Random Trees and the Comparative Method: A Cautionary Tale

One of the toughest problems facing comparative biology is the paucity of robust phylogenetic hypotheses for many taxonomic groups. Martins (1996) proposed a method to analyze comparative data in the absence of a known phylogeny using randomly generated trees. Before applying this method, however, researchers should be aware that (1) parameter estimates derived from this method essentially assume a star phylogeny, and thus, estimate the same evolutionary regression or correlation coefficient as traditional cross-species analyses; and (2) statistical conclusions derived from this method may be so conservative as to mask evolutionary patterns, such as Rensch's rule, and should be interpreted with caution.