--- In cybalist@..., "Piotr Gasiorowski" <gpiotr@...> wrote:
> Notice that no single satem language has labiovelars, whereas they
all preserve *k^w _and_ *kw (*kw and *qw?) clusters. This looks like
a shared innovation rather than strange coincidence, and suggests a
very early loss of labialisation as a feature of dorsal stop
phonemes, most likely contemporal with the satem shift proper. With
*k^/*kW/*k redefined as *k/*kW/*q we get this:
>
> PIE *k > Satem *k^
> PIE *kW > Satem *k
> PIE *q > Satem k
>
> -- with the fronting of the whole *k/*q subsystem, but not of *kW,
which becomes delabialised and merges with reflexes of old uvular *q.
The merger is irreversible, as speakers have no means of knowing
which *k comes from older *kW. Albanian may be an exception (it's
imaginable that in Albanian *q wasn't fronted until rather late and
thus escaped the palatalisations that affected *k < *kW), but even in
Albanian there is no trace of retained labialisation. At best we'd
have a three-way system like *k^/*k/*q, and it's hard to imagine how
it could have "devolved" back into anything centum-like.
>
> The satem group is very sharply defined (there are no satem
developments outside it), which suggests a late fast-spreading
innovation rather than a dialectal feature in PIE times. In the
latter case we would expect a "grey zone" between satem and centum
areas, with fractures partly obliterating the [+/- satem] boundary.
The facts that Tocharian and Greek were unaffected (despite the
numerous affinities between Greek and Indo-Iranian and Armenian), and
that there is a slight south-to-north cline with regard to the
thoroughness of the change _within_ the Satemic group, suggest that
satemisation was initiated somewhere in the southern part of the
Pontic area and diffused northwards at a time when Proto-Tocharian
and Proto-Hellenic occupied sufficiently peripheral locations east
and west of the core area. The optimal timing, in my opinion, would
be after 3000 BC but well before 2000 BC.
>
> Piotr
>
>
So, I hear: a linguistic development dividing two languages, rules
leaving a few stragglers behind...
I wonder if the shibbolethisation mechanism I proposed
http://www.angelfire.com/rant/tgpedersen/Shibbolethisation.html
might come in handy to explain the centum/satem thing?
Torsten