First Person Agents in (h)a- < *wa

[Koontz John E]

On Tue, 17 Oct 2006, Rankin, Robert L wrote:
> If Siouan is related distantly to Caddoan and/or Iroquoian, then the
> fact that those languages have a paucity of labials may be significant.
> On the other hand, I recall that in Uto-Aztecan initial */p/ > /w/ in
> some of the languages, even though in initial one would expect the
> opposite (w>p).

I probably should have added that *t and *k are similarly rare.  I was
getting at the scarcity of unaspirated stops in initial position, barring
the quite productive verb-stem-forming instrumentals.   I don't know about
the comparitive numbers of labials vs. dentals vs. velars per se, with or
without restricting matters to unaspirated forms.

Actually, I think there might be more *p than *t or *k forms, though I
could be wrong.  My impression is that it's *p > *k > *t (initially), but
that's in line with the numbers of verb stems including
instrumental-initial forms, and it might be influenced by it.

As far as *p > w, as you can see I have been wondering about that.  I
wasn't aware of the UA parallel.  We definitely have *p > w in Winnebago
and Ioway-Otoe, in initial position.

I've been wondering if some of the oddities in reconstructed *w-initial
sets might also be explained by assuming *p-initial instead, as in the
case of the first person *wa- and indefinite object *wa(a)-.  Perhaps the
first person is *pa- and only the indefinite object for is *wa(a)-.

This sort of thinking obviously ties in with Bob's work on characterizing
final, sonantized stops in Dakotan.  In effect, Bob suggests that in weak
(ening) contexts like stem final position and cluster initial position
*ptk are sonantized to *bdg, and these have the variety of familiar
reflexes in (C)CVC roots in Dakotan, e.g., Teton sab- ~ sapa, xol- ~ xota,
... (anyone remember an oral velar example? ..., blaska, gleska (but no
dl...).  I'm just wondering if prefix initial position might be another
weak position, further back in Siouan prehistory, with syncopated *pa-
Agt1 becoming *b in clusters, as in like *p-r... Agt1 of r-stem, etc., but
*ba- and eventually wa- and ha- in unsyncopated contexts.  In this context
w is definitely acting much like f in other language families, e.g., in
Germanic or Japanese.

This suggestion that *p and some initial *w (as currently reconstructed)
might be more or less related is independent of my one time postulate that
*W (pronounced "funny" w) might be the intial variant of *p and *R
("funny" r) the initial variant of *t.  I believe Bob has argued that
*W... and *R... must be *w(a)-w... and *w(a)-r...  In that case *wa-
behaves much like *pa-, I guess, and I have no good explanation for the
different behavior of *wa- Agt1 and *wa(a)- indefintie object elsewhere.

In any event *W and *R are far more obligatorily initial-only than *p and
*t are scarce in initial position, and there doesn't seem to be any
corresponding "funny" initial set for *k.

Incidentally, though I christened *W and *R and spent a lot of ink on
them, they were known to Dorsey.  Kaufman postulates them as sets (*?w and
*?r, I think), and though I think Matthews doesn't discuss them as such
anywhere they certainly seem to have had an influence on his thinking
about *w and *r and related clusters.  I may be safe in saying that Wolff
missed them, but I wouldn't want to bet on it.