Tests of a sample of 206 cladograms of mammals show that morphological data seem to predict phylogenies that match the known fossil record better than molecular trees. Three metrics that assess the rank order of branching points, the stratigraphic consistency of those nodes, and the ratio of ghost range to known range show a considerable diversity of values. Some published trees show excellent matching with fossil-record data; others show almost no correspondence whatsoever. Morphological trees are nearly twice as good as molecular trees in terms of matching of the rank orders of nodes and oldest fossils, while morphological trees are 10% better than molecular in terms of stratigraphic consistency of the nodes. The ratios of ghost range to known range are lower for molecular trees. Among the molecular trees, those based on gene data are considerably better than those based on protein sequences, at least in terms of the rank order of nodes and the stratigraphic consistency of nodes. Protein trees, however, were best of all in terms of minimizing the proportion of ghost range. These findings probably indicate real phenomena, but the match of molecular trees to the expectations of stratigraphy may improve as the study of molecular phylogeny matures. (C) 1998 Academic Press.
