Few sound changes anywhere are 100% exceptionless, and the analysis
of exceptions found in the Satem languages (especially Baltic,
Slavic and Albanian) reveals certain subregularities pointing to the
existence of inhibiting contexts where the change was blocked or
less likely to occur (e.g. before a liquid in Albanian and Baltic).
That's also normal linguistic stuff. Leaving all that apart, we
consider the Slavic, Baltic and Albanian languages "Satem" (in
addition to Armenian, Indo-Iranian and presumably Thracian) for two
reasons:
> (1) They show the Satem shift in the first place (no
documented "centum"
> language has anything of the kind, even in isolated cases; the
only
> doubtful case is Luwian, but there the "Satem" layer may be partly
of
> Indo-Iranian origin, partly an illusion).
>
> (2) The proportion of exceptions (whether real or apparent) to
the "as
> expected" cases is insignificant. There are not enough exceptions
to
> call into question the essentially regular character of the Satem
shift.
>
> Piotr
************
For me, the crutial problems are not inhibiting contexts of further
satemization, but stimulating contexts.
If we take a look on derivatives of PIE root *ag^- 'to lead' in
Albanian, we will see that some derivatives undergo satemization,
like: <udhë> 'road' (*ong^-, nasalized o-grade forme. For
evolution /-on-/ > /-u-/ see: kuvend < Lat. conventum,
kushërin/kushërir < Lat. consobrinus), <përudhë> <-- për- +
udhë 'lead', <i shpërdhitë> <-- shpër- + *ag^- + -ito 'agile', etc.
So, in these cases, I think that problem of satemization was caused
by preceding nasals and liquids. Other derivatives: <ogiq> 'main ram
(with bell) of the herd', <ugar> 'fallow' preserved centum forms.
To my view, the same thing happens with Alb. word <uthull> 'vinegar'
< *onk^- + -ull 'sharp, sour'.
Konushevci