CLICK TO ENLARGE
a, Single tree obtained by equal weighted maximum-parsimony analysis based on morphological data of four cranial morphological modules. Bremer support values are displayed, as well as a numeric label for each node. The Bremer support values were determined by examination of the strict consensus of trees 0.01–0.12 steps longer than the shortest tree found for the data set. b, Single tree obtained by maximum likelihood. The ingroup (Homo specimens) and outgroups are displayed in purple and black respectively.Abstract: Evolutionary novelties in the skeleton are usually expressed as changes in the timing of growth of features intrinsically integrated at different hierarchical levels of development. As a consequence, most of the shape-traits observed across species do vary quantitatively rather than qualitatively, in a multivariate space and in a modularized way. Because most phylogenetic analyses normally use discrete, hypothetically independent characters, previous attempts have disregarded the phylogenetic signals potentially enclosed in the shape of morphological structures.
When analysing low taxonomic levels, where most variation is quantitative in nature, solving basic requirements like the choice of characters and the capacity of using continuous, integrated traits is of crucial importance in recovering wider phylogenetic information. This is particularly relevant when analysing extinct lineages, where available data are limited to fossilized structures.
Here we show that when continuous, multivariant and modularized characters are treated as such, cladistic analysis successfully solves relationships among main Homo taxa. Our attempt is based on a combination of cladistics, evolutionary- development-derived selection of characters, and geometric morphometrics methods.
In contrast with previous cladistic analyses of hominid phylogeny, our method accounts for the quantitative nature of the traits, and respects their morphological integration patterns. Because complex phenotypes are observable across different taxonomic groups and are potentially informative about phylogenetic relationships, future analyses should point strongly to the incorporation of these types of trait.
Ancestral states corresponding to the first principal component of each trait, estimated as values across the principal component of each character, and visualized as deformation grids from the reference (the origin of the first principal component) towards the estimated principal component score of each node.
