2015 GSA Annual Meeting in Baltimore, Maryland, USA (1-4 November 2015)

Paper No. 176-12
Presentation Time: 10:45 AM


HEINS, Liam E., Department of Ecology and Evolutionary Biology, University of Kansas, 1345 Jayhawk Blvd, Lawrence, KS 66045, HOLDER, Mark T., Department of Ecology and Evolutionary Biology, University of Kansas, 1200 Sunnyside Ave, Lawrence, KS 66045 and LIEBERMAN, Bruce S., Department of Ecology & Evolutionary Biology, University of Kansas, 1345 Jayhawk Blvd, Dyche Hall, Lawrence, KS 66045, lheins@ku.edu

Accurately resolving phylogenetic relationships among fossil species is essential for understanding macroevolutionary patterns and processes. The overwhelming majority of phylogenetic analyses of fossil data have employed parsimony. In recent years attempts to adapt maximum likelihood (ML) and Bayesian methods to fossil data have yielded promising results towards overcoming some of the limitations of parsimony analysis such as long-branch attraction and problematic handling of missing data. However, application of existing ML approaches to fossil data requires adjustments based on the manner in which morphological data is collected. For instance, when adapting these parametric approaches to fossil data it is necessary to take into account various ascertainment biases related to the way that only parsimony-informative data, which excludes invariable and autapomorphic characters, is typically coded. Here we present results from reimplementing an algorithm developed in collaboration with Jordan Koch for calculating likelihoods that are conditional on parsimony-informative data. We have integrated of this algorithm into a ML framework that properly accounts for ascertainment bias and missing data. We use character datasets from two distinct and diverse groups of trilobites, the Cheiruridae and the Olenellina, to illustrate the method. Further, we show that these approaches provide a more computationally efficient way of accounting for ascertainment bias with regards to parsimony-informative morphological data, and allow for statistically rigorous testing of models of character evolution and phylogenetic reconstruction in the context of fossil data, with the potential to provide more realistic phylogenetic results than previously available ML methods.