> >One of the marks of the satem group is the merger of *kW and *k.
>
> In Albanian and Armenian, *k and *kW are kept distinct, under certain
> circumstances. This suggests that the satem shift is independent (and
> earlier than) the merger of *k and *kW in Balto-Slavic and Indo-Iranian.
Steady on! This is a little overstated. In Albanian there are a few
(?four) words that show *kW > s before an original front vowel, while *k >
g' in the same context. The lack of counter-examples has allowed us to
suggest that these are the regular reflexes, but the basis for this rule is
still slight! In most cases the outcome of PIE *kW and *k is identical.
With the voiced equivalents, *gW and *g, I believe the evidence is even
slighter (pace Piotr). In 1974 S Mann called the evidence "ambiguous" -
but perhaps more has come to light since then.
I am unaware of the survival of three reflexes in Armenian, and would be
glad of the evidence.
In essence I am not saying your claim is wrong, only that as a bald
statement it is made to carry more weight than it should bear -
overwhelmingly the satem languages merge *kW with *k, *gW with *g, and
(where the aspirates are distinguished) *gWH with *gH.
I like the argument of Piotr some time ago, where he picked up the idea that
traditional *k^ is [k], traditional *k = [q] and traditional *kW = [kW],
then said "it would seem, then, that satemisation proper is the *K > *K^ and
*KW > *K part, while *Q > *K is a later development." He therefore agrees
with you that the merger is after the palatalisation, and I would not wish
to disagree. But it still means that this merger is an almost universal
mark of the satem languages.
Peter