Dubious historical linguistics (was Re: Koryak Vowel harmony (was Re: DECAL: Examples #2: Phonotactics))

Hallo!

On Thu, 20 Jan 2005 14:11:39 +0200,
Isaac Penzev <isaacp@...> wrote:

> > Greenberg surely was a brilliant typologist, but his contributions
> > to historical linguistics are dubious.
>
> Historical linguistics is always dubious ;)
Especially when it gets to long-range comparisons, but even in
more conventional fields one sees loads of nonsense published
- and taken seriously.

What regards the long-range proposals such as Illich-Svitych's,
Dolgopolsky's and Bomhard's "Nostratic" or Greenbergs "Eurasiatic",
it is indeed likely that some of the language families included
in these macrofamilies are related to each other, but the "proofs"
published are insufficient.  Greenberg uses "multilateral comparison"
which doesn't prove anything, because it registers only similar
forms (which could be coincidence, borrowings or whatever) and not
sound correspondences.  This is at best a heuristic for finding
relationship *candidates* (which then must be examined using the
comparative method) and at most a snake oil technique that can be
used to "prove" relationship between any set of languages.  It would
not surprise me if, using Greenberg's technique, one could
"demonstrate" that Eskimo-Aleut and Na-Dené, the two American
families Greenberg excluded from "Amerind", are "Amerind" as well.

The Russian Nostraticists, as well as Bomhard and Kerns, do not
use Greenberg's technique, but what at first glance looks like
perfectly good old comparative method.  However, they often reach
down into individual languages, allow great semantic latitude
and posit vast phoneme inventories for the proto-language.

One of the few works on long-range comparison that seems to hold
water to me is Michael Fortescue's _Language Relations across
Bering Strait_, which attempts to relate Uralic, Yukaghir,
Chukotko-Kamchatkan and Eskimo-Aleut into a single stock or,
as the author cautiously calls it, "mesh".  The reasoning and
the sound correspondences they give look sane to me.

My personal opinion is that "Uralo-Siberian", as Fortescue calls
his grouping, is in turn related to Indo-European, and I am going
to try to establish Indo-Uralic sound laws, mostly based on
morphology.  I think system comparison - applying the comparative
method to entire phonological and grammatical systems - is the
best way to do long-range comparisons, where lexical cognates
are rare and hard to find.  (Fortescue, whom I have mentioned
above as a positive example of good long-range work, does this;
system comparison also has a good tradition in more "orthodox"
historical linguistics.)

> > something Samuel Johnson would have called
> > "milking the bull"
>
> Hehe. I like this phrase!
BP Jonsson used to have the full quote in his .sig for a while:
"Truth is a cow that would give them no more milk, so they are
gone to milk the bull."  Johnson apparently said that about
some kind of scholars who used what he considered bad methodology,
or something like that.  I do not know the original context,
but it describes what one sometimes reads in the field of
historical linguistics quite well.

Greetings,

Jörg.