Re: [tied] The disappearance of *-s -- The saga continues

--- In cybalist@yahoogroups.com, Miguel Carrasquer <mcv@...> wrote:

> >> >[MCV:]
> >> >> I'm not inventing anything.  Everybody knows that the
> >> >> acc.pl. comes from *-m plus *-s.  Isn't it obvious?
> >> >
> >> >Yes, that's why I want the *-s to be there, while I observe you
> >> >pulling it off.
> >>
> >> No.  I'm adding it.
> >
> >Yes, you have to add it, because you stroke it from the evidence

in the

> >first place. I just accept it where I see it. I do see what you

mean

> >though: The form *toy is plural already, so the accusative plural

may be

> >expected to add only an accusative marker, which would give *toy-

m. If we

> >find *to:ms, it may be that *toy-m developed into *to:m before

the extra

> >*-s was added (if it was).
>
> Yes, that would be the idea more or less.
>
> But I have to protest again against allegations that I'm
> striking things from the evidence. This is a complex issue,
> and it's almost unavoidable that some of the evidence
> contradicts some of the other evidence.

I'm not insinuating this, I'm saying it flat out. But you are right,
we all have to edit the evidence to one degree or another. I think
we differ precisely as to the degree we will allow this to attain.

> My first statement regarding this issue was:
> "The nominative ending *-s (perhaps still *-z) disappears
> after _all_ PIE sonorants except /w/."
>
> You'll notice that I already allowed for the possibility
> that it was the _voicedness_ of nom.sg. *-z which caused its
> subsequent disappearance, though not without first having
> caused lengthening of a preceding vowel. I think we're
> agreed on that.

Yes, that is exactly the way I see it, rightly or wrongly.

> Then, while I was in Paris over Easter weekend, Glen
> objected that *-s in acc.pl. *-ms does not disappear, and
> you (Jens) countered with a statement that the *-s in the
> acc.pl. (unlike the nom.sg. *-z) does not disappear [of
> course] and does not lengthen.
>
> I'm not ready to follow you there. I don't really disagree
> on the first part (the /s/ in *-ms stays), although I said
> that I can envisage a scenario where the /s/ is only added
> _after_ the working of a soundlaw -Vms > -Vm. The second
> part ("plural" *-s does not lengthen) contradicts a certain
> interpretation of the plural endings acc.pl. *-o:ns (c.q.
> *-i:ns, *-u:ns), and ins.pl. *-o:ys. They are not currently
> my favourite interpretations, but I cannot exclude them
> altogether.
>
> >> >> What we find is an ins.pl. in *-o:ys, which _could_ mean
> >> >> that the "plural" *-s _did_ have a lengthening effect, and
> >> >> an acc.pl. in *-o:ms which _could_ mean the same thing.  So,
> >> >> unless you can show convincingly that those possibilities do
> >> >> not apply, I wouldn't exclude them from consideration.
> >> >
> >> >If the acc.pl. contained a lengthening sibilant we could not

have

> >> >forms like *kWet-ur-m.s (Ved. catúras, Lith. ke~turis), but

would

> >> >have to have something ending in *-wor-m.s . There are no

acc.pl.

> >> >forms of this structure, ergo its *-s did not lengthen.
> >>
> >> Doesn't follow.  I could reconstruct *kWet-wér-ms >
> >> (lengthening, zero grade) kWtwé:rms > (shortening before
> >> CCC) *kWtwérms > (*wé > *ú) *kW(e)túrm.s > catúras.
> >> A form like Arm. c^`ors (*kWét-wor-ms) represents the
> >> analogical type of acc.pl., which is simply the acc.sg. +
> >> *-s.

c^`ork` has c^or- in all (four) case forms and reveals nothing with
specific reference to the stem-formation of its accusative.

> >But wé does not yield ú, it stays wé.
>
> It yields *ú in my view, like *yé yields *í (unless *w- and
> *y- are initial, or a *h2 follows).

But -u- often has a full-grade variant -we-, and -i- a full-grade
variant -ye-.

> The Sanskrit accent is
> original here, as it in in e.g. the gen.sg. of the
> ptc.pf.act. *'-wot-s, *-ús-os.

The accent of the type cakr.vá:n, cakr.vá:m.sam, cakrús.as is
strange because it sort of moves the wrong way. Still it is
constantly on the vowel following the root, and in the type with
asyllabic root dadvá:n, dadús.as, dadús.a:m it is fully columnal,
being constantly on the second syllable. Now, columnal accent is
very like Sanskrit, and not unlike PIE itself, so I assume that the
short type set the model for the longer type in deciding which
segments were to be accented. A columnal accent does not have to
have been brought about by the ablaut alone, it can easily be
secondary. Therefore, this does not prove original accent on -us-,
for which I will assume that it is what it looks like, the zero-
grade variant of the suffix reduced in originally unstressed
position.

> >You are not seriously reasoning with
> >the accusative singular of the word for 'four', are you?
>
> I was of course referring to the accusative singular in
> general. There are two models for the acc.pl.: one where
> the form is identical to the acc.sg. (i.e. a strong case
> followed by *-m) with *-s added (e.g. sg. *h2ák^-mon-m =>
> pl. *h2ák^-mon-m-s; sg. *p&2-tér-m => pl. *p&2-tér-m-s; sg.
> *h1ek^w-o-m => pl. *h1ek^w-o(:)-m-s). The other (as usually
> in Sanskrit) is a weak case, different from the acc.sg. (and
> ultimately to be equated with a syncopated form of the
> gen.pl.), with added *-s (e.g. gen.pl. *&2k^-mén-om =>
> acc.pl. *h2k-mén-m-s; gen.pl. *p&2-tr-óm => acc.pl.
> *p&2-tr-m'-s; gen.pl. *h1ek^w-oy + -m [*h1ek^w-o(:)-m] =>
> acc.pl. *h1ek^w-oy-m + -s [*h1ek^w-o(:)-m-s].

There are a few forms showing that an unaccented suffix segment had
zero-grade in the acc.sg., just as we expect: *tri-dk^m.t-m. in Av.
thri-sat&m '30' (and likewise for '40') and *pónt-&2-m 'path' in
OPers. pa(n)thim. The normal acc.sg. type *swé-sor-m. can be
analogical on the type *p&2-tér-m. by the easy proportion *p&2-té:r,
*p&2-tér-m. = *swé-so:r : x; x = *swé-sor-m.; the zero-grade
accusatives do not lend themselves to this analogy. Conversely, in
the acc.pl. we would not expect *swé-sr-m.s from a preform *swé-ser-
m-s if the plural sibilant lengthened; that would give *swé-sor-m-s

> *swé-so:r-m-s and then remain that way or be shortened to

*swésorm.s (depending on the exact application of the shortening
rule). Do we really have to assume that, by some diabolic turn of
events, the language came to present exactly the opposite? I think
this is taking the lack of respect for the evidence to new heights -
I find you in contempt of court.

Jens