A CATCH-22 OR SIREN OF TITAN? ASSESSING THE IMPACT OF CHARACTER EVOLUTION MODELS ON PHYLOGENETIC AND MACROEVOLUTIONARY INFERENCES OF FOSSIL DATA

Investigating patterns of morphological character evolution is a major component of macroevolution. In phylogenetic paleobiology, morphologic characters are typically the only source of information available for reconstructing phylogenies. Thus, paleobiologists are often faced with what seems like an inescapable dilemma: given a phylogenetic hypothesis and a model of character evolution, phylogenies make predictions regarding patterns of character change. But character evolution models are used to infer phylogenetic hypotheses—a seemingly catch-22 situation. Luckily, probabilistic phylogenetic methods offer a way out “where all the different kinds of truths fit together,” i.e., simultaneous inference of topology, divergences, macroevolutionary parameters, etc. Nevertheless, even in this framework systematists must make decisions regarding how characters are coded, whether they are modeled as continuous or discrete, and how to address biological sources of rate variation, which can vary among characters, through time, among lineages, and may be influenced by changes in species ecology or developmental shifts.

To determine the impact of how different models of morphological evolution influence phylogenetic inferences and downstream comparative analyses of fossil data, we used Bayesian stepping stone sampling to compete a series of increasingly complex models of character evolution to a dataset of Cambrian-Ordovician trilobites, the Olenidae, comprising 38 species, 62 discrete characters, and 12 continuous traits. Competed models vary in complexity, ranging from simple constant rate scenarios with only discretized traits, to complex models with discrete and continuous traits, accounting for multiple sources of rate variation and ecological vs. non-ecological partitions. We also compared topological distributions across competed models by visualizing their distances in multidimensional treespace. Results indicate the addition of continuous traits dramatically improves support for complex evolutionary scenarios. Remarkably, model complexity also has a major impact on which regions of treespace are explored by MCMC, which suggests a critical need to carefully model morphologic evolution even when systematists only want to infer patterns of phylogenetic relationships alone.