Re: THEORY: A possible Proto-World phonology

> Date:         Thu, 29 Jun 2000 10:23:16 -0600
> From: dirk elzinga <dirk.elzinga@...>
>
> On Thu, 29 Jun 2000, Ed Heil wrote:
>
> > And in case everything wasn't confusing enough, Winifred Lehmann rides up and
> > claims "Pre-Indo-European" as a technical term for an older stage than
> > Proto-Indo-European but one which has nothing to do with Nostratic;
> > Pre-Indo-European was active, though Proto- became accusative;
> > Pre-Indo-European had no reconstructible phonemic vowels*, while Proto- had e
> > and o.
> >
> > Basically anything that is a really radical bit of reconstruction, and which
> > is not *directly* reflected in the IE languages but is theoretically elegant
> > or compelling in some way, he throws into Pre-Indo-European.
> >
> > *RE the lack of reconstructible phonemic vowels: Lehmann has been criticized
> > on this list and elsewhere for this point, which seems rather like
> > structuralism pushed too far -- it's rather like pointing out that because
> > they are in complementary distribution, english [h] and [N] could be regarded
> > as two different realizations of the same phoneme; while it's perfectly good
> > on a certain theoretical level it is intuitively abhorrent.  I'm not sure I
> > feel that strongly about the "lack of vowels reconstruction" but in any case,
> > Lehmann doesn't claim that Pre-IE was *pronounced* without distinctive vowels
> > or anything nutty like that, only that the vowels were completely conditioned
> > in quality by their context.
>
> Many accounts of North-West Caucasian (Abkhaz, Ubykh,
> Circassian, Kabardian, etc) posit a minimal vowel system (1 or 2
> or 3) which balloons into 12-16 surface vowel qualities
> depending on the environment. What Lehmann proposes is something
> similar, IIRC. The contrast isn't among vowels of different
> qualities, but rather between a Vowel and Not a Vowel. Where
> there is a Vowel, its quality is determined solely according to
> context, just as you've said. Typological reconstructionists
> have used this feature (among others) to posit a link between
> PIE and Proto-NW-Caucasian.
How does Lehmann argue that his model is better than one that says

        The language probably had 3 to 10 different vowel phonemes,
        like languages usualy do --- but the data do not allow us to
        tell them apart, so we just write **@ for all of them.

? I often get the feeling that the underlying motivation for such
theories is a distaste for admitting unknown quantities into a theory:
"if we can't tell them apart, there was only one." Occam's razor does
not apply here --- when we have no data, any theory will fit the fact,
and being the simplest is not a great virtue when experience shows
that it does not usually fit cases where data do exist.

Curiously, it seems to work the other way with consonants. "If we
can't tell why the consonants in two different words developed in
different ways, they were probably different in the protolanguage."
Well, perhaps they developed differently because the **@'s next to
them were in fact different vowels.

This makes me have great doubts about typological reconstruction. What
type the protolanguage gets can to a large extent be an expression of
the biases of the reconstructor. Without external comparison, it must
be very hard to tell in which direction a language has changed its
type. IE is large enough that internal comparison helps, but only for
changes since 'Common IE'.

AFAIR, the modern NW-Caucasian languages are all of the same 'vowel-
poor' type, but 7000 years is a long time --- for all we know the
protolanguage might have had a phonology like Hawai'ian 5-7000 years
ago, when the link to PIE would have existed. (The change would not
even have to be before the split into branches --- people don't seem
to believe in 'inherited tendencies' anymore, but we're got areal
features to replace them).

From what I've seen, syntactic type seems to be less constant than
phonological type --- and even that changes. Common German, perhaps
2500 BP, had about 4 vowels and three series of consonants in all
positions. Modern Danish has 13 vowels, and two series of consonants
that only differ when alone in initial position.

Besides, there are only so many types. All in all, the evidentiary
force of typological similarity is not huge.

Lars Mathiesen (U of Copenhagen CS Dep) <thorinn@...> (Humour NOT marked)