Landmarks, Localization and the Use of Morphometrics in Phylogenetic Analysis

The analysis of morphology is crucial to the study of phylogeny in many ancient and modern organismal groups. Recently a number of arguments have been made in favor of regarding certain kinds of morphometric variables as putatively homologous characters and allowing them to participate, along with other non-morphometric variables, in parsimony-based cladistic analyses. These arguments rest on the assumptions that geometric landmarks incorporate the concept of biological homology and that partial warp and principal warp morphometric variables uniquely incorporate and operationalize the concept of spatial localization, thus providing investigators with the ability to assess patterns and directions of geometrical variation in unmeasured regions of organic forms. Literature review, coupled with empirical investigations of similarities and differences among three different morphometric data analysis methods on two different datasets suggest that there is no support for these assertions. Geometric landmarks, which form the basis for all morphometric measurements and latent shape variables, have no necessary correspondence to biological homology. Partial warp variables, coordinate-point eigenshape variables, and inter-landmark distance-based singular vector shape-change variables all express localized (= regionally-weighted) geometric deformations. However, none of these deformation patterns are localized in the sense of being truly independent from globally-distributed aspects of shape change. All three of the morphometric data analysis methods investigated are capable of inferring patterns of shape change in remote regions of the form so long as a spatially adequate array of landmarks is used as the basis for the shape analysis. Based on these results it is concluded that: (1) all types of morphometric procedures are potentially useful in phylogenetic inference to the extent that they are able to summarize patterns of biotic shape change within the Euclidean space of the original measurements; (2) morphometric interpretations are strengthened by using alternative analytic approaches to probe the geometric system in a hypothetico-deductive manner and achieve a robust and consistent interpretation of underlying differences among operational taxonomic units (OTU s); and (3) while deformationbased morphometric variables can be used to analyze the distribution of organic shapes in a variety of measurement and latent variable spaces, they represent abstract investigative tools that are tied irreducibly to particular samples and data. Such variables can be used to study morphological characters, refine character and character state definitions, and aid in the assignment of character states to OTU s. However, they are not characters themselves and may not be suitable for phylogenetic analyses directly. In addition, a new morphometric data analysis technique coordinate-point eigenshape analysis is described and its use illustrated.