A. Satemic languages
usually have a number of conditioned or unpredictable exceptions from the satem
palatalisation. The latter may be borrowings from non-satem languages or
sporadic survivals of sound-change spread. Indo-Iranian has fewer such
exceptions than Slavic, Albanian or (especially) Baltic, but Armenian has
practically none: whatever could possibly be satemised, _was_ satemised there.
E.g. Armenian loys 'light' < *leuk^-o- contrasts with such forms as OCS luc^I
'ray' < *louk-jo-, Skt. rocate 'shines' < *leuk-e-toi or Av. raoc^ah-
'light' < *leuk-es-.
B. The satem shift was
too complex a change to be reversible. The fronting of the *K^ series was
accompanied by the unrounding of labialised velars and their merger
with the inherited *K series as plain velars (except perhaps in Albanian, but
the evidence for that is weak); there is no indication of a temporal gap between
those processes, which means that at a certain point the whole system of
phonological contrasts was redefined. Secondary "re-centumisation" of Germanic
would have involved the faithful restoration of the former system, reintroducing
etymological labialisation as well as losing all traces of satem palatalisation.
Such things simply don't happen. To use a biological metaphor, mammals can
evolve into marine animals, but not back into fish. The same applies to
Greek.
It is well known that
cladistic analyses of the IE family are easier if Germanic is excluded. Germanic
just doesn't seem to fit anywhere significantly better than it fits anywhere
else. The reason why in some proposed phylogenies Germanic clusters with Baltic
and Slavic is the choice of features used in the "distance matrix" -- if you use
easily borrowable innovations, false affinities due to areal diffusion will
distort the topology of the tree derived from such data. Lexical isoglosses, in
particular, are deceptive. All the languages of the North European Plain share a
lot of exclusively "northern" vocabulary thanks to millennia of contact and
areal convergence, irrespective of their original genetic affiliations.
Non-trivial systemic changes in phonology would be more reliable indicators of
relatedness, but in mechanical computer-assisted analyses few of them are ever
used and they are easily outweighed by more doubtful
"resemblances".
Piotr
----- Original Message -----
Sent: Friday, July 20, 2001 4:54 PM
Subject: [tied] Re: Armenian.
A. What is meant by "more consistently Satemic than any other
branch?
B. How 'bout this one: last year a study was published that
concluded
the Germanic languages were originally Satem, then later
Centumized.
If this is true, then I would suspect that the Satems were the
Kurgan
or (more probably) the Battle Axe people coming from eastern Europe,
assimilating with centum speakers in central Europe, then moving into
Scandinavia.
Might this model have been duplicated by the
proto-Greeks before they
moved into Greece?