To:
cybalist@egroups.com
From:
chriscrawford@...
Date sent: Fri, 07 Jul 2000 02:24:30 -0000
Send reply to:
cybalist@egroups.com
Subject: [TIED] Re: IE, AA, Nostratic and Ringo
I've seen attempts to computerise cognate-hunting, but
for various reasons the results to date have been
pathetically uninteresting. The problem is that with
dozens of phonological changes having taken place
between the ancestral language and its modern offspring
the task of reconstructing and verifying the historical
sequence of changes (usually complicated by the
operation of analogy and by interdialectal borrowing)
becomes too complex to be reduced to a mechanical
routine.
Perhaps an AI genius could design an expert system to
assist linguists in coping with this problem, but there
are further difficulties. Statistical comparison is not
enough to determine relationships: language contact
leads to word-borrowing and structural convergence,
producing mirages of distant relationship; at the time
depths in question such mirages might overshadow
genuine relationships (if any).
Some people have compared e.g. Basque vocabulary with
Aharon Dolgopolsky's Nostratic word-list (Trask) or
Austronesian with IE (Dyen) -- finding a surprisingly
large number of "matches", ranging in quality from
fairly decent to magnificent. They had no intention to
propose new superfamily schemes -- only to warn their
colleagues that pure coincidence may produce more
spurious cognates than even a specialist would expect.
Piotr
Piotr, please patronize an ignorant lurker. It seems to
me that the
word-comparison strategy could be handled with some
utility if it
weren't so arbitrary, as you point out. I would expect
that one would
define a list of the thousand most basic words in all
languages --
most
likely body parts, numerals, family relations, and
crucial
environmental components like sun, sky, earth, etc --
and then
perform
a statistical comparison across languages. That
comparison would have
to be adjusted for the specific language phoneme
peculiarities, but
my
impression is that you linguists already have a good
system for
defining phonemes: all those weird little letter
combinations you use
appear to precisely specify pronunciation. You should
be able to come
up with some sort of measure of "pronunciation
distance" (e.g.,
English
"sh" is close to English "ch", but more distant from
German "z") I
realize that all this sounds pretty far out on a limb,
but you could
establish a baseline for comparison using relationships
among
existing
languages. Anyway, all this seems rather
straightforward to me, so I
suppose that it's already been done, or perhaps already
been shown to
be balderdash.
And please forgive my layman's terminology; I still
can't figure out
what "fricative" and "glottal" mean.
><><><PIOTR><><><
[pyotr gonshorofski]
School of English
Adam Mickiewicz University
Poznan, Poland
><><><>BYE<><><><