On 2007-09-08 03:57, stlatos wrote:
> You said that wasn't from a laryngeal last time (but l>& after
> non-syl. h dropped causing t>tH).
Oops, sorry. It was an idea that occurred to me in the course of our
discussion. I forgot my own solution so now I've had to go back to my
original posting to check what I wrote on that occasion. OK, I suggested
*tl.h2-tlah2 > *tl.tHlah2 > *t&tlah2 with a "schwa secundum", i.e. an
epenthetic vowel breaking up difficult consonant clusters in some
branches and possibly in PIE as well, even if it wasn't phonemic:
http://www.humnet.ucla.edu/pies/pdfs/IESV/1/BV_rhiza.pdf
It tends to surface as /a/ in Italic and Celtic, /i/ in Greek and /I/ in
Slavic at least when flanked by obstruents. The standard notation for it
in the literature is a raised circle, but what I normally use here is
/&/ without an index.
> How many schwas are there? Do you
> say there's only one "laryngeal schwa"?
No. The syllabic allophones of the three laryngeals must have been
distinguishable in PIE, just as their reflexes are distinguishable in
Greek. They were also distinct from "schwa secundum". The latter was
usually coloured by a following resonant in branch-specific ways. In
Lindeman variants of monosyllabic words with *CR- onsets we get *C&R-,
just as *Cw- ans *Cj- may produce byforms with *Cuw- and *Cij-. It seems
that in Latin *C&RV- developed into CaRV- at least when the resonant was
/r/ or /n/ (I can't think of a secure example involving /l/):
*k[&]r-o:n > caro: (from *ker- 'cut')
*m[&]n-eh1-t > manet
In Germanic, syllabic laryngeals merged with *a in initial syllables and
seem to have been lost in most other cases (as in *Duxte:r). The 'duck'
word *anuD-, however, shows *u, which may be a regular reflex in a final
syllable; however, other examples are difficult either to find or
evaluate, especially as secondarily epenthesised unstressed vowels are
common in Germanic.
The vocalisation of syllabic resonants followed by a laryngeal plus a
vowel is *uR in Germanic (*gWr.h2u- > *kuru-, etc.), and it seems to me
the Lindeman treatment of *CR- is the same, as in the *Bulan- example.
Piotr
Piotr