Testing species boundaries in biodiversity studies

A paper recently published by Phillips et al. (1996) reported a molecular genetic study of all recognized subspecies of the common snapping turtle (Chelydra serpentina) and concluded from patterns of geographic variation in isozyme and mitochondrial DNA restriction fragment pattern data that three evolutionary species should be recognized in this group. We suggest that this paper is fundamentally flawed because it fails to present any species concept as a testable hypothesis. Data are collected in the absence of any conceptual framework for diagnosing species boundaries, so no criteria for acceptance or rejection of a preferred hypothesis are formulated, and species boundaries are determined in a nonrigorous, Post hoc manner. Further, the absence of specific criteria for species diagnosis in this case leads to flaws in sampling design, data collection, and data analysis. Because these design flaws are typical of many studies, ute briefly outline three different lineage-based operational species concepts (phylogenetic, concordance, and cohesion) and present an alternative interpretation of the Chelydra data, insofar as this is possible given the design limitations We conclude the following: (1) species status may not be warranted for the Central and South American taxa, (2) more-detailed analysis is warranted in the US population because distinct lineage may be obscured by poor lab technique or introgression of mtDNA, and (3) the Ecuadorean population may deserve species status based on fixed nuclear isozyme loci. We recommend the following procedures for implementation of lineage-based species concepts within a rigorous hypothesis testing framework. First, if an animal is to be sacrificed, proper care should be taken to utilize different tissues for multiple pass electrophoresis. This will unmask hidden heterogeneity and maximize the number of resolved loci. Second, when phylogenetic relationships are reconstructed with restriction fragment length polymorphism data, fragment data should be converted to site data to avoid uncertainties in homology. Finally, a proper sampling scheme should be designed to address the question of species status in terms of numbers of individuals, genetic loci, populations, and geographic regions.