stlatos wrote:
>>> G kná:o: 'I scrape'; MIr cnáim 'I chew/gnaw'
>> In both Greek and Celtic *n.h2 > na:.
>
> But why would only the 0-grade remain in both?
Because the present stem that survived in both branches happened to be
based on the nil-grade. Unaccented *CR.h2- and *CRah2- are
indistinguishable in Greek and Celtic (_and_ Italic), which doesn't mean
that any /CRa:/ must be a reflex of a *CRah2- root. You have to use
external cognates to make sure that it isn't the nil-grade of *CeRh2-.
In this particular case, by the way, a connection with Skt. kHánati is
possible but far from certain, based more on the similarity of the form
than a fully satisfactory correspondence.
> What do you say
> caused the optional *gYnYoxW+ in PIE?
Optional? I'd say it's the normal form of the root and doesn't have to
be explained. PIE root-structure constraints permit //CReC// as well as
//CeRC//. *//g^neh3-// isn't anomalous in any way.
Piotr