This paper investigates the performance of new marginal likelihood estimators, specifically path sampling (PS) and stepping-stone (SS) sampling, for comparing models of demographic change and relaxed molecular clocks. The authors compare these methods to the harmonic mean estimator (HME) and a posterior simulation-based analogue of Akaike’s information criterion (AICM) through Markov chain Monte Carlo (MCMC). They find that PS and SS sampling outperform HME and AICM in terms of accuracy and reliability, particularly in complex evolutionary and population genetic models. The methods are implemented in BEAST, a powerful software package for Bayesian evolutionary analyses, and are demonstrated on synthetic data and real-world examples. The results show that PS and SS consistently yield more realistic model classifications compared to HME and AICM, making them a recommended choice for model selection in Bayesian phylogenetics.This paper investigates the performance of new marginal likelihood estimators, specifically path sampling (PS) and stepping-stone (SS) sampling, for comparing models of demographic change and relaxed molecular clocks. The authors compare these methods to the harmonic mean estimator (HME) and a posterior simulation-based analogue of Akaike’s information criterion (AICM) through Markov chain Monte Carlo (MCMC). They find that PS and SS sampling outperform HME and AICM in terms of accuracy and reliability, particularly in complex evolutionary and population genetic models. The methods are implemented in BEAST, a powerful software package for Bayesian evolutionary analyses, and are demonstrated on synthetic data and real-world examples. The results show that PS and SS consistently yield more realistic model classifications compared to HME and AICM, making them a recommended choice for model selection in Bayesian phylogenetics.