Using the Volumetric Indices of Telencephalic Structures to Distinguish Salamandridae andPlethodontidae : Comparison of Three Statistical Methods

The aim of this study was to establish whether appropriate statistical analysis of 16 volumetric indices corresponding to 16 structures making up the entire telencephalon of Urodela could distinguish between two families,Salamandridae and Plethodontidae. We compared the efficiency of three statistical methods (stepwise discriminant analysis, artificial neural networks, correspondence factor analysis) and the information they provide. All three methods found the same species difficult to classify. However, only correspondence factor analysis could suggest explanations for “misclassifications” as it superimposes the two sets of variables, (sub)species and anatomical variables, thus revealing the correlations between them. The bulbus olfactorius accessorius and the caudal mitral cell layer of thebulbus olfactorius principalis were the most discriminatory structures in separating Salamandridae andPlethodontidae . The correspondence factor analysis mapped species very much in line with accepted taxonomy and highlighted several current controversies [e.g. positioning of certain newts (T. marmoratus, T. vulgaris, T. alpestris), ofSalamandrina terdigitata, and of members of the genus Euproctus]. Mapping of Plethodontidae was less clear-cut than that of Salamandridae with more overlap among genera but was quite consistent with knowledge of brain structure complexification. We conclude that relationships derived from analyses of telencephalic structures provide valuable information that might help resolve ambiguities; we have coined the term “neurotaxonomy” for this approach.