Abstract

Abstract The spectacular variation in species forms and richness across space and time can be explored using sophisticated and powerful tools recently developed by evolutionary modellers. In this contribution, we ask if the classic ‘Simpsonian’ view of tachytelic (fast), horotelic (standard) and bradytelic (slow) diversification rates can be distinguished with currently available tools and data. A neglected topic here is the role that the uncertainty of diversification rate estimates plays, where the lack of in‐depth uncertainty measures could hinder our ability to confidently suggest differences in speciation or extinction rates in any given comparison. We propose quantifying the relative uncertainty of diversification estimates, to better compare diversification tempo across phylogenies of different sizes and ages. We present three case studies, using the most popular models for diversification rate estimation, with or without fossils, to investigate claims of bradytely or tachytely. Using summary statistics and linear models, we ask if point estimates of diversification rates are comparable across clades. More specifically, we fit a linear model to understand which phylogenetic tree properties (including size and age) may affect the uncertainty of diversification estimates. We found the ‘Goldilocks of uncertainty’: Phylogenies that are young with insufficient tips or that are old increase the uncertainty of diversification estimates. The choice of diversification modelling approach is independent of the pattern of diversification rates decaying exponentially with clade age. In practice, we still cannot confidently compare diversification rates or their variation, due to uncertainties stemming from clade age, sample size and biased sampling. We emphasize the need for researchers to focus on estimating and presenting uncertainty in their estimates. Such uncertainty estimates are currently absent from many publications, limiting our ability to compare the tempo of diversifications across the tree of life. We conclude by proposing solutions and guidelines to encourage new studies for measure uncertainty.