I've been thinking about the alleged
three-way reflexes of the three series in Albanian. *k^ changes into fricative
<th> there (probably via a prehistoric affricate like *ts), but *k and *kW
apparently fail to merge before front vowels, *k being retained and *kW
undergoing palatalisation and ending up as <s>. There are, to be sure, few
good examples of *k- preserved before *e or *i, but some look secure, e.g.
*keh1so- > kohë 'time'.
This situation looks very strange if we
accept the traditional phonetic interpretation of the PIE consonants. Other
things being equal, [kW] is u-coloured and so inherently more resistant to
palatalisation than [k], cf. French [se] < [ke] vs. [ke] < [kwe]. But if
*k = [q] and *kW = [kW], there is a way to circumvent the paradox. We first
get *k > *k^ (> *ts) and *kW > k, whereas uvular *q survives as such
(resisting paltalisation, as uvulars will). By the time *q becomes *k, _older_
*k (from *kW) has already been palatalised, so the eventual merger takes place
only in non-palatalising contexts.
It would seem, then, that satemisation
proper is the *K > *K^ and *KW > *K part, while *Q > *K is a later
development.
Piotr
----- Original Message -----
Sent: Friday, July 20, 2001 8:26 PM
Subject: [tied] Satem shift
Piotr spoke earlier of uvularisation and a reinterpretation of
*k^/*k/*kW as *k/*q/*kW to resolve typological problems in IE. If so, the satem
dialect area would have simply "fronted" *k/*q, forming *k^/*k. A one-step
process rather than a messy, two-step palatalisation + delabialisation
event.