Seems as if time-calibration must be working to some degree

This week's journal club discussion covered McIntyre et al. 2017, Global biogeography since Pangaea, Proc. R. Soc. B 284: 20170716.

The authors set out to compare estimated continental break-up (and, to a lesser degree, collision) times as estimated from palaeomagnetic data with species divergence times as estimated from phylogenetic analyses using molecular clocks. They selected 42 vertebrate sister taxa for their presumed lack of dispersibility to exclude groups whose distribution may have been influenced by long-distance dispersal. Even among the selected taxa they tried to account for dispersibility by, if I understand correctly, extending the error bars around the divergence times for lineages that seemed more dispersible.

In the end they arrived at a very nice correlation between continental break-up times and divergence times. What does this tell us?

There were some concerns in our group about the argumentation being somewhat circular. I do not actually see that myself; one dataset was palaeomagnetic, and times in the other would presumably have been based on fossils and nucleotide substitution rates, so really two independent data sources would have been compared. (The time-calibrated phylogenies were sourced from the timetree.org database, which I have not yet used myself.)

To the degree that I found the methodology odd it is because of the decision to extend error bars when dispersal was considered somewhat probable. Yes, admittedly the immediately obvious way of identifying confounding dispersal - comparing divergence times against continental break-up times - would be circular in a study explicitly setting out to compare those two; using that approach would have amounted to massaging the data. But I would still find it more logical to have some way of categorically identifying suspected cases of dispersal and kicking them out of the dataset instead of leaving them in but making the relevant data points fuzzy.

What I found most puzzling, however, is that the paper is not actually very clear on what the research question was. It is thus somewhat up to the reader to draw a conclusion. If you already trust time-calibrated phylogenies, you could take the study to confirm the reliability of palaeomagnetic data. If you already believe that palaeomagnetics works but are somewhat skeptical about time-calibrated phylogenies, this study should at least show that molecular clocks can't be that bad after all, otherwise they wouldn't have got such a neat calibration out at the end.

And this is also what I take away from our reading, especially in the light of the criticism of molecular clocks that is still regularly advanced by vicariance biogeographers and panbiogeographers. Yes, this study did show that the fit is pretty good except where there is reason to suspect dispersal. And that brings us to the last point:

The present paper carefully excluded cases of suspected dispersal to examine only cases of vicariance, so the authors must be biogeographers (and geologists) who accept the existence of both long-distance dispersal and vicariance. And the same was true of our journal club. Nobody I know has any problem whatsoever reading a paper that concludes "this pattern is best explained by vicariance" if that is indeed what the data say.

But let's be clear here, it does not work the other way. Just read the papers I discussed a few weeks ago; pan- and vicariance biogeographers generally do have a problem reading a paper that concludes "this pattern is best explained by long-distance dispersal" and will instinctively start questioning the methodology.

The situation is just not symmetrical. The "dispersalist" who tries to explain every pattern with dispersal, no matter what the data say, is a non-existent straw-man. I have never met or read such as colleague. The panbiogeographer who tries to explain every pattern with vicariance, no matter what the data say, does, however, seem to be alive and kicking.