--- In
cybalist@yahoogroups.com, "Anders R. Joergensen"
<ollga_loudec@...> wrote:
>
> There is good evidence against a development *e > o by labiovelar
> rounding:
>
> OIr. geilt 'wild' < *gWelti- (MW gwyllt)
> OIr. crenaid 'buys' < *kWrenath(i) (lowering) < *kWrinati (MW pryn-
)
> OIr. cethair 'four' < *kWetwores (MW pedwar)
> OIr. ceirt 'bush' < *kWert- (MW perth)
> etc.
>
> [Material lifted from Schrijver, "Vowel Rounding by Primitive
Irish
> Labiovelars", Ériu L, 133-137]
I may have been a bit hasty: In Schrijver's formulation, *kWe gives
*ku when followed by a consonant palatalized in the first round of
palatalization (or followed by -Cu-). However, since -i- and -u-
provokes raising of a preceding -e- > -i-, this (*CWeC'-, CWeCu-) is
quite rare.
So we might expect a 3sg. *gWenet(i) > *gWen'e- > goinid, but a 3pl.
*gWenont(i) > *genait. However, in *gWel-e- 'graze', the unrounded
allomorph is generalized, OIr. geilid, gelait.
And in view of Brit. Celt. *gwan-, a reconstruction *gWen- does not
seem possible.
Rounding of *a > o following *kW, *gW is unconditioned (OIr.
coire 'cauldron' < *kWario-, MW peir, Romance *pario-, OIr.
Cothriche 'St. Patrick' < *kWatrixio- < Lat. Patricius).
Anders