New approaches to late Miocene hominoid systematics: Ranking morphological characters by phylogenetic signal

Abstract

The phylogenetic relationships of Eurasian fossil hominoids from the later Miocene (∼13–5 Ma) have been in a state of flux for over 150 years. Disagreement focuses on how fossil genera are related to each other and to extant clades, and which characters are useful for resolving these evolutionary relationships. The incongruities among opposing phylogenetic hypotheses reflect, in part, different methodological choices made by researchers at each stage of analysis, in particular, how characters are selected, described, and coded. This project is both an investigation into the evolutionary relationships of Miocene hominoids from Eurasia and an attempt to implement new methods for inferring these relationships. The aims of the study are to: (1) clarify the relationships of later Miocene hominoids; and, (2) establish new protocols for the selection of morphological characters for phylogenetic inference.

Quantitative (n=115) and qualitative (n=20) morphological character data were collected from the cranium (n=457) and postcranium (n=264) for ten extant anthropoid taxa and seven Eurasian fossil apes from the time period ∼13–6 Ma. I presented a new method for selecting characters for analysis involving fossil taxa, based on the relative phylogenetic signal of each trait for a closely related group of extant taxa. Individual characters were ranked on their ability to recover phylogenetic information (topology and branch lengths) independently inferred from molecular data. Ranks were derived from maximum likelihood and parsimony-based null models representing hypotheses of no phylogenetic signal in each character for the sampled taxa. Characters deemed to exhibit high-signal levels for extant species were then used preferentially in phylogenetic analyses including fossil taxa.

Methodological choices relating to character selection and coding were demonstrated to have a marked effect on both hypotheses of primary homology and resulting phylogenetic trees. For extant taxa, trees inferred from character matrices composed of only highly-ranked traits performed significantly better (shorter tree lengths and greater branch support) than those composed of all traits. Preferential selection of characters that were conserved over phylogeny proved an effective means of improving phylogenetic accuracy for the extant taxa sampled in this study. Characters that ranked highly for living species also produced well-supported phylogenetic hypotheses for later Miocene hominoids.