Rabosky 2010

Rabosky, D.L. 2010.  Extinction rates should not be estimated from molecular phylogenies.  Evolution.  Early View Date: January 2010.

Hmm weird.  If a paper is accepted in the journal Evolution, it goes through an early viewing thing where you can download the pdf but it is not actually printed in the journal yet, so it doesn't get a volume/issue number yet.

First off, I should state that I enjoyed this paper, mainly because it confirms (using fancy-math) the obvious: that you can't infer extinction rates using phylogenies without fossils.

True, the shape of a tree can be influenced by both the net rate of lineage diversification through time as well as the ratio of the extinction rate to the speciation rate; however, this does not give us license to use living species to provide information on the historical extinction rates.  I mean, if an organism goes extinct without a trace, you really can't know for certain that it existed, right?

One thing that confused me was in the abstract when he states, "molecular phylogenies contain information about the tempo and mode of species diversification through time".  What does he mean by the 'mode'?

In the article, the author shows that when rates vary across the branches of a phylogenetic tree or among clades (as they usually do), estimators that assume rate-constancy among lineages perform poorly (as they should!).  To do this, he first took a number of diversification rates from a set of clades with known extinction rates (an avian families dataset).  He then simulated clade diversity under those rates, and finally estimated the extinction rate for each set of clades.

It is clear from the author's results that among-lineage variation in diversification rates results in messed-up or directionally biased estimates of extinction rate.  He strongly urges that fossils should be used in conjunction with molecular phylogenetic studies to generate a richer perspective on the dynamics of speciation and extinction.  Agreed.

Graur & Martin 2004

Graur, D. and W. Martin.  2004.  Reading the entrails of chickens: molecular timescales of evolution and the illusion of precision.  TRENDS in Genetics.  20(2): 80-86.

The purpose of this paper was to mostly make fun of some other papers (mainly by Hedges).  While I agree that the science was bad in those papers and Graur and Martin were kind of funny in their paper, I felt it was also kind of unprofessional and I think there may have been a more diplomatic way to make it clear that the science was bad.  I don't know, maybe that's just me.  Science is supposed to be objective after all, not emotional.

Here are the main points:

-the findings summarized in a Trends in Genetics review are all based on a single calibration point and tenous methodology

-the calibration point is a supposedly well-constrained fossil divergence time between the ancestor of birds (diapsid reptiles) and mammals (synapsid reptiles) at precisely 310 ma

-the first problem is that this calibration has no errors associated with it.  The second is Graur and Martin could not find the original reference to this calibration

-the authors suggest that a solution to the single-calibration conundrum would be to use multiple primary calibrations

-the use of standard errors as error bars in highly misleading, it is more appropriate to calculate the 95% or 99% confidence intervals

-the authors state that the appearance of these outlandish reviews have resulted in hundreds of citations in which such dates were accepted as factual, but they don't actually cite any.  Was it really in the hundreds? I have my doubts.

Although this paper is mostly bashing another paper(s), it does get the point across that you have to be really careful with this methodology and not base your calibrations on someone else's calibrations.  This paper is getting a bit old now and hopefully most researchers who work with divergence dating know this already and would never take calibrations from someone else's work.  I mean, seriously.  I do appreciate that the authors sate that molecular estimates of divergence times are useful when based on solid statistical methodology and multiple fossil calibrations.  See, it's not all bad! :)