Effects of morphometric descriptor changes on statistical classification and morphospaces

Ten morphometric descriptors (five pairs of form and shape parameters) are used to describe the complex morphology of the first lower molar of two morphologically similar species, Microtus arvalis and M. agrestis. These descriptors are derived either from linear measurements or from outline analysis. The effects of these different descriptors on classical analysis as used in biology or palaeobiology are explored. First, the reliability of results in statistical classification is assessed. All of the descriptors discriminate well between the two species. The initial morphometric scheme (linear or outline) does not induce marked differences in statistical classification and the major discrepancies are between standardized and non-standardized versions of descriptors, and between amplitude- and coefficient-based or linear-based descriptors. Subsequently, the similarity of morphospaces based on partial least squares analysis and of intraspecific variance (estimated from the morphospace analysis) are observed. This is done within a morphospace-disparity framework and procedures used here for testing are directed at this research area. Similarities between morphospaces are relatively high. In this case, the initial morphometric scheme is a major factor inducing dissimilarity. However, the patterns of intraspecific dispersion inferred from morphospaces are roughly similar. Major differences in results correspond to the two classes of form or shape descriptors. Similarity of intraspecific variance is obtained when standardized descriptors are used (except for amplitude-based descriptors); conversely, dissimilarity is obtained when non-standardized descriptors are used. In many cases, the results of the various analyses are robust despite changes in descriptor. Moreover, the developmental pathway of vole teeth can frequently explain major dissimilarity or even similarity. (C) 2004 The Linnean Society of London.