EVALUATING SHIFTS IN ANGIOSPERM DIVERSIFICATION RATES

Origination and extinction rates of taxa vary non-randomly over geologic time and among clades. Exploration of the processes underlying this variation and the identification of key biological, ecological, and physical factors that affect evolutionary rates is a fundamental component of macroevolutionary research. Given a well-supported phylogenetic node and reliable estimates of species richness in sister taxa at a subsequent point in time, one may compare relative net diversification without data on exact times of origin. By definition, monophyletic sister taxa share a common time of origin; greater species richness in one taxon relative to its sister may be interpreted as the result of higher origination rates or lower extinction rates (or both). However, given the range of factors proposed as potentially affecting evolutionary rates, the validity of comparing sister pairs to a null distribution derived from a constant rates birth-death model is questionable. To what degree is a trait’s effect on rates overprinted by effects of other traits or environmental changes over geologic time? Here we use species richness estimates of extant angiosperm genera and families and sister taxon relationships from published molecular phylogenies to test (1) how sister pair tests are affected by inaccuracies in phylogenetic relationships and composition and (2) how well distributions of species among higher taxa match model predictions. Simulations indicate that a constant rates model accurately predicts species distributions among angiosperm genera but not among families, paralleling work on some animal clades (e.g., rodents). The distribution of pair asymmetries (i.e., ratio of the smaller taxon to its larger sister taxon) is a hollow curve, skewed toward 1.0. Pair asymmetry is significantly correlated with the total richness of the pair at the genus level but not at the family level. Results suggest that some uncertainty in phylogenetic relationships and composition should not bias estimates of relative diversification from sister pairs. The utility of a constant rates model as the null is questionable for higher taxa, though reasonable for genera, potentially indicating that lineages descending from deeper nodes generally associated with higher taxonomic rank are more likely to have undergone rate changes.