--- In
cybalist@yahoogroups.com, Miguel Carrasquer <mcv@...> wrote:
> I wrote the following on sci.lang:
>
> On Wed, 31 Dec 2003 17:39:27 GMT, "Peter T. Daniels"
> <grammatim@...> wrote:
>
> >I'm no fan of Henning Andersen, so if you can refute his 1968
article,
> >after explaining what he said (Collinge's summary isn't terribly
clear),
> >I'd welcome it.
[...]
As I understand it, the main point in Andersen's reasoning is that
both branches of BSl. generalized the sibilant that could only come
from *s and not *k^. It's a matter of morphophonemic identity:
Slavic /x/ is different from the non-alternating /s/ (*k^), and
Lithuanian /s/ is different from non-alternating /s^/ from (*k^). In
both cases there is generalization of the alternant of restricted
distribution.
To enter the relevant universe one may note that Henning Andersen
proposes an interpretation of the ruki rule in terms of markedness:
The vowels /i, u/ are non-optimal as far as vowels go, velars are
non-optimal consonants, and /r/ is even non-optimal with regard to
an interrupted : continuous opposition. Thus, *by assimilation*, /s/
in the position after these marked phonemes is replaced by a marked
variant.
Andersen says such abstract assimilation is frequent. He must be
right, if only because any phoneme is either marked or unmarked, and
whatever it is changed into will also be either marked or unmarked,
so half the time a changed segment will correspond to its neighbour
with regard to markedness (and since most segments have two
neighbours, it may be quite close to all the time).
Jens