STATISTICAL CLUSTERING ANALYSIS OF ARTIODACTYL POSTCRANIAL MORPHOLOGY

Vertebrate paleontology often concentrates on craniodental material, but postcranial material can be much more common in any assemblage that underwent significant transport and often represents the only evidence of taxonomic membership. Yet standard morphological comparisons miss subtle differences in size and shape of postcranial material of functionally constrained bones. Previous work demonstrates that statistical analysis of 2-D and 3-D morphometrics can be used to differentiate taxa, but shape-differences arise from more than just species or genus level differences. In this study, we examine statistical distributions of linear measurements of postcranial structure to identify how the shapes of these structures cluster or diverge in closely related species and within sexually dimorphic species. We use measurements of the astragali, as this durable ankle bone is common enough in the Cenozoic fossil record and to be compared in detail with modern assemblages. A Principal Component Analysis of eight standardized measurements on the astragali reveals clustering of these morphological proxies at a family level, however does not allow for assignment to particular families or genera and is most impacted by size differences. We therefore applied a novel statistical approach of calculated k-nearest-neighbor to remove overall body size as a variance factor and then ran a clustering analysis. Distinct clusters emerge for all families included (bovids, cervids, antilocaprids, camelids, giraffids, and palaeomerycids), however within family analysis produce different results by different families. Antilocaprids, when corrected for size differences between fossil and extant taxa do not show any sexual dimorphism in astragali, but do differentiate species included in the study. Cervids show little species level differences, but maintain a strong signal of sexual dimorphism in shape characteristics for the astragali. Giraffid and bovid taxa included typically cluster by both size-corrected species and by sex. Camelids, when corrected for very large size differences do not show size dimorphism and very few species are distinguishable from each other. This last finding is both interesting and troubling, as the astragali of the extinct family palaeomerycidae is morphologically similar to camelids. While clusters did not emerge for palaeomerycids, the variance is most similar to the cluster spreads in strongly sexually dimporphic taxa, like cervids, and the lack of two distinct clusters could be confounded by time averaging.