Yakhontov

[Larry Trask]

----------------------------Original message----------------------------
On Sun, 7 Feb 1999, Alexis Manaster-Ramer wrote:

[on the "second version" of Yakhontov's principle, the one depending on
miscellaneous resemblances]

> But where do we find a clear statement by Starostin of this
> second version?  Certainly not in the 1991 book on Altaic.

OK.  I quote from Starostin's letter, published in Mother Tongue.
Sergei states Yakhontov's principle, quite properly, in terms of
cognates, and asserts that he knows of no counterexamples among
established language families.  He then dismisses my joke Basque-English
comparisons because they apparently fail to satisfy Yakhontov's
principle.  I might query this, but I'm not interested in it here.  He
then turns to Bengtson's Basque-Caucasian comparisons, and I quote:

"Let us take Bengtson's Basque-Caucasian list again.  Here we have, out
of 19 items on the 100-word list, 13 items belonging to the 35-word list
[snip list], which gives us 37%, and leaves us with only 9% matches
within the 65-word list.  This certainly seems like a significant result
to me."

Now, recall Yakhontov's principle, which is based upon Yakhontov's
division of a modified version of the Swadesh 100-word list into a
35-word sublist and a 65-word list:

When two languages are genetically related, the proportion of cognates
in the 35-word list is always greater than the proportion of cognates in
the 65-word list."

OK?  This is part of Yakhontov's efforts to identify those vocabulary
items which are maximally resistant to replacement.  But now see what
Starostin has done.

First, Basque is not known to be related to any version of Caucasian,
and not a single cognate pair is known to exist in Basque and Caucasian.
Therefore, Yakhontov's principle, apud Yakhontov, has *nothing whatever*
to say about Basque and Caucasian.

Second, what Bengtson presents is no more than a collection of
*miscellaneous resemblances* between Basque and Caucasian.  Starostin,
by attempting to apply Yakhontov's principle to Bengtson's comparisons,
is therefore silently replacing the crucial word `cognates' with the
entirely different term `miscellaneous resemblances'.  This is clearly
illegitimate.

Third, having done this, Starostin is turning the (now altered)
Yakhontov's principle on its head.  In effect, he is arguing as follows:
if the proportion of miscellaneous resemblances in the 35-word list is
greater than the proportion of miscellaneous resemblances in the 65-word
list, that constitutes evidence that the languages are related.  But
Yakhontov himself apparently never said any such thing.  And this is why
I took Starostin to task in my response, and this is why I got confused
about Yakhontov himself was claiming.

I find it impossible to make the slightest sense out of Starostin's
remarks here without concluding that he is appealing to a principle
which is very different from what Yakhontov himself proposed.

Moreover, there is yet another flaw in Sergei's reasoning.  Whatever
Sergei is trying to claim, he is claiming it in terms of Yakhontov's
35-word list and his 65-word list.  *But Bengtson has not presented
these 100 words.*  By Starostin's count, Bengtson has presented only 13
words from the 35-word list and an unknown number of words from the
65-word list.  If we have only a modest subset of each of the two lists,
then *no version* of Yakhontov's principle can even be contemplated.
All we have to look at is Bengtson's *self-selected* data.  Therefore,
the final version of what I will now call "Starostin's principle" is
this:

If the proportion of miscellaneous resemblances in a modest subset of
the 35-word list selected by John Bengtson for his own purposes is
greater than the proportion of miscellaneous resemblances in a modest
subset of the 65-word list selected by John Bengtson for his own
purposes, then this constitutes evidence that the languages are related.

Anybody want to defend this position?  Anybody still want to maintain
that Starostin is proceeding entirely according to Yakhontov's
principle?

> Starostin's book is entirely clear. Anybody can judge by
> reading the passage which I have e-distributed.

Maybe it is, but I have never seen that book.  All I had was Sergei's
remarks in Mother Tongue -- and those remarks show clearly that Sergei
was doing something entirely different from Yakhontov.

The prosecution rests. ;-)


Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk