Bayesian inference of phylogenetic trees is not misled by correlated discrete morphological characters

Abstract Morphological characters are central to phylogenetic inference, especially for fossil taxa for which genomic data are unavailable. While Bayesian methods have gained popularity in recent years, they typically assume characters evolve independently, despite known correlations among characters. Here, we assess the impact of character correlation and evolutionary rate heterogeneity on Bayesian phylogenetic inference using extensive simulations of binary characters evolving under independent and correlated models. We find that Bayesian inference assuming character independence accurately recovers tree topologies even when characters are strongly correlated or evolve under heterogeneous rates. However, branch lengths or clock rates tend to be underestimated, particularly under extreme rate heterogeneity. These biases are partially corrected using models that integrate over character-state heterogeneity. Our results demonstrate that Bayesian methods are robust to violations of character independence in topological inference, supporting their continued use in morphological phylogenetics.