes and lampreys. In this study I present results of phylogenetic analyses of combined phenotypic and molecular data sets that focus on relationships among hagfishes, lampreys, and gnathostomes. Maximum parsimony analyses of 115 phenotypic characters combined with 4,638 rRNA sites and more than 10,000 amino acids each result in monophyly of lampreys and gnathostomes, demonstrating that the addition of relatively few phenotypic characters can alter phylogenetic inferences from large molecular data sets. On the other hand, Bayesian analyses of the combined data sets support monophyly of hagfish and lampreys, indicating that model-based analyses may be prone to data “swamping,” where the phylogenetic signal of the larger molecular data sets overwhelm the signal present in the much smaller phenotypic data set. Nodes that relate hagfish and lampreys were recovered at a low frequency in parametric bootstrapping analyses, indicating that the timing of diversification among hagfishes, lampreys, and gnathostomes has created a difficult phylogenetic problem for molecular data. The fact that addition of relatively few phenotypic characters can alter phylogenetic inferences of cyclostome monophyly obtained from molecular data sets, and the inability of simulated data sets to recover key nodes in the craniate phylogeny provide reasons to view the strong support for cyclostome monophyly inferred from molecular data sets with a measured degree of skepticism. J. Exp. Zool. (Mol. Dev. Evol.) 312B:749–761, 2009. © 2009 Wiley-Liss, Inc.