> I think I might just have confused the issue. Maybe what I want to
> argue for is that *-NgW- (standard: *-gW-) went
> *-´NgW- -> *-´h3w-
> *-NgW´- -> *-ngW´-
> so that it was the normal course of events, not an application of
> a classical root constraint? There might be a ton of
> counterexamples, but I'll plead a Meillet on that: the weirder
> the data, the truer it must be. We might explain 'n-infixes' by
> sticking to the II data and calling them morphological, but that
> doesn't explain how they came about.
>
A phoneme like *NgW (or is it *n,W ?) has a lot of interesting
properties. It's [+nasal, +velar, +labial] which is nice for
something one wants to use for 1sg.:
PPIE *-áNgW -> PIE *-óh3 (h3 = [-nasal, +velar, +labial])
PPIE *-aNgWí -> PIE *-mí (m = [+nasal, -velar, -labial])
(note the use of the velar -k for 1sg in other language families)
also, possibly
PPIE *NgWa -> Hittite nu
PPIE *sa -> Hittite sa
PPIE *ta -> Hittite ta
the three conjunctions, perhaps the origin of sg. mi-endings,
when placed after the verb of preceding subordinate clauses.
In a verb inflected
*stórh3-i, *str.n.gW-énti (hi-inflection)
anaptyxis would happen
*stórh3-i, *str.n&gW-énti
and since there is now actually a vowel in the root in 3pl,
that vowel gets stressed in 3sg
*str.nóh3-i, *str.n&.gW-énti
Torsten