Jens:
> I must say I am being widely understood quite correctly here which
> is rewarding and rare in itself.
Now if only the treatment was reciprocal <:/
> In the real debate we may perhaps be at a stalemate if we go on
> passing the buck and sweeping the problem under the next carpet.
I'm not doing that. All I want out of you is a firm reason why it
can't be anything other than a z-allophone of *s. However the
explanations you give can always be used to support z-allophony.
The best arguement I guess is...
> I find it hard to regard word-final voicing a preferred solution,
> but who am I to tell?
Alright, so I do an internet search on this apparent restriction
and yes I guess linguists do find final voicing rare... but not
non-existent (and one might wonder whether the assertion is only
substantiated by 'lack of evidence', by dismissing the potential
data from languages which _do_ have it, like Lezghian).
This article is very significant for this debate on IE:
http://home.uchicago.edu/~aclyu/papers/Language80.pdf
The thing that blew my laderhosen right off my body was Figure (3).
It lists possible conditions in which final voicing can arise. One
of the notable possibilities that are predicted by the universal
"LBC" rule is the option (3c) which theorizes a language with "Voice
licensed after V and before sonorants."
!!!
Oh my god! That's IE in a nutshell! Then it says "In Lezgian, a Nakh
Daghestanian language, final and preconsonantal ejectives and voiceless
unaspirated obstruents are voiced in certain monosyllabic nouns, as
demonstrated here with acoustic evidence, thus violating this major
prediction of the LBC approach." (Why does IE have to parallel a
neighbouring NEC language, I wonder, but that's another can of worms.)
The point is that Lezgian is like IE in that respect and option (3c)
describes IE well. Interestingly, the rule also asserts that:
"The universal and fixed nature of the constraint hierarchy
‘precludes the existence of grammars in which voicing is
neutralized finally but not before obstruents and more
generally grammars in which voicing is licensed in a less
informative context than the ones where it is neutralizedÂ’
So does IE show voicing of *s before obstruents? Yes! It's that
pesky *nisdos example again which would be pronounced ['nizdoz]!
Just like in Lezghian too.
It therefore doesn't violate anything. The pattern in IE is instead
_predicted_ by the same rules that you claim dismiss it. That's
kind of exciting.
> One possibility is now the one I have suggested by positing nom. *-z
> as opposed to 2sg *-s. That would cost a phoneme, but that's all. Is
> that to be avoided at all costs?
Yes, see above, address it.
> I can't see why it should be, but that is not really the point; the
> point is whether it reflects the truth. So the alternatives should
> be considered with an open mind,
It hardly can reflect the truth. You focus on just the phonetics and
put blinders on for the rest. You ignore the etymology. The etymology
can only say that they are _not_ different phonemes from those
in "voiceless *s". Even Miguel asked how can the animate-inanimate
pattern *-s/*-d and *so/*to- be a coincidence? It's common sense.
People readily see the obvious connection and I don't know of any other
connection that comes close to making as much sense.
The phoneme appears to be *s, and yet alternates between voiced and
voiceless. Ergo, accounting for ALL the facts, both phonetic and
_etymological_, z-allophony is definitely preferred over a new phoneme
that unnecessarily destroys these etymologies and then opens the door
to more questions of where all these morphemes come from, all for
a lousy new phoneme.
I can certainly handle _one_ bizarre feature in IE as opposed to a
number of them in your theory:
1. R-infixes that become *o (rare)
2. your account of *V > *e/*o diffusion by avoiding
the optimal and commonplace choice of _lengthening_
before voiced consonants (rarer and rarer)
3. double length (rare, rare, rare!)
My theory is doing just fine in comparison, thank you.
> If the nominative marker was just /-s/, realized with subphonemic
> voicing [-z], the nominative is all right, but then the 2sg ending
> must be given a different form. We would like to posit *-t for that
> anyway at some point, and it should of course be considered if it
> could still be *-t at this stage.
Alright, so you would say that wherever *-s derives from *-t, that
*s was voiceless and wherever it derives from plain ol' *s it's *z.
Then since the aorist and 2ps derive from forms in *t, they must be
voiceless but then that works against your view. It also doesn't
explain why the genitive is _voiceless_ in *-e-syo when we know
that it is from an earlier form in *s, not *t! Oh that's right, I'm
supposed to drop all the logical etymologies and support a view that
infuses rarities in preIE and disconnects it completely from Uralic,
right? Whatever.
> I have no problem accepting *-d for the pronominal neuter
> a simply /-d/, but there is a clash between 2sg *-s and 3sg *-t
> which cannot both be /-t/. So, do we conjure [...]
Conjecture nothing. As I said, you're confusing four thousand
years together. What can come of such nonsense. The *-t had
become *-s very early. It had to in order to show up in Tyrrhenian
with plural /-r/, a rhoticized former sibilant.
> I do see indications that the 3sg marker was once a cluster /nt/ or
> rather a unit phoneme /Nt/,
No sense and divorced of any connection to Uralic but that won't stop
you. The 3ps is assuredly bare in the oldest stage and anyone can
see how *-t < *ta "this". That's why *-t is only found in the 3ps
and 3pp. Demonstratives go well with the 3p.
> Now, if in the stage with *-z, *-t, *-Nt, thematic vowel *-o- (or
> its prestage) was selected by the voiced *-z (leading to *-oz > *-
> os), and not by the voiceless *-t
Okay, I can't even discuss this with you seriously anymore. You've
gone crazy with layers of conjecture that are too sloppy for me
to listen to.
> To that environment belonged the preceding vowel, while in the case
> of pre-PIE *-z there is quite strong evidence that there was no such
> vowel present at the relevant time.
Option 3c of LBC.
= gLeN