Accepting fewer etymologies

[Sean Crist]

In my last post, I said that it is better to miss a real cognation than to
accept a false one.  I'm partly writing to explain why I say that and what
I mean by it.  I'm partly also responding the the separate discussion
about Larry Trask's set of criteria for judging a word in Basque to be
native.  I'm sure it will surprise nobody that I agree with Trask on this
matter.

Rather than restate the matter in my own words, I'd like to give a quote
from the introduction of Don Ringe's book on the relative chronology of
the sound changes in Tocharian.

------------------------------

p. xvi (emphasis, when it occurs, is Don Ringe's):

	There is more uncertainty and disagreement concerning the sound
changes in Tocharian than in almost any other area of IE linguistics.
This is due in part to the number and complexity of the changes involved,
and in even greater part to the fact that relatively few Tocharian words
have obvious IE etymologies.  However, methodological difficulties are
also involved.
	One might suppose that the uncertainties regarding sound changes
in Tocharian can be resolved by proposing many new etymologies for
Tocharian words; presumably such a procedure would greatly increase the
number of examples of individual sound changes, and so would place the
proposed changes on a firmer footing.  Apparently many scholars have drawn
just such a conclusion; at any rate, the "etymological method" of research
in Tocharian has been and continues to be extremely popular.  Yet the
etymological approach is BAD METHODOLOGY, which leads to fatal errors for
the following reasons.
	Strictly speaking, a relationship between languages can only be
PROVED by demonstrating that the similarities between them cannot
reasonably be attributed to borrowing or to chance.  For Tocharian this is
not problematic: the extraordinarily high preponderance of r-stems among
the basic kinship terms and the reappearance of word-final -r in the
mediopassive endings 3sg. -ta"r, 3pl. -nta"r are sufficient to prove that
these languages belong to the Indo-European family.  But in order to
progress beyond a mere demonstration of relationship one must employ
traditional comparative techniques based on the discovery and exploitation
of sound correspondences.  Unfortunately, unless the linguistic
relationship being investigated is very close, only a small minority of
the historically genuine correspondences will appear to be statistically
significant; the rest will be too rare.  How, then, can we distinguish
genuine correspondences and cognates from mirages?
	There is one procedure, and only one, which elevates comparative
reconstruction above the level of mere guesswork.  That procedure is the
RIGOROUS application of the comparative method, based on the recognition
of STRICT sound correspondences and ultimately on the observation that
sound changes which have been carried to completion in a linguistic
community are almost always completely regular (i.e. are "sound laws").
ALL etymologies not based on those principles are in effect
_Gleichklangsetymologien_; by themselves they have no probative value at
all, and any hypothesis which crucially depends on such etymologies will
be forever beyond proof.  (That is true even of putative cognate sets
which contribute to probabilitic proofs of relationship; in the aggregate
the similarities they exhibit are most unlikely to be the result of
chance, but individual examples can and do exhibit chance resemblances.)
But this circumstance creates a methodological paradox: we cannot propose
reliable Tocharian etymologies until we have discovered the sound laws,
yet we can only discover the Tocharian sound laws by the analysis of
reliable etymologies!
	The paradox can be resolved only by adducing support of another
kind for our etymologies, and for most the only available support is
inherent plausibility.  Tocharian words which agree closely in meaning
with solidly reconstructed PIE words, and which strongly resemble those
PIE words in shape, and which are unlikely to have undergone analogical
changes (because they are derivationally isolated), can be accepted as the
reflexes of those PIE words on the grounds on inherent plausibility.
Sound laws can be worked out on the basis of these basic etymologies;
further etymological proposals can then be tested against those sound
laws, and the reliable etymological base of Tocharian can thus be expanded
step by step.  But even though this is the only feasible course, we must
not forget that the core of "plausible" etymologies on which it is based
constitutes an inherent weakness, and that weakness must be minimized by
keeping the basic etymological core to a minimum.  As Jochem Schindler
observed to me some years ago, "It is a matter of which etymologies one
cannot do without."
	It follows that we can improve the reliability of our proposed
Tocharian sound laws not by finding more etymologies, but by accepting
FEWER.  That is exactly what I have tried to do.  The vast majority of the
Tocharian etymologies that I have accepted in this book are so plausible
that they were proposed within the first few decades of serious research
on Tocharian; few later etymologies have been admitted, and I have
proposed very few of my own.  I have also tried to apply rigorous
standards to etymologies in other IE languages and to the reconstruction
of the PIE lexicon, rejecting, for example, a great deal of the material
included in Pokorny 1959.
	It should be emphasized that this is not in any sense a
revolutionary idea or procedure.  It is the logical result of a
determination to take the regularity of sound change seriously; as such it
is implicit in and exemplified by the work of all my teachers and mentors
and of careful historical linguists generally.  It is also the most
clearly principled way to meet the challenge of Klaus Schmidt
(1980:396-7), who rightly calls for higher standards in Tocharian
etymological studies.  I am therefore resolved to ignore any criticism
which suggests that i have accepted substantially too few etymologies
(though of course individual cases are always subject to discussion).
Conversely, there is a real danger that I have accepted too many
etymologies--especially my own etymological suggestions--and the reader
should bear that in mind.

[end quote]

  \/ __ __    _\_     --Sean Crist  (kurisuto at unagi.cis.upenn.edu)
 ---  |  |    \ /     http://www.ling.upenn.edu/~kurisuto/
  _| ,| ,|   -----
  _| ,| ,|    [_]
   |  |  |    [_]