27-04-2004 23:55, elmeras2000 wrote:
> I don't see how one can know anything about this. The only
> compelling reason to posit /c/ would be the observation of an
> opposition /c/ : /ts/ which is not in the system as far as we know.
A contrast between an affricate and a stop + fricative cluster is not
essential for the phonemicisation of the affricate. Spanish /c^/, which
is undoubtedly a phoneme and a segmental unit, is the only affricate in
the Spanish phonemic inventory, and doesn't contrast either with */ts^/
(there's no /s^/ in the language) or with /ts/. English /3^/ is older
than /z^/, and it can't contrast with /dz^/ (an impossible sequence in
English); but /3^/ is an phoneme by all accounts. If my putative *c can
be treated as a single segment and if it contrasts with both *t and *s,
its phonemic status can be established.
As for its unitary character, the evidence is as follows:
(1) Hittite orthography, which clearly treats <z> as a unit; note that
the palatalised reflex of *t (unlikely to be anything else but a true
affricate) is spelt in the same way. The evidence of a language which
retains /ts/ surely carries some weight.
(2) The syllabification of *-tst-, which is clearly /-ts.t-/ rather than
/-t.st-/ (the /ts/ develops just as it does word-finally). Since
syllabification rules normally maximise onsets (if permissible) if the
alternative is to create a complex coda, /-ts.t-/ is odd unless /ts/
doesn't count as two segments.
(3) The behaviour of *ts in "thorny" metathesis (assuming that
affrication was the first stage of the "thorn" development: *tK > *tsk >
*kts (the affricate is metathesised as a whole).
As for the contrast, we find it both word-finally and prevocalically
after dorsals:
PIE Gk. Skt.
*-t zero -t
*-s -s -h.
*-c -s -t
*k^t kt s.t.
*k^s ks ks.
*k^c kt ks.
etc.
> A unit interpretation could also be attractive on the basis of an
> idea of duration. But I do not think we have any such knowledge.
> My phonemicization /setstos/ is meant as a protest against the
> superscript s generally seen in /setStos/ (where S is superscript s,
> just this time). I dislike this because it would mean that potential
> sequences of original *-t-s-t- which also surface as "-tSt-" would
> occasion a rule "-tst- > -tt-", i.e. one of *loss* of /s/ between
> two dentals because the output is one in which the -s- is not
> phonemic; that strikes me as odd if the truth is that an -s- is
> automatically *present* here. But it's no big deal.
I see that as the positional neutralisation of the contrast //-T-t-// :
//-T-s-t-// (and //-T-st-//, if any examples exist), all of them
surfacing as /-ct-/ (not /-tt-/). No big deal, as you say.
Piotr