Fw: Glottochronology

----- Original Message -----

From: Juha Savolainen

To: testwham

Sent: Friday, July 04, 2003 12:56 PM

Subject: Re: Glottochronology

"Arbitrary" in the sense that if the "core list" is not objectively identifiable, the choice of the "core words" will subjective and hence the "lexical clocks" will run differently. (They will run differently because the very idea of picking "core words" presumes that words outside the "core list" will not be replaced at a constant rate, of course.)

 

Are there ways to circumvent this problem? Your comments suggest a positive answer. As I have not studied myself your references, it would be rather pointless for me to start guessing whether the in-principle objection has been met ot not. I leave this task for those who do know the relevant literature.    

 

You also seem to believe that statistical techniques could take care of the vexing problems selective forces create for time keeping. Well, while I certainly have sympathies for the use of advanced statistical techniques (preferably Bayesian) for taming tricky problems, I would like to see how they are being used before rushing to pass any judgements. Once again, I leave the field to those of us who have some first-hand knowledge of such attempts.

 

And on the margin of 3000 years: whether this is unacceptable of not depends on the nature of the problem, of course. Dating the splits in the IE language family with a margin of error of 3000 years would be too much, I think...It could be suggestive and heuristically interesting but too weak to serve as an independent dating scheme.

 

Having said that, we must keep in our minds the distinction between absolute and relative dating. A dating scheme that would get the relative order of the splits right while giving very little information about the absolute dates would be valuable, but a dating scheme that cannot reliably give even the relative dates is pretty useless.

 

Regards, Juha

testwham <test@...> wrote:

--- In cybalist@yahoogroups.com, Juha Savolainen wrote:
>
> Richard,
>
> Hmmm...I think I must disagree with you. First, my point about
the "core words" is based on my understanding (faulty as it may be)
that Swadesh did not claim that all words change at a constant rate.
Rather, if my memory serves me well here, Swadesh claimed that
the "core words" changed at a constant rate. If so, it is upon a
defender of such a view to show that such "core words" can be
unproblematically identified. If not, the whole exercise becomes
rather pointless as no non-arbitrary "lexical clocks" can be put
together.

The clock is not arbitrary; however each meaning then has to be
calibrated independently. This has been done in two big studies, of
Austronesian and Indo-European. Absolute dates were available for
the first; for Indo-European only relative replacement rates were
calculated.

> As for the "molecular clocks": there is a world of difference
between evolution driven by natural selection and evolution driven
by random accumulation of copying errors. I claim that only the
latter give us some realistic hope of constructing
meaningful "molecular clocks" and fail to understand what the
compensating "error terms" could be here. Of course, certain bases
are more prone to replacement than others, but this is not an
obstacle in itself (although it makes more difficult the
construction of molecular clocks as the rate of change must be
studied case by case). What does count is the selective neutrality
of the evolution. And unless you spell out how the "error termr" lay
rest the (chronologywise) disturbing selective forces in (a) genetic
evolution and (b) memetic evolution, I am still asking what might
play the role of "junk genes" in memetic evolution.

The problem with evolution driven by natural selction is that it may
affect the rate of change - preserve a good base sequence and
encourage change in a bad base sequence. Perhaps the 'tempo of
linguistic change' is analogous to the selective pressures. The
error terms relate to a concept of the aging of the morpheme for a
meaning. If this be valid, then if the morpheme is 'old' at the
time of the split, the time since the split is likely to be
overestimated, and if the morpheme is 'young', i.e. resistant to
replacement, the time since the split wil be underestimated. In
principle one should be able to use statistical techniques to
estimate and then allow for these effects.
> And I did hope that my Keynesian joke would give some insight on
my views: of course, if you take a very Olympian view of a river,
you will "compensate away" the turbulences. But such an "Olympian
clock" would permit margins of errors that are unacceptable for
dating purposes.

I recall a fairly wide margin (3000 years?) being quoted on the
results.

Richard.

---

Do you Yahoo!?
SBC Yahoo! DSL - Now only $29.95 per month!