PCA and DFA as multivariate tools to resolve ambiguous species relationships in museum collections

Abstract Taxonomic relationships in genus Berylmys (B. bowersi, B. mackenziei, B. manipulus, B. berdmorei) have traditionally been ambiguous and made difficult by their geographically disjunct and scattered populations. Several specimens held in museum collections are still of unknown or disputed species assignations. Nor have cranial characters in this genus been tested for significance in variance or for usefulness in separating species. Appropriate species identifications are important for any museum-based study of phylogeny, phylogeography, population biology, functional morphology, or conservation. This study tested diagnostic characters and species assignments on one-hundred two specimens of Berylmys housed in American Museum of Natural History (AMNH), Smithsonian Museum of Natural History (NMNH), and Field Museum of Natural History (FMNH) collections. Discriminant function (DFA), principal components (PCA), and linear regression were used to analyze sixteen morphometric measurements. DFA correctly assigned >90% of specimens to their predefined species group. Length of palate (PL) and auditory bulla (LAB), length (RL) and depth of rostrum (RD), and breadth of incisive foramina (BIF) were the best discriminators among species. Six specimens of B. bowersi were re-classified as B. mackenziei, and two B. mackenziei were re-classifed as B. bowersi. One specimen of B. bowersi, of unknown locality, is indicative of Malaysian affinity under DFA.