On 2006-09-14 10:14, tgpedersen wrote:
> I don't understand the compensatory lengthening Latin pe:s,
> Greek pous. I seems to me they are compensated for something
> they didn't quite lose. There is not much phonological sense
> in a rule that removes part of the consonants in auslaut. All
> or nothing I say. It must have been nominative pe:, in composites
> -i-po:, and the nominative -s analogically restored later.
The actual form of the nom.sg. was *po:ds, phonetically *[po:ts], and
simplifications of the final cluster are branch-specific (cf. Skt. pa:t
with regular t < *ts). The length isn't compensatory, or to be more
precise it doesn't compensate for a lost segment. The nom.sg. *-s (but
not just any *-s!) lengthens vowels in the final syllables of
consonantal stems irrespective of whether it stays or goes (as in
*p&2té:r, etc.). It was lost after *r, *n, *s and *j (probably also
after *m and *l), but not e.g. in *wo:kW-s, *k^lo:p-s, *népo:t-s or
*dié:u-s, where we find length nevertheless. Jens's hypothesis is that
the nom.sg. ending was originally voiced *-z rather than *-s, and that
the difference has something to do with the phonetic lengthening
(phonemicised upon the merger of *-z with *-s). The o-colour of the
thematic vowel in the nom.sg. *-o-s is also ascribed to the original
voicing of *-z.
Curiously, the collective suffix *-h2 also lengthens vowels in the same
fashion, but final *-h2 can't have been voiced, judging from the fact
that there is no o-colouring of the thematic vowel before it.
The whole business has been discussed before. Try key phrases such as
"Szemerényi lengthening" in searching the archived messages.
Piotr