TEMPORAL AND PHYLOGENETIC VARIATION IN MORPHOLOGICAL AND TAXONOMIC DIVERSIFICATION RATES OF SEED PLANTS

Since seed plants appeared in the mid-Devonian, they have come to dominate most terrestrial biomes and represent the most abundant and taxonomically rich plant clade on Earth. Their evolutionary history reflects diversification and decline of multiple subclades, most recently the Cretaceous radiation of flowering plants (angiosperms). Of ±300,000 extant seed plant species, less than 0.2% are non-flowering seed plants (i.e., ‘gymnosperms,’ currently considered paraphyletic). The dominance of angiosperms in modern ecosystems has prompted extensive speculation on intrinsic and/or extrinsic drivers for their success yet the asymmetry in extant species richness makes comparative analyses based solely on modern data highly tenuous. Here I integrate modern and fossil data for 52 monophyletic seed plant clades, including 41 extant angiosperm clades (n = 18,152 species), 5 extant ‘gymnosperm’ clades (n = 313 species), and 5 extinct seed plant clades (n = 221 species). Clade-level relationships were based on robust phylogenetic hypotheses; clades with uncertain positions or divergence ages were excluded from the analysis. In some situations (e.g., Cycadales), the effects of alternate positions were explored. In addition to species-level taxonomic diversification rates, morphological diversification rates were estimated as change in variance per million years based on species-level seed size data (i.e., dry weight in g for extant taxa, ellipsoidal volume in mm³ for extinct taxa). Results from Phylogenetic Independent Contrast (PIC) analyses indicate significant phylogenetic signal in variation of morphological rates across extant angiosperms (K = 0.35, p = 0.013) but not in extant or extinct ‘gymnosperms’. Variation in taxonomic rates shows a consistent pattern. PIC-based correlation tests (i.e., Pearson’s) indicate that neither taxonomic richness nor morphological variance are significantly correlated with clade age. Finally, branching simulations were used to test if observed variation differed significantly from expectations of a neutral model in which continuous characters may diverge indefinitely after divergence from the same values (i.e., Brownian motion) or in which characters are constrained to evolve around a given value (i.e., Ornstein-Uhlenbeck).