A parametric bootstrap approach to the detection of phylogenetic signals in landmark data

A phylogenetic signal is present in a morphometric data set if similarities in form reflect genealogical relationships. The degree to which such a reflection exists can be measured by comparing the topology of a morphometric-based hierarchical clustering with the topology of a cladogram that is specified a priori using other sources of data. A strong phylogenetic signal is indicated by a high degree of agreement between topologies. A lack of agreement is indicative either of data with a strong “alternative” signal (attributable to homoplasy) or of data with a lack of a signal of any kind. In considering the uncertainties inherent in morphometric data, we present a new method for detecting phylogenetic signals when form is described using landmark coordinate data. We provide a parametric bootstrapping algorithm that, while applied to landmarks, is general enough to be applied to any sort of morphometric data where a reasonable model of within-sample variation can be specified. We then demonstrate how the bootstrap data can be used to make topological comparisons between morphometric clusterings and the cladogram, using: 1) bootstrap proportions attached to cladogram nodes; 2) tree-comparison statistics; and 3) analysis of the frequencies of morphometric-based clusterings that occur when bootstrapping under the model. We then demonstrate our method by examining phylogenetic patterning in midfacial shape for ateline primates. We conclude by discussing topics where more research is needed, concentrating on efforts to partition morphometric data into homologous and homoplasious components.