An Improved Procedure for Testing the Effects of Key Innovations on Rate of Speciation.

The effect of key innovations (e.g., phytophagy in insects, Mitter et al. 1988; viviparity in fishes, Slowinski and Guyer 1993) on speciation rates has been a major focus of evolutionary biology in recent years (Sanderson and Donoghue 1996). Much of the apparent variability in species number might be consistent with simple stochastic models of phylogenesis, since all degrees of species diversity are equally likely under a null model of random speciation and extinction (Farris 1976; Slowinski and Guyer 1993; but see Losos and Adler 1995). Therefore, invoking a key innovation to explain the diversity of one single group possessing the innovation against a sister group not possessing it is not tenable. However, if several to many sister groups are considered, tests of positive association between the possession of the trait and species number can be designed (Slowinski and Guyer 1993). The first to follow this path were Mitter et al. (1988), who found a significant positive association between phytophagy and number of insect species. To test the association, they used a sign test. Slowinski and Guyer (1993) suggested an improved, less conservative method, based on a null model of random extinction and speciation. Under this model, the probability (P value or pi) of observing disparity as large or larger in species number between each group possessing the innovation and its sister group is first calculated. The different P values obtained are then combined using Fisher’s widely advocated pro-