Why Not to Do Two-Species Comparative Studies: Limitations on Inferring Adaptation

Claims about adaptation are commonly made from comparative studies involving only two species (or only two populations of a single species). Our main purpose here is to alert practitioners to several logical and statistical problems associated with using two-species comparisons for studying adaptation and to outline some alternative approaches. Multispecies comparisons are one such alternative. However, data from multiple species may not be independent or identically distributed in the statistical sense, which violates assumptions of ordinary statistical methods (Harvey and Pagel 1991). We therefore also discuss one phylogenetically based statistical method that can be employed for valid hypothesis testing with comparative data, Felsenstein’s (1985) method of phylogenetically independent contrasts. We discuss briefly how such methods can be employed, even with incomplete phylogenetic information, and also how data for multiple populations within species can enhance comparative analyses. Phylogenetically based analyses come in a variety of flavors, and our penultimate section discusses some differences in perspectives regarding statistical hypothesis testing in a phylogenetic context. We conclude by pointing out that many of our criticisms of two-species comparisons apply also to comparisons aimed at discovering mechanisms underlying physiological differences between species