Re: [tied] *h3 (More deja-vu)

From: Jens Elmegaard Rasmussen
Message: 15818
Date: 2002-09-30

On Mon, 30 Sep 2002, Miguel Carrasquer wrote:

> >>So you would like to have two sources for PIE *H1 - why only two?
> Surely
> >>you are not being guided by the principle of not inventing entities you
> do
> >>not need, so why stop here?

> >I'm being guided by the observation that the vast majority of languages
> have
> >words beginning with V- (with or without automatic glottal stop ?V-) as
> well as
> >hV-, if they have /h/ at all.  A matter of common sense.  In practice,
> since the
> >matter is mostly unknowable (barring considerations of a Nostratic
> nature), I'm
> >perfectly happy to adopt an algebraic position and use the symbol *h1.'

I do not think this applies to Semitic, at least it is not what I always
hear, not being anything near an expert on any kind of it myself.

> Put differently: Hittite zero-grade forms like as-, ad- (*&1s-, *&1d-)
> force me
> to accept that PIE had no words starting with a vowel, but they do not
> force me
> to accept something as unnatural as that the initial phoneme in all such
> cases
> was /h-/.

I do not see how Hitt. as-, ad- can force one to any such inference. If
laryngeals not adjacent to syllabics were realized with a miniscule
prop-vowel which took the shade of the laryngeal itself (as we see in the
three prothetic vowels in Greek), then all we need to get the Hitt.
reflexes out of *H1s-, *H1d- is a change of the prop-vowel to something
written in a way that makes us transcribe it as <a>, and then of course
loss of the consonantal *H1 as always.

> I accept that PIE had a /h/ (since it had /bh/, /dh/ etc.), and I accept
> that in
> cases like *h1t > *th, *h1 must have been /h/.  On the other hand, in a
> form
> like *h1wih1k^m.tih1, I'm willing to bet that the first two *h1's were
> surely
> [?], not [h], and I suspect the last one was [ç] (that's three, and I'll
> stop
> here).

Some of the cases of -Ht- > -th- actually have -H2- which is then found to
aspirate to both sides. So the statement that such a thing *must* be /h/
is false for (at least) one of them. I see no problem with /H1/ and /H2/
sharing a feature and having in part the same effect (in addition to their
differences). As long as the case story of 'twenty' is not parallelled by
anything really resembling it, it remains a case story and, being a
numeral, may quite well be a spontaneous event with no claim on the status
of regularity. All I see is lack of the initial *d-, which is not so
strange since the second part has /dk^-/ underlyingly. We are yet to find
a rule replacing the second -d- with simple length, but that is a separate
problem encountered with all decades. I am not quite sure whether the
first part is in the stem-form of *dwo-, which in compounds always
surfaces as *dwi-, or is instead inflected in concord with the final
ntr.du. with an ending *-iH1 - or both for that matter. Since
*(d)wi-(d)k^mt- and *(d)wi-iH1-(d)k^mt- equally yield einzelsprachlich
*wi:k^m.t-, I see no way of really knowing. Still, if Greek /ewi:k-/
points to *H1wi-, as I am quite willing to believe, then the first part
was most probably inflected; for that would give *dwi-iH1-dk^mt-iH1, in
which the initial could then be not only dissimilated (against the second
-d- while it was still there), but also assimilated, a series of events
that would account fully for the resulting PIE *H1wi-iH1-(d)k^m.t-iH1. I
find it very strange that you take the two last laryngeals to be
originally different: to me they are the same morpheme in concord: "two