Competing methods of divergence time estimation recover diverse evolutionary timescales

Abstract Molecular clocks convert evolutionary distances into geological times to infer time-scaled phylogenies. Methods differ in terms of fossil calibration strategy (node versus tip) and whether they use molecular and/or morphological data, but their relative performance remains unclear. Given their well preserved fossil record, we use Crocodylia to interrogate common approaches to timescale estimation using a combination of calibration strategies and datatypes in MrBayes. We found that calibration strategies had a much greater impact than datatype, with node-calibrated timescales producing younger ages than tip calibration. The older ages recovered by tip-calibrated approaches imply an Early Cretaceous origin (approx. 125–102 Ma) of Crocodylia, inconsistent with the biostratigraphic record, while node calibration produces more credible estimates, including a Late Cretaceous origin (approx. 95–78 Ma). Performing tip calibration with a fossilized birth–death process produced an intermediate timescale, estimating an Early–Late Cretaceous origin (approx. 118–95 Ma). These timescales support materially distinct scenarios for crocodylian evolution. We show that these disparities among calibration strategies are driven by differences in the structure of their underlying priors, with tip-calibrated timescales biased toward unrealistically ancient ages. Our results highlight the need for future analyses to explore hybrid calibration strategies that treat node and tip calibration as complementary, rather than competing, approaches.