LINGUIST List 5.1483

Mon 19 Dec 1994

Disc: Comparative Method

Editor for this issue: <>

Directory

1. "Larry Trask", Comparative method
2. Ecological Linguistics,Anderson,PRT, Compar.Ling. item Sci.Am.Jan95

I've been following with great interest the long discussion of the
comparative method, and more particularly of the absence of the
comparative method in a number of recent attempts at establishing
remote genetic links. Quite a few people have drawn attention to the
very large number of errors which specialists have found in
Greenberg's data, and asked whether such errors are, all by
themselves, enough to destroy Greenberg's conclusions.

But I've been struck by the fact that no one has asked what I would
suggest is a more fundamental question, and so I've decided to ask it
myself. It's this:

Is it possible to do ANY useful comparative work on languages you know
nothing about?

In other words, can you establish anything about the genetic
affiliations of a language merely by extracting data from secondary
sources, without yourself having any kind of specialist knowledge of
that language?

My reason for asking is my exasperation with the numerous recent
attempts at locating some relatives for the genetic isolate Basque,
attempts which in every case have been accompanied by claims of
success: Basque, it has been asserted, has been shown to be related to
North Caucasian, to Burushaski, to Yeniseian, and to ____ (you may
fill in the blank at will without falsifying my statement).

Now, outright errors in the citation of Basque data are not exactly
lacking in this work, but, more importantly, all of it, without
exception, has been carried out by scholars who clearly know nothing
at all about Basque, and who have proceeded merely by extracting
Basque material incomprehendingly from bilingual dictionaries and
other secondary sources. As a result, a huge proportion of the
"evidence" adduced on the Basque side consists of the following:

 --Obvious loan words;
 --Obvious neologisms;
 --Dialect variants of words whose more widespread, and more conservative,
 forms would destroy the proposed matchups;
 --Words whose phonological forms show that they could not possibly have
 been in the language even 2000 years ago;
 --Words which are known for certain to have had, less than 2000 years
 ago, phonological forms which were so different as to destroy the
 proposed matchups;
 --Words which have been arbitrarily, and quite wrongly, segmented
 in order to extract non-existent "roots" for purposes of finding
 matches.

And this is only on the Basque side. I have no reason to suppose that
the instigators of such attempts have, in general, a comparatively
magisterial command of the other languages involved.

The shortcomings deriving from the investigators' complete ignorance
of Basque are so severe as to render all this work meaningless. But
is the work on Basque a special case? Have other people been more
successful while using the same approach? Has anybody ever certainly
identified a relative for Tzotzil merely by leafing through a
Tzotzil-English dictionary and picking out the bits he liked?

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
England

larrytcogs.susx.ac.uk

Two notes from Scientific American, January, 1995, which are relevant to our
concerns in deep language connections.

One is the review pp.102-103 of

The History and Geography fo Human Genes
by L. Luc Cavalli-Sforze, Paolo Menozzi and Alberto Piazza.
Princeton University Press, 1994, $250.

The phrasing of one sentence in the review provokes a question. The sentence
is:

"One reassuring result is the remarkably smooth relation obeyed by the "genetic
distance" between pairs of groups and the simple geographical distance between
the group locations."

My question is whether there is an analog in the field of genetics to the
distinction linguists make between inherited cognate vocabulary vs. (at least
relatively) later loanwords. The only one I can think of right off is that if
there were a family tree of the mutations of mitochondrial DNA, it should be
possible to detect a late reintroduction of descendents of one branch of the
tree into a human population which did not generally share all the other forms
from that branch of the tree because they represented what was basically a
different branch of the tree at older time periods. Have such phenomena been
factored out? Is it possible to do so? I have not followed the details of
this field, but could others reading the messages on Comparative Linguistics
who have followed it please enlighten me or us?

Of course genetics and language family groupings need not correspond neatly,
since people can adopt the language of others. But they will more often tend
to, we expect.

The other item in this month's issue on pp.17-20 contains yet further examples
of inherited DNA keeping its same function more or less across gigantic
evolutionary divides, so that look-alikes represent not merely convergence
dictated by function but rather true inherited cognates. This concerns
proteins which guide the developing neurons to grow towards them.

Netrin-1 and Netrin-2 in vertebrates (chicks, rodents)
Unc-6 in a Nematode (the netrins "resemble" unc-6)
a netrin gene in the fruit fly Drosophila

Lloyd Anderson