Detecting evolutionarily signi?cant units above the species level using the generalised mixed Yule coalescent method

There is renewed interest in inferring evolutionary history by modelling diversi?cation rates using phylogenies. Understanding the performance of the methods used under di?erent scenarios is essential for assessing empirical results. Recently, we introduced a new approach for analysing broadscale diversity patterns, using the gener-alised mixed Yule coalescent (GMYC) method to test for the existence of evolutionarily signi?cant units above the species (higher ESUs). This approach focuses on identifying clades as well as estimating rates, and we refer to it as clade-dependent. However, the ability of the GMYC to detect the phylogenetic signature of higher ESUs has not been fully explored, nor has it been placed in the context of other, clade-independent approaches. We simulated >32 000 trees under two clade-independent models: constant-rate birth-death (CRBD) and variable-rate birth-death (VRBD), using parameter estimates from nine empirical trees and more general param-eter values. The simulated trees were used to evaluate scenarios under which GMYC might incorrectly detect the presence of higher ESUs. The GMYC null model was rejected at a high rate on CRBD-simulatedtrees. This wouldleadtospurious inference of higher ESUs. However, the support for the GMYC model was signi?cantly greater in most of the empirical clades than expected under a CRBD process. Simulations with empirically derived parameter values could therefore be used to exclude CRBD as an explanation for diversi?cation patterns. In contrast, a VRBD process could not be ruled out as an alternative explanation for the apparent signature of hESUs in the empirical clades, based on the GMYC method alone. Other metrics of tree shape, however, di?ered notably between the empirical and VRBD-simulated trees. These metrics could be used in future to distinguish clade-dependent and clade-independent models. In conclusion, detection of higher ESUs using the GMYC is robust against some clade-independent models, as long as simulations are used to evaluate these alternatives, but not against others. The di?erences between clade-dependent and clade-independent processes are biologically interesting, but most current models focus on the latter. We advocate more research into clade-dependent models for broad diversity patterns.