March 12, 2007

Not just bad linguistics, but shoddy historical genetics too?

Last week, Sally Thomason posted about the bad historical linguistics in Stephen Oppenheimer's work, at least as reported by Nicholas Wade ("Nutty journalists' (and others') language theories", 3/6/2007). At the same time, she conceded to his authority in genetics ("I have no expertise whatsoever in genetics and I therefore have no comment on Dr. Oppenheimer's proposals in this highly technical and well-developed field of inquiry."). But Razib at Gene Expression does have expertise in genetics, and his comment ("Bad historical population genetics?", 2/6/2007) was:

I've haven't read Oppenheimer's book, but I have read his The Real Eve. He's not one to be modest, and he's trying to make another splash. I've stated earlier that I thought the Etruscan studies were historical population genetics done right, and I think here Oppenheimer in particular is all about the discipline done wrong. From a Popperian perspective I suppose one could say that Oppenheimer is making bold claims which demand to be tested, but, his idea that Germanic speech predates the Anglo-Saxons, and that the Celts brought agriculture to England, rest upon revisionst and extreme minority positions within history, archaeology and linguistics. It would be one thing if the genetics was rock solid, but it isn't. The whole model seems an intellectual mess, more ego than experiment. The populations of northwestern Europe may simply be genetically too close to use uniparental phylogenies to definitively decide between historical hypotheses, other fields need to offer concurrent evidence, and that just isn't happening here.

And one of his commenters adds another complaint about allegedly "shoddy dating techniques" in another piece of work on historical population genetics:

I completely agree with you here, Oppenheimer's got a rigid view of prehistory that he's dreamed up, and he's sticking to it.

For example, his genetics papers that I'm familiar with (Austronesian) use shoddy dating techniques (one molecular clock calibration point, _rho_, at orders of magnitudes greater than the history he's trying to date) to get a huge range of age estimates (6,000 - 50,000 years ago). Despite this, they apparently still support his "revolutionary" view of Pacific prehistory which dates the spread of these people to circa 13,000 BP. This is completely in contradiction of all the linguistic and archaeological evidence.

I've ordered Oppenheimer's book, and I'll report back after I read it and go over it with some experts in relevant areas of computational biology.

Posted by Mark Liberman at March 12, 2007 08:00 AM