ONE SIZE DOES NOT FIT ALL: USING BAYESIAN FOSSIL BIRTH DEATH PROCESSES TO DETERMINE CHARACTER PARTITION FIT ACROSS MULTIPLE AMMONOID DATASETS

Both ecological and developmental theory predict that different aspects of anatomy will evolve at different rates from each other. On one hand, this enables us to test macroevolutionary hypotheses about the effects of ecology and development on rates and disparity; on the other hand, this provides an important nuisance parameter to overcome in phylogenetic analyses. Among ammonites, it is commonly suspected that adult shell characters commonly reflect adaptations to “immediate” ecological pressures more often than do juvenile and suture characters, and thus might change more frequently. Several previous ammonoid phylogenetic analyses analyzed these types of characters using parsimony; these studies suggest that suture characters appear to be the most phylogenetically informative. We reassess that here by applying Bayesian methods to three ammonoid datasets, Moyne and Neige’s Middle Jurassic 2004 dataset, Yacobucci’s Late Cretaceous 1999 dataset, and a new dataset from the Late Cretaceous Acanthoceratoidea. We partitioned the characters into every combination of suture characters, adult ornamentation characters, juvenile ornamentation characters, and shell shape characters, including a “null” model in which all classes belong to a single partition. We then allowed the rates of change for each partition to vary independently. These analyses reinforced the idea that there is not a single model that can be applied to all datasets, but also show that 2+ rate partitions generate appreciably better posterior probabilities than single partition schemes. Every partition scheme had differing levels of fit for each dataset. In this study, we were able to show the importance of model testing and recommend testing different partition schemes for each dataset rather than assuming a single “best” model for all analyses.