On Sat, Feb 24, 2001 at 11:29:32PM +0100, Jörg Rhiemeier wrote:
> But as we are at it, I am sceptical of the palatal series in PIE as
> well.  I'd rather say that they were velars, which somehow were missed
> by the satemization.  (I am not alone here;

You're definitely not alone. According to my PIE grammar of choice (_New
Comparative Grammar of Greek and Latin_ by Andrew Sihler), they were most
likely velars; but the term "palatals" for some reason is still used by some
IEists, as a vestige of past reconstructions I guess. I don't remember all
of Sihler's reasoning, but I remember specifically that he says shifts from
palatal stop to velar stop are completely unheard of in any other language
(but probably not impossible even if very rare, IMHO), and furthermore that
the "satem" language groups would have to have gone through separate changes
of palatal->velar, rather than all going through the change once (because of
their geographic locations), meaning that given the extreme unlikeliness of
a palatal->velar shift, it would be astronomically unlikely for that to
happen several times in PIE.

I find it surprising though that a shift of palatal stops to velars hasn't
ever been conclusively demonstrated. It doesn't seem any less reasonable
than the very common shift in the opposite direction. Besides that, there
have been shifts of palatal (or nearly palatal, like post-alveolar)
*fricatives* to velare (e.g. Spanish /S/ > /x/). But I think his reasoning
makes sense.

> Note that in Baltic (a geographically marginal satem branch), some words
> have /k/ where other satem languages have /s/.  Do they justify
> reconstructing yet another series of stops between palatal and velar?  I
> don't think so.

That's Sihler's position too. Apparently the satem/centum isogloss wasn't
really clear-cut, and borrowings across it probably played a part in the
mixture of forms.

Eric Christopherson / *Aiworegs Ghristobhorosyo
Conlang code:
CU !lh:m cN:R:S:G a+ y n2d:1d !R* A-- E L* N1 Id:m k- ia- p+ m- o+ P-- d* b+++ lainesco