When can decreasing diversification rates be detected with molecular phylogenies and the fossil record?

Traditionally, patterns and processes of diversification could only be inferred from the fossil record. However, there are an increasing number of tools that enable diversification dynamics to be inferred from molecular phylogenies. The application of these tools to new data sets has renewed interest in the question of the prevalence of diversity-dependent diversification. However, there is growing recognition that the absence of extinct species in molecular phylogenies may prevent accurate inferences about the underlying diversification dynamics. On the other hand, even though the fossil record provides direct data on extinct species, its incompleteness can also mask true diversification processes. Here, using computer-generated diversity-dependent phylogenies, we mimicked molecular phylogenies by eliminating extinct lineages. We also simulated the fossil record by converting the temporal axis into discrete intervals and imposing a variety of preservation processes on the lineages. Given the lack of reliable phylogenies for many fossil marine taxa, we also stripped away phylogenetic information from the computer-generated phylogenies. For the simulated molecular phylogenies, we examined the efficacy of the standard metric (the γ statistic) for identifying decreasing rates of diversification. We find that the underlying decreasing rate of diversification is detected only when the rate of change in the diversification rate is high, and if the molecular phylogeny happens to capture the diversification process as the equilibrium diversity is first reached or shortly thereafter. In contrast, estimating rates of diversification from the simulated fossil record captures the expected zero rate of diversification after equilibrium is reached under a wide range of preservation scenarios. The ability to detect the initial decreasing rate of diversification is lost as the temporal resolution of the fossil record drops and with a decreased quality of preservation. When the rate of change of the diversification rate is low, the γ statistic will typically fail to detect the decreasing rate of diversification, as will the fossil record, although the fossil record still retains the signature of the diversity dependence in yielding approximately zero diversification rates. Thus, although a significantly negative γ value for a molecular phylogeny indicates a decreasing rate of diversification, a nonsignificantly negative or positive γ value might mean exponential diversification, or a slowly decreasing rate of diversification, or simply species turnover at a constant diversity. The fossil record can be of assistance in helping choose among these possibilities.