"centum"/"satem" "exceptions" [was Re: Northwest IE attributes]

[Richard M. Alderson III]

On Mon, 14 Feb 2000, Stanley Friesen (sarima at friesen.net) wrote:

> At 10:11 AM 2/8/00 +0100, Miguel Carrasquer Vidal wrote:

>> Compare the number of Ce-/Co-roots (e.g. in Pokorny) with the number of
>> Ci-/Cu-roots.  I guesstimate the difference is a factor 10 or so.

> Certainly.  But in my book, even a handful of such roots is enough to
> establish i and u as PIE vowel phonemes.

Of course, there is no question of the phonemicity of *i and *u, which has
nothing to do with their occurrence in roots.

Benveniste's root theory is very specific:  *All* roots are of the form C1VC2,
where the V is the apophonic *e/o vowel, and C1 and C2 have some co-occurrence
restrictions:

1.  C1 may not be identical to C2.
2.  C1 and C2 may not both be "voiced plain" (*b *d *g *g{^w}).
3.  If either C1 or C2 is "voiceless" (*p *t *k *k{^w}), the other may not be
    "voiced aspirated" (*bh *dh *gh *g{^w}h).

Benveniste's theory treats *i and *u as conditioned variants of *y and *w, only
occurring on the surface when *e/o is not present for accentual reasons.

The real problem is that there are occurrences of *i and *u which do not ever
vary with *y and *w, so they must be phonemic, and the interchange is no longer
phonetically or phonemically automatic.

								Rich Alderson