Assessing calibration uncertainty in molecular dating: the assignment of fossils to alternative calibration points

Although recent methodological advances have allowed the incorporation of rate variation in molecular dating analyses, the calibration procedure, performed mainly through fossils, remains resistant to improvements. One source of uncertainty pertains to the assignment of fossils to specific nodes in a phylogeny, especially when alternative possibilities exist that can be equally justified on morphological grounds. Here we expand on a recently developed fossil cross-validation method to evaluate whether alternative nodal assignments of multiple fossils produce calibration sets that differ in their internal consistency. We use an enlarged Crypteroniaceae-centered phylogeny of Myrtales, six fossils, and 72 combinations of calibration points, termed calibration sets, to identify (i) the fossil assignments that produce the most internally consistent calibration sets and (ii) the mean ages, derived from these calibration sets, for the split of the Southeast Asian Crypteroniaceae from their West Gondwanan sister clade (node X). We found that a correlation exists between s values, devised to measure the consistency among the calibration points of a calibration set (Near and Sanderson, 2004), and nodal distances among calibration points. By ranking all sets according to the percent deviation of s from the regression line with nodal distance, we identified the sets with the highest level of corrected calibration-set consistency. These sets generated lower standard deviations associated with the ages of node X than sets characterized by lower corrected consistency. The three calibration sets with the highest corrected consistencies produced mean age estimates for node X of 79.70, 79.14, and 78.15 My. These timeframes are most compatible with the hypothesis that the Crypteroniaceae stem lineage dispersed from Africa to the Deccan plate as it drifted northward during the Late Cretaceous.