Grzegorz Jagodzinski wrote:
> This material is not convincing. As Hermann pointed out in his discussion of
> the problem (1907: 47f.), the velars may have been restored analogically in
> these words.
Hardly in causatives (like *sworgH-éje/o- > dergjem), which never had a
back vowel after the root-final consonant in any of their conjugational
forms, and so no obvious source of analogical influence can be proposed.
In roots with a securely reconstructed final labiovelar we do find the
expected palatalisalion (e.g. ndez 'ignite' < *-dHogWH-éje/o-) and I'm
not aware of any velar restoration there.
Piotr