Abstract
A major concern in molecular clock dating is how to use information from the fossil record to calibrate genetic distances from DNA sequences. Here we apply three Bayesian dating methods that differ in how calibration is achieved—‘node dating’(ND) in BEAST, ‘total evidence’(TE) dating in MrBayes, and the ‘fossilised birth-death’(FBD) in FDPPDiv—to infer divergence times in the Osmundaceae or royal ferns. Osmundaceae have 13 species in four genera, two mainly in the Northern Hemisphere and two in South Africa and Australasia; they are the sister clade to the remaining leptosporangiate ferns. Their fossil record consists of at least 150 species in ∼17 genera and three extinct families. For ND, we used the five oldest fossils, while for TE and FBD dating, which do not require forcing fossils to nodes and thus can use more fossils, we included up to 36 rhizome and frond compression/impression fossils, which for TE dating were scored for 33 morphological characters. We also subsampled 10%, 25%, and 50% of the 36 fossils to assess model sensitivity. FBD-derived divergence dates were generally greater than ages inferred from ND dating; two of seven TE-derived ages agreed with FBD-obtained ages, the others were much younger or much older than ND or FBD ages. We favour the FBD-derived ages because they best match the Osmundales fossil record (including Triassic fossils not used in our study). Under the preferred model, the clade encompassing extant Osmundaceae (and many fossils) dates to the latest Palaeozoic to Early Triassic; divergences of the extant species occurred during the Neogene. Under the assumption of constant speciation and extinction rates, FBD yielded 0.0299 (0.0099–0.0549) and 0.0240 (0.0039–0.0495) for these rates, whereas neontological data yielded 0.0314 and 0.0339. However, FBD estimates of speciation and extinction are sensitive to violations in the assumption of continuous fossil sampling, therefore these estimates should be treated with caution.#
