I thank Nikos Kokkinos for the thoughtful and informed
engagement with the Qumran radiocarbon datings,
and his comments are most welcome. I respond below 
point by point.

> Although I have not had the time to follow closely the 
> 14C saga of the DSS, working coincidently at present 
> on 14C questions in an area 1000 years earlier, I felt I 
> should look at what Doudna said on Orion.  I know 
> Doudna's paper of 1998, and although I disagree with 
> his conclusion, I have liked his criticisms of the radiocarbon 
> method (informed it seems by Timothy Jull at Arizona).  
> In fact, I (and colleagues) have expressed similar criticisms 
> some years earlier (but also published in 1998 in Balmuth 
> & Tykot's 'Sardinian and Aegean Chronology', Oxford, 
> 29-43).  Yet, I was surprised to read a somewhat defensive 
> Orion e-mail, when the problem really lies with the 
> radiocarbon method itself.  I shall briefly explain.

The defensiveness was not over radiocarbon issues, 
but over issues of misrepresentation with some history
between Goranson and me on this point. 

> To begin with, it is futile to argue against "dismissal of outliers" 
> (Goranson) or against "data inerrancy" (Doudna).  In any 
> focussed collection of 14C determinations the first rule of 
> interpretation should be "data inerrancy" (against Doudna) 
> and yet subsequent rules may inevitably involve the "dismissal 
> of outliers" (against Goranson), depending on individual 
> interpretations.  This is the problem of 'probabilistic' statistics 
> applied to a scientific dating method, C14, which has a series 
> of inbuilt uncertainties - from the field to the lab and inter-lab 
> to the trees and stratosphere to stratigraphical interpretation.  
> The publications mentioned above give only a taste of such 
> uncertainties.


> But in connection with the DSS the example given by 
> Doudna (4Q266) is not as good as it sounds.  First, the 
> palaeographical date of this text is not as clearly cut as 
> presented ("pre-63 BC"). The script is semicursive and 
> quite careless, so a + date would be wise, most probably +. 

You are correct that the DJD palaeographic dating
by Yardeni in the Baumgarten volume is 'should be dated
to the first half or to the middle of the first century BCE', 
rather than "pre-63 BCE". 
But I referred to 'the pre-63 BCE palaeographic dating 
of 4Q266 (Da) *of Cross*" (emphasis added) in my
Orion post. Cross indeed did determine that 4Q266
preceded 63 BCE on palaeographic grounds. I cited
the reference at Doudna 1998: 460: Cross, _Ancient
Library_, 1995 ed., 96, "[4Q266] can be no later than
the first half of the first century BC...before the Roman
conquest of 63 BC...by palaeographical evidence".
(That is, Cross cited 'palaeographic evidence' as
a positive argument for pre-63 BCE dating of 4Q266,
which is meaningless unless he sees the palaeographic
dating as requiring pre-63 BCE.)

> Second, Doudna's calibration of the 14C determination 
> of the sample from this text (as with all of his calibrations) 
> is not to be accepted.  

Here Kokkinos is mistaken (though it is an honest
mistake, and fruitful to explain).

> Doudna used a version of CALIB 
> which he modified - not without reason but not an indefinite 
> one.  He attempted to account for short-lived samples 
> based on a partial study reflecting single-year calibration 
> on wood from the US Pacific Northwest, post-AD1510. 

There is some misunderstanding here. There are
two modifications at issue. CALIB is one of 3 or 4 
standard computer calibration programs in common 
use. CALIB was produced at the U. of Washington 
at that time. Kokkinos uses OxCal, produced at Oxford, 
which is another standard one. Each of these programs 
use calibration datasets and are the interface to users
being able to calculate and manipulate output
based on the calibration datasets. The calibration
datasets are the same in each of these programs,
but the output is slightly different for each of the
standard programs based on slight differences in
how the math is done. These differences are 
tiny and in almost all cases insignificant, and again,
these all derive from use of the same calibration

However the calibration dataset itself has gone through
corrections. The Zurich and Tucson datings were done
using a standard 1986 dataset; later some problems
were found with that dataset and for a few years an
interim corrected 1995 dataset was widely used by
the world's labs. In 1998 all of the labs and the
standard calibration computer programs (CALIB,
OxCal, etc.) switched to use a new 1998 dataset,
which remains the one in use now in 2002.

My calibrations were done on a CALIB program
which was programmed for the 1995 dataset, which
was reprogrammed with the 1998 dataset--that is,
the *correct* dataset, the *same* dataset which
Kokkinos's OxCal program (and all calibrations
done today) use. The reason I had to modify
a 1995-based CALIB was because the new 1998 
dataset was not out yet and my article was due to 
Flint and VanderKam. I basically cajoled Paula Reimer 
at the U. of Washington lab--where the calibration
dataset was being produced and updated--to
give me the hot, new dataset before it was released
or available to anyone. (In fact let me make this
simple: she gave me a new disc with the 1998 program,
but told me to call it a 'modification' of my old one,
because officially no one was supposed to have this
data before the release date. She was being helpful--
and it sort of had to be obscured by calling it
'modification' of my program for bureaucratic reasons.)

When the 1998 CALIB became "officially" available, 
the underlying dataset had not been changed
any from what was installed in my CALIB program
(in the era at issue with the DSS). (I checked on that
point.) Therefore the calibrations I have in Doudna 1998
are correct in terms of the most current basis used by
all labs today in 2002.

A second point is the detail of the modification for
single-year uncertainty. The option of increased
uncertainty factor is an option that all of the programs 
allow one to enter when getting calibrations. But here 
is the point, and it is important to grasp this: the *effect* 
on the calibrated datings--the output--compared to the 
calibrated datings that would have resulted if I had not 
entered that factor, was in no case greater than *2 
calendar years*, and in most cases was 
1 or 0 calendar years difference. (I ran the numbers both
ways to check on this.) Since OxCal rounds up to the 
next 10 anyway, this second factor of the increased 
uncertainty based on the single-year data I entered can 
be elimininated as having any significance at all.

> We now know that regional offsets vary considerably, and 
> we should have no idea what was the real 14C distribution 
> in Palestine of the first centuries.  Until new and regional 
> calibrations become available the best thing to do is to 
> use the standard (and latest) OxCal98/2000 programme.  

The OxCal 98/2000 program uses the same dataset as 
CALIB 1998, which is what my calibrations in Doudna 
1998 were from. The regional offset issue (discussed in
Doudna 1998) is a separate issue and unrelated to
anything to do with rejection of my calibrations using
CALIB or preference for Kokkinos's using OxCal.
(The regional offset issue is an additional uncertainty
affecting both my and Kokkinos's numbers, and all
reported radiocarbon calibrations in the world, 

I am not aware of any argument that OxCal is any 
better in principle in terms of its math/accuracy in 
running calibrations than CALIB, though having used 
both I can see why many users like OxCal for its 
user-friendliness. (Incidentally, anyone
can download OxCal [also CALIB] for free from
the Oxford lab web page; just enter 'radiocarbon
lab' in any search engine and it is easy to find.)

> This on my computer gives a date at two-sigma for 4Q266 
> of 50 BC-AD 130.  The upper end is not far from the 
> conventional palaeography at all.

I published in Doudna 1998 for this:  44 BCE-129 CE.
Again, this is from the identical dataset used by OxCal.
The only difference is OxCal is programmed to round
up (larger) to the next 10. 44 BCE gets rounded up to
50 BCE, and 129 to 130 CE. (I think OxCal's rounding 
to next 10 is sound reporting since single-year reporting 
is a misleading precision impression, though that is what 
CALIB does, as well as the original Zurich and
Tucson DSS publications. It is a minor point.)

> Doudna may say that he worked with one-sigma throughout. 

No, I agree with Kokkinos here on his emphasis that
two-sigma (the 95% range) should be the key calibrated
result of focus, and not one-sigma (the 68% range). 
For purposes of the 4Q266 example, I was thinking in
terms of two-sigma. It goes as early as 44 BCE (or 50 BCE
by OxCal's rounding up to nearest 10) compared 
to Cross's pre-63 BCE dating on palaeographic grounds.
Cross's pre-63 BCE is outside of the two-sigma (though
it is not outside of it by very much). 
> He should not have been advised to do so consistently 
> (though it is unfortunately the practice).  The method has 
> yet to reach the point of dating confidently at one-sigma in 
> all areas and times involved (despite pretensions by 
> 'know-all-statisticians-cum-archaeologists who do not 
> understand what context, association and 'terminus post 
> quem' really mean).  

Here Kokkinos is right and I agree (though
Kokkinos thinks he is arguing different from me).

> The evidence is clear even within the 
> 14C saga of the DSS - namely the internally-dated Judaean 
> documents from the era of Bar Kokhba.  For example, 
> XHev/Se 8a, the deed of sale at Kefar Barou (misspelled 
> by Doudna and others) dating to April/May AD 134 or 

(I used Milik's original mistaken reading Kefar Bebayou
because that was the convention but I noted at 
Doudna 1998: 455 n. 67 that the correct reading is 
actually "Kefar Baru".) 

> February/ March AD 135 is calibrated at one-sigma as 
> AD 230-350 - only the two-sigma age range covers it 
> (AD 130-390).  And  there is no point for Doudna to 
> plead for 'contamination' here (which may or may not 
> be true) for this can be the case with almost all of the 
> DSS samples tested.  The fact is that the two-sigma 
> age range indeed found the absolutely true date!  Thus, 
> it is also a fact that radiocarbon dating cannot at present 
> determine a more precise chronology than +100 (or 
> perhaps even +150) years for any ancient document - 
> and particularly if this comes from a mini-14C disaster 
> area on the curve.
> Goranson will say (or rather in Doudna's interpretation 
> of Goranson) that therefore "data inerrancy" wins the 
> day.  

Goranson has stated that he does not wish to be identified 
with 'data inerrancy', and therefore my guess at Goranson's 
meaning was wrong. 

> But it does not do that, as I said at the beginning.  
> Look at the single, most strange example of the lot: 
> 4Q258 (Community Rule d).  My OxCal cannot cope 
> with it even at the two-sigma age range: AD 120-320!  
> Actually not even the three-sigma (at 99.7% probability!) 
> is adequate: AD 80-330.  The easy way out will again 
> be 'modern contamination' which I would strongly pray 
> it is the case (before I begin crying at Zeitlin's grave).  
> Another way will be to begin imagining copies of original 
> DSS deposited at the time of the Second Revolt (minus 
> palaeography) - as in followers of Bar Kokhba reading 
> old sectarian documents!  But please do not raise your 
> fictional hopes - the 14C method has enough uncertainties 
> as already illustrated above.  So at last here is an "outlier"
>  for Doudna.   
> I need to go back now to the 14C results I am currently 
> checking of 1000 years earlier (there are other strange 
> battles there to be fought).  I just thought to butt in - I 
> hope I said something useful.
> (This may be posted to Orion if anyone thinks it should.)
> Nikos Kokkinos

Yes it was useful. Thank you.

Greg Doudna

For private reply, e-mail to "Greg Doudna" <[EMAIL PROTECTED]>
To unsubscribe from Orion, e-mail to [EMAIL PROTECTED] with the
message: "unsubscribe Orion." Archives are on the Orion Web
site, http://orion.mscc.huji.ac.il.

Reply via email to