Are 'Comparative Methods' Always Necessary?

Comparative analyses of traits across taxa taking explicit notice of phylogenetic relationships among the taxa involved, are today commonplace in modern evolutionary ecology. This approach grew from an increasing awareness among ecologists that evolution has a clear historical component such that closely related taxa are not solely a product of their current environment but have also inherited a certain proportion of their phenotypes from a common ancestor (see reviews by Brooks and McLennan 1991, Harvey and Pagel 1991, Maddison and Maddison 1992). Since evolution modifies what already exists this is to be expected. This fact posed a problem, for example, for ecologists who are interested in convergent evolution as a response to common selection pressures. If two species share the same adaptation, shared selection pressures need not be the cause of the similarity, but the fact that their common ancestor had already evolved this particular adaptation. Thus, from an evolutionary point of view, the two species do not represent two independent evolutionary events. This problem is opposite to the one faced in phylogenetic systematics, where the null hypothesis is that similarities are homologous rather than being a result of convergence. Characters that vary widely are generally avoided because of the risk of finding similarities that in fact are an effect of convergence, which makes the character potentially misleading in a phylogenetic analysis. For ecologists, however, convergence due to ecological reasons is often what is of interest, and thus similarity due to shared ancestry is the confounding factor in comparative studies. Consequently, there has been a considerable interest among ecologists and evolutionary biologists to try to find ways of sorting out to what extent the distribution of traits among taxa is due to convergence and what is due to shared ancestry (see reviews by Brooks and McLennan 1991, Harvey and Pagel 1991, Maddison and Maddison 1992). However, while it is increasingly appreciated that many character distributions may have a component of shared ancestry, this must not necessarily be the case. As is evident from systematic studies some characters may vary so much that they are of no use in systematics, and this conversely must mean that the very same characters can be used in comparative ecological studies without any precautions with regard to phylogeny. As no comparative methods are free from assumptions that may limit their usefulness or may have other properties that may render them less suitable (e.g. Bjdrklund 1994, Ricklefs and Starck 1996), use of raw data seems desirable as much as possible. I will argue here that the approaches commonly used in systematics can be of great value in analysing the phylogenetic component of character state distribution over a set of taxa, and can help to decide whether the data can be used as they stand, or whether we will have to use comparative methods. This will perhaps also help to settle the conflict among the widely different schools of thought with regard to comparative methods and their usefulness (see review by Ricklefs and Starck 1996). Several methods explicitly try to estimate the phylogenetic component of trait distributions by means of elaborate statistical procedures. In particular, these methods aim to either estimate the relative importance of shared ancestry vs convergence (e.g. Cheverud et al. 1985, Lynch 1991), or to find the taxonomic level of analysis where the influence of shared ancestry is minimal (e.g. Steams 1983, Derrickson and Ricklefs 1988, Miles and Dunham 1992). The method presented here is a means of testing if there is any influence of shared ancestry in the trait distribution that requires special concern in further analysis. The method has also the virtue of being very simple and relies on well-established methods used in phylogenetic systematics.