ABSTRACT
Evolutionary biology abounds with theories and scenarios for the origins of the major chordate clades but little attempt has been made to constrain knowledge over the dating of these evolutionary events. The fossil record of early chordates, including stem-gnathostomes and basal crown-gnathostomes, as well as the sisterclade Ambulacraria (Hemichordata plus Echinodermata), is critically re-evaluated. This is achieved through both qualitative and quantitative assessment of the fit of phylogenetic hypotheses to stratigraphic range data, and through assessment of the internal consistency of stratigraphic range data. The results suggest that the fossil record of early chordates is of variable quality; the fossil record of basal chordates appears to be a poor reflection of their evolutionary history, while the fossil record of many stem-gnathostomes, such as conodonts and heterostracans, appears to be very good, albeit poorly understood in places. Thus, palaeontological data provide little constraint on the origin of chordates, craniates, and vertebrates, other than to indicate that these clades were established by 530 Ma. The origin of total-group Gnathostomata has a well-supported fossil estimate of 495 Ma, a date which falls within the error calculations of published molecular clock estimates. The origin of crown-gnathostomes is dated at 457 Ma using the fossil record, with a confidence interval extending to 463 Ma, implying an incomplete record; this lies just outside molecular estimates (e.g. 528 Ma±56.4 myr). Finally, the fossil record suggests the divergence of actinopterygians and sarcopterygians at 425 Ma, with a very narrow confidence interval (+580 Ka) and falls within molecular estimates (450 Ma±35.5 myr). Thus, where internal assessments of palaeontological data imply a good record there is correlation with molecular clock estimates, and where these assessments suggest a poor record there is poor correlation. Where correlation occurs we may assume that our estimates are a good reflection of the true time of divergence of the various clades, and where there is conflict we must assume nothing. We note that even where corroboration between datasets occurs, error bars on divergence times remain too coarse to attempt correlation to evolutionary events in other clades, and extrinsic events in Earth history.
