How to date a molecular phylogeny: comparison of effective priors between node calibration and fossilized birth–death

Abstract Time-calibrating a phylogenetic tree is a fundamental step in phylogenetic inference, as it allows the study of macroevolutionary processes such as lineage diversification, trait evolution and historical biogeography. To this end, the fossilized birth–death (FBD) process, a stochastic process that coherently integrates fossils into phylogenies, is increasingly used as an alternative to traditional ad hoc node-calibration densities. However, the effective prior distribution on node ages induced by the FBD has never been investigated before, hindering an informed choice between the two approaches. Here, we analyse two empirical datasets (crocodylians and fireflies) by applying several models of time-calibration, including traditional node calibrations and FBD. We show that the effective node age priors induced by the FBD process in the absence of morphological data are comparable to those induced by flat node calibrations when only the oldest fossil occurrence per calibrated clade is included in the FBD analysis. However, effective node age priors become more informative when several fossil occurrences per calibrated clade are included. Our exploration sheds light on how palaeontological information is translated to node ages by the FBD process, and suggests that node calibration approaches remain an important alternative when the fossil record of the studied group is scarce and other prior information can be used to devise informative calibration densities.