elmeras2000 wrote:
> A sequence //g^herHy-// will yield full-grade *g^her- before clusters
> and zero-grade *g^hriH- in any environment.
Perhaps, though independent examples would not be easy to find. Lat.
hordeum seems to reflect the zero grade *r., not *-iH-, and the Albanian
vocalism is of course ambiguous. On second thoughts, we should perhaps
separate the Greek word from the rest. The Latin, Germanic and Albanian
forms are regularly derivable from *g^Herzd-/*g^Hr.zd-. Greek shows *dH
rather that *d, and no reflex of *z, whose loss is strange, given the
normal development of *-zdH- to Gk. -stH-. Who knows, <kri:tHe:> could
even go back to *krih1-tah2 or the like. Because of Albanian -d-, at any
rate, I'd favour a derivation of <drithë> from something with *-zd(H)-
in it, since *-i(:)dah2 would have lost the intervocalic stop. That's
one more reason for grouping it together with the Latin and Germanic
words even at the cost of giving up the Greek connection.
Piotr