Homology and Comparative Methods in the Study of Avian Cooperative Breeding

In an earlier article (Edwards and Naeem 1993), we presented a comparative analysis of cooperative breeding (CB) in perching birds (order Passeriformes). Using both discrete (categorical) and continuous data sets, we illustrated a number of ways in which comparative methods could help clarify the temporal origins of CB in various clades, ecological correlates of CB, and patterns of interspecific variation in attributes such as group size. The commentary on our article by McLennan and Brooks (1993) makes three main points, each of which raises important issues about the biology and ecology of CB in birds as well as the application of comparative methods to behavioral data. The first point is that comparative analyses of behavioral traits based on trees derived from taxonomies or noncladistic methods (phenetic clustering), "neither of which produces a robust phylogenetic hypothesis of genealogy" (McLennan and Brooks 1993, p. 790), are nonetheless useful for those researchers recognizing the importance of examining evolutionary hypotheses in a phylogenetic context. This point echoes our own plea for increased interest in building robust trees for ecologically and behaviorally diverse groups of birds (see, e.g., Richman and Price 1992), particularly for cooperative breeders (see, e.g., Edwards and Wilson 1990). The analyses in our article were based on trees built with varied sources of data and methods of analysis: cladistic analysis of morphology (fairy wrens, Malurus), phenetic analysis of morphology (New World wrens, Campylorhynchus), taxonomy (e.g., Australian treecreepers, Climnacteris), phenetic analysis of molecular distance data (e.g., New World jays), and cladistic analysis of DNA sequence data (Australian babblers, Pomatostomus). We agree that phylogenies are always preferable to taxonomies in comparative analyses, and, to the extent that it is difficult o model morphological evolution phenetically (Felsenstein 1988), our analysis in these cases should be treated as tentative. But to state that "phenetic" analyses categorically do not produce "robust" phylogenetic hypothesesthereby implying that cladistic methods categorically produce better trees-is simply outdated. The philosophical underpinnings of cladistics are different from those of other methods, but this does not mean that "noncladistic" methods cannot yield truth in practice. Uncertainties in phylogenetic inference often lie more in limitations of data than in the particular method used (Miyamoto and