FOSSILS WITHOUT CHARACTERS CAN IMPROVE BAYESIAN PHYLOGENETIC ANALYSES

The fossilized birth death (FBD) model allows researchers to more completely incorporate the intricacies of fossil data into phylogenetic inference. Fossil data, in the form of morphological and temporal information, are integrated directly into the tree, leading to statistically coherent divergence times, whereby more fossil data should result in higher precision. Particularly for analyses of only fossil taxa, temporal information is usually only included for taxa in which morphological information is also available, or only for a few fossils as calibrations. However, even when morphological data are not available, temporal data can inform an FBD analysis. Here, we tested the impact of including relatively large amounts of fossil taxa with only temporal data (i.e. without morphological data), along with those for which morphological data have been collected. We tested for any impact on phylogenetic analyses of empirical data for trilobites using the FBD model. We constructed two datasets in service of this: 1) 56 taxa with 254 discrete morphological characters and temporal information in the form of stratigraphic intervals, and 2) a dataset the same as the first but with an additional 194 taxa with only temporal information, giving a total of 250 taxa. The additional temporal information in the second dataset are stratigraphic intervals taken from the Paleobiology Database for other congeneric species as those included with morphological information. We used the same analytical setup for both datasets, but for the 250-taxon dataset, we used taxonomic constraints to restrict the topological position of the taxa without morphology to their genus. We then assessed the impact on parameters inferred directly from the analyses, such as origin and divergence times, as well as metrics of stratigraphic congruence and leaf stability. We found that including more taxa, even with only temporal information, substantially reduces the uncertainty around inferred origin and divergence times. The additional data also produced trees that are significantly more stratigraphically congruent (as measured by multiple metrics) and contain fewer rogue taxa. This case study demonstrates the utility of the inclusion of more fossils, even if they lack morphology.