Reassessing hominoid phylogeny: evaluating congruence in the morphological and temporal data

Paleobiology, Fall 2004 by Finarelli, John A, Clyde, William C

Bootstrap Analysis

Bootstrapping was also used to evaluate support for MMPC topologies (Felsenstein 1985). Two hundred characters were randomly resampled with replacement for 1000 heuristic search replicates (ten random sequence additions per replicate) and a majority-rule consensus cladogram was produced. Most of the nodes observed in the MMPC are left ambiguous in the 50% majority-rule cladogram for the bootstrap (Fig. 3A). Additionally, those nodes that were resolved have low bootstrap proportions, indicating only marginal support. It is likely that this, too, is a function of the incompleteness of certain taxa in the data matrix, which increases ambiguity under resampling techniques such as the bootstrap. We noted above that Kenyapithecus played a large role in the low Bremer support for many of the nodes in the MMPC. Therefore, a second bootstrap was performed on the data set excluding Kenyapithecus (Fig. 3B). The topology of this second bootstrap is fully consistent with that of the first, only more resolved, and the bootstrap proportions indicate stronger support for several of the internal nodes. Upon removal of Kenyapithecus, an Equatorius/Griphopithecus clade is also recovered. Removal of Kenyapithecus also serves to separate the clade of "modern" hominoids (albeit without Morotopithecus) relative to early Miocene hominoids from East Africa with strong support, as well as increasing the support for a distinct clade of great apes, including Oreopithecus.

The bootstrap topologies do differ significantly from the strict consensus cladogram in several respects. In the strict consensus of the MMPC, both the pongine clade and Ouranopithecus are hypothesized as derived relative to Dryopithecus and Oreopithecus. In both bootstraps (with and without Kenyapithecus), all European hominoids are joined in an unresolved polytomy with the hominines (Fig. 3A,B). The support for this group as a clade that is both distinct from and derived in relation to the South Asian hominoids is extremely weak (52% in both bootstraps), indicating that the relative positions of the Homininae and Ponginae with respect to late Miocene European hominoids is rather weakly supported by the morphological data, despite its apparently clear resolution in the MMPC. The polytomy in the bootstrap topologies formed by the South Asian hominoids does distinguish a Pongo/Sivapithecus clade with reasonable support. The lack of resolution of a monophyletic Ponginae here is likely also a function of incompleteness of Ankarapithecus and Lufengpithecus. These taxa are 46% and 38% coded, respectively, whereas Sivapithecus and Pongo are 72% and 100% coded, respectively. As the incompleteness of a taxon increases, it becomes more likely that resampling techniques, such as the bootstrap, will sample heavily from missing character data for that taxon. This leads to an increase in the ambiguity of its placement within and among bootstrap replicates, and serves to lower bootstrap proportions. It is likely that much of the uncertainty in the position of Lufengpithecus and Ankarapithecus is caused by the relatively poor preservation of a few key taxa.