Hallo conlangers!

On Monday 13 August 2012 20:17:26 And Rosta wrote:

> Alex Fink, On 12/08/2012 16:09:
> > On Sun, 12 Aug 2012 02:39:29 +0200, BPJ<[log in to unmask]>  wrote:
> >> The fact that *b is rare in PIE is a bit of a snag, since the
> >> bilabial implosive is the most common, but I think that the
> >> bilabial implosive may have merged with *w anyway. It's even
> >> possible that original /b/ merged with /w/ causing original
> >> /ɓ/ to lose implosion and become /b/ earlier than the changes
> >> which affected remaining implosives in Indic, Greek and
> >> Germanic.
> > 
> > If it were the latter, /b_</>  /b/>  /w/ i.e. *b>  *bh>  *w
> > chain-shiftwise, we should be able to see it in the root cooccurrence
> > restrictions, against *TeDh *DheT *DeD.  That is, unlike the other
> > members of Dh, *bheD would be illicit but √bheT √bheDh would be okay. 
> > Do we see that?
> > 
> > Come to think of it, this should shed light on *b>  *w as well; we would
> > have no *wReD but some √wReT √wReDh, and probably a bias against simple
> > *weD too.  Do we see that? ... hm, "root" is *wreh2d-, isn't it.  That's
> > not promising.
> 
> So the /b/--/w/ merger would antedate the *DeD constraint;

This would be necessary, yes.

>       and possibly
> between the two, /D/ acquired a realization with a phonation that
> motivates *DeD. Presumably, DeD roots would have become TeD, or DeT, or
> DheD, or DeDh; I don't know if there are statistical patterns or the
> existence of plesionymous doublets (that would indicate different
> dialectal outcomes of DeD) that would shed any light on what DeD tended to
> become, which might then shed light on the realization of /D/ at the time
> the *DeD constraint kicked in (-- I can imagine ejectives changing to /T/
> or /Dh/ (assuming those to be ordinary pulmonics with a normal phonation
> contrast), depending on whether voicing or aspiration was more prominent
> as the basis of /T/:/Dh/ at the time, but you would expect implosives to
> become /Dh/ rather than /T/).

The paper you linked in your second post looks interesting;
I have to examine it more closely.  It also depends on how the
three-grade system came into being in the first place.  The
best candidate for the nearest living kin of Indo-European is
Uralic, which doesn't have it - it has a single set of (plain
voiceless) stops.  Of course, Uralic may have merged the three
grades Tocharian-wise, but the next closest kin seems to be
Eskimo-Aleut, which is like Uralic in this regard, so Proto-
Indo-Uralic may also have had only one grade of stops.  But we
are treading on dangerously unknown terrain here ;)

> Jörg Rhiemeier, On 12/08/2012 20:29:
> > I think the glottalic hypothesis makes a better account for the
> > absence of */b/ than any speculations about a /b/-/w/ merger.
> > If */b/ merged with */w/, why didn't */bh/?
> 
> Just because /b/ merged with /w/, there'd be no reason to expect /bh/ to
> join in the merger. If you mean "why didn't /bh/ rather than /b/ merge
> with /w/", then the answer is simply that the evidence is for /b/--/w/
> merger.

It depends on the features of the phonemes in question. If
*b merged with *w and *bh not, then *bh must have had some
property that allowed it to escape being "taken along" by
*b on its way to *w.  A chain shift *b > *bh > *w looks
more likely, and if it antedated the stop grade assimilation
and the *DeD constraint, it may indeed have happened!
 
> > Also, as Alex says,
> > a chain shift /b/>  /bh/>  /w/ can be ruled out, as that would
> > have created irregularities in the root structure constraints
> > (which are in themselves a piece of evidence for glottalism).
> 
> It can't be ruled out; it just entails a particular ordering of the
> changes.

Yes.  If it happened before the *DeD constraint took force,
it may well have happened.
 
> > In PIE, no two plain voiced stops occur in a single root, while
> > otherwise two stops in a root have to agree in their voicing
> > and aspiration.  PIE */bh/ behaves as it should regarding these
> > rules, and not like */b/.
> 
> The pluses of the glottalic theory are that it can tie together the absence
> of /b/ with *DeD.

Yes.

>       The pluses of the /b/--/w/ merger theory are principally
> that it explains the weird /wr/ and /wl/ and secondarily that it explains
> the high frequency of /w/. I think the latter has the better pluses.

Sure.  These pluses are not to be ignored at any cost.

>       *DeD
> suggests that at some point /D/ had glottalic realization, either [t'] or
> "[d']" (implosive). Assuming that late PIE did have a [dh]:[d]:[t(h)]
> contrast (or was it [Th]:[d]:[t]?),

The former, I think; the reflexes of *Dh are voiced in so many
branches that it must have been voiced.

>       it seems to me slightly easier to get
> to [dh]:[d]:[t(h)] from [d(h)]:[d']:[t(h)] (/d/ becomes pulmonic and
> accordingly aspiration (or voicing) on /dh/ becomes contrastive) than from
> [d(h):[t']:[t(h)] (> [d(h)]:[t]:[th] > [dh]:[d]:[t(h)]).

Yes, but AFAIK, implosive grades do not tend to lack the labial
member.  But ejectives may turn into implosives first before
they are deglottalized:

[t]:[t']:[d] > [t]:[d']:[d(h)] > [t]:[d]:[dh]
 
> In all of this I am merely an interested and not very knowledgeable
> amateur, I hasten to add.

I am also an amateur, perhaps better-versed in PIE matters than
you, but still only an amateur!

--
... brought to you by the Weeping Elf
http://www.joerg-rhiemeier.de/Conlang/index.html
"Bêsel asa Éam, a Éam atha cvanthal a cvanth atha Éamal." - SiM 1:1