GENEALOGY-DNA-L ArchivesArchiver > GENEALOGY-DNA > 2007-04 > 1176833311
From: James Heald <>
Subject: Re: [DNA] Malaspina et al (2001)'s Cohens -- an answer ?
Date: Tue, 17 Apr 2007 19:08:31 +0100
References: <firstname.lastname@example.org> <email@example.com><firstname.lastname@example.org>
Dienekes Pontikos wrote:
> The conclusions of the earlier studies regarding the dating of the
> CMH-6 cluster which would make it compatible with "Old Testament
> priests" are no longer persuasive. There are two reasons why this is
> the case:
> 1. It is now known that the CMH-6 occurs in both J1 and J2. This means
> that the calculation of the age in Cohanim needs to be redone
> separately for J1 and J2, and indeeed at even finer phylogenetic
> 2. The mutation rate of 0.0021 used in that paper is no longer widely
> used, while a slower one proposed by Zhivotovsky is now used. (I have
> my reservations about the applicability of both to diverse
> populations, but that is not relevant here). Thus, the original
> Bronze-Age-compatible date is now suspect.
You're absolutely right (and can I say what a fantastic resource your
blog is, btw!)
However, there is some hope the two effects might cancel each other out;
and the original date might not be too far out after all.
Your point #1 already is in tune with that "founder effect" complaint on
the Wikipedia page that too many of the data were clustering too closely
around the CMH-6 for the story Thomas et al were telling. The Thomas et
al date was only saved by including samples from J2 (and Hg I and other
haplogroups) that were much too far from the CMH-6 for the story to be
Regarding #2, you're also right, sort of, but I would quibble that to
say baldly that "the mutation rate 0.0021 is incorrect" is to mislead
In fact, as an estimate of the average mutation rate of the six markers
0.0021 is not so far off, compared to say John Chandler's estimates.
Perhaps about 20% high, but not bad. (And the error bars are still
huge). What has changed is an appreciation of how different the
mutation rates are for different STR loci, which can be used to somewhat
sharpen the calculation.
What is proposed by Zhivotovsky shouldn't be called a mutation rate.
Rather, it is a correction factor for the ASD equation used in the
Thomas et al paper.
We now appreciate that saying ASD = mu t only applies if there is
perfect homogeneity in the sample -- ie if the distribution for the
number of lines surviving to 2007 is the same for distant members of the
population at a time t as it is for close relatives.
But inevitably this is not the case and there are fluctuations, so the
ASD < mu t ; so you need to put in a fudge factor alpha, so
t = alpha (ASD / mu)
where alpha > 1.
This is especially likely to be the case in this situation, where the
Ashkenazi as a group have bred much faster than the Sephardi, so the
untweaked ASD will be overweighted in the (presumably smaller)
intra-Ashkenazi variance, and underweighted in the (presumably larger)
((Which would be fine, except the Thomas et al data show the Sephardi
clustering /even more closely/ around the CMH-6. This I do not
Unfortunately also, AFAIK, nobody has a particularly convincing story
for calculating the tweak factor alpha.
Yes, you can just set alpha to 3 and hope for the best, but it's not
One other thing to flag about the Thomas et al date is the absurdly
narrow estimate of its uncertainty.
The quoted very tight error-bar would only have been correct if the 130
data-points gave 130 /independent/ samples of the variation from the
mean. But this is false, because the lineages must actually coalesce
more and more into shared common lines, going back in time. So the
number of actual effective /independent/ is (unavoidably) going to be
very much smaller; the uncertainty inevitably must remain significantly
|Re: [DNA] Malaspina et al (2001)'s Cohens -- an answer ? by James Heald <>|