Re: [tied] *kW- "?"

--- In cybalist@yahoogroups.com, "tgpedersen" <tgpedersen@...> wrote:

> --- In cybalist@yahoogroups.com, "etherman23" <etherman23@...> wrote:
> > --- In cybalist@yahoogroups.com, glen gordon <glengordon01@...>
> wrote:
> > > The traditional theory is INVALID by markedness, Mate,
> > > because, simply, a logical theory must be _OPTIMALLY_
> > > EFFICIENT. A theorist must weed out any non-
> > > necessities, otherwise the theory suffers. Period.
> >
> > Actually it's quite easy to see why *k^ would be more common. If *e
> > palatalized velars and *e is the most common vowel it's perfectly
> > understandable that the velar series would violate markedness
> > constraints. But then we'd expect to see some kind of regularization,
> > which we do: the merger of unpalatalized velars and labiovelars.
>
> But if *e occurs only as an allophone in the set /e/, /o/, zero, of the
> phoneme *a, then 'palatalised velars' are palatalised only before the
> allophone /e/. In other words, the phoneme sequence *-k^a- (which is
> realised as *-ka- in proto-IE) has the allophones *-c^e-, *-ko-, *-k-
> in PIE.

With the caveat that we'd expect to see some analogical leveling.
Which perhaps explains why we have pairs like

*ger to turn *g'ers to turn
*ghrebh to grasp *g'her to grip, seize
*ghrebh to scratch *g'her to scratch

I disagree that /e/, /o/, and zero are allophones of *a, though.