Circularity and Independence in Phylogenetic Tests of Ecological Hypotheses

It has been asserted that in order to avoid circularity in phylogenetic tests of ecological hypotheses, one must exclude from the cladistic analysis any characters that might be correlated with that hypothesis. The argument assumes that selective correlation leads to lack of independence among characters and may thus bias the analysis. This argument conflates the idea of independence between the ecological hypothesis and the phylogeny with independence among characters used to construct the tree. We argue that adaptation or selection does not necessarily result in the non-independence of characters, and that characters for a cladistic analysis should be evaluated as homology statements rather than functional ones. As with any partitioning of data, character exclusion may lead to weaker phylogenetic hypotheses, and the practice of mapping characters onto a tree, rather than including them in the analysis, should be avoided. Examples from pollination biology are used to illustrate some of the theoretical and practical problems inherent in character exclusion.