Meeting Abstract
Traditionally, to infer diet from tooth morphology across Carnivora, three linear measurements on the lower carnassial are used. Unfortunately, these measurements lose much of their explanatory power when phylogeny is taken into account, which may generate problems when applied to distantly related fossils. We therefore investigate an alternative method for inferring diet using Geometric morphometrics. We examined 232 specimens from 125 extant terrestrial carnivorans and identified 4 homologous landmarks on the occlusal view of the lower carnassial. Landmarks were aligned using Generalized Procrustes Analysis and the mean shape calculated for each species. The main axes of shape variation represent two ways to achieve elongation, one found in feliforms and the other in caniforms. We estimated the multivariate phylogenetic signal within shape and used it to parameterize discriminant function analyses (DFA) to estimate how well shape predicts diet. The strength of the phylogenetic signal differs between sub-orders: feliforms exhibit a strong signal (K=0.9, p-value=0.001) while caniforms have a much weaker but still significant signal (K=0.3, p-value=0.001) compared to the null Brownian motion expectation (K=1). Using DFA to identify carnivores from omnivores we get 30-40% misclassification, but it drops significantly if the phylogeny is ignored for feliforms. We conclude that the inference of diet using landmark geometric morphometrics is influenced by phylogeny but the strength of the effect differs between the suborders: feliforms exhibit a very tight correlation between phylogeny, tooth shape and diet. We are therefore investigating whether the upper carnassial or tooth outlines will provide more reliable, phylogenetically-independent, estimates of diet across carnivorans.
