Back to the *h3 business.
(a) I don't think it's correct to derive
the attested perfects of *HeC- stems (in any branch) from the formally expected
reduplications *He-HoC- or *He-HC-. Perfects with a lengthened vowel must have
developed very early, since the hypothetical alternation *HeHoC- : *HeHC- would
have produced extremely untransparent vowel patterns after the loss of the
laryngeals. The pivot of the shift was the perfect of *h1eC-, where *h1eh1C-
(and perhaps *h1e-h1od- as well) ended up as *e:C- (possibly well before the
loss of *h2) producing present : perfect alternations like *ed-mi :
*e:d-h2a and providing a model for later vowel-initial stems.
Latin o:di may be a perfect of this type,
from the root *od-, with morphological lengthening rather than reduplication.
But apart from e:di, e:mi, etc., quite a few Latin stems with an initial
consonant also occur with a lengthened vowel (fodio : fo:di; lego : le:gi;
scabo : sca:bi), and whatever their origin, neither the length nor the colour of
the root vowel is of laryngeal origin. Why shouldn't o:di be classed together
with fo:di in a "single pattern for verb formation"? The Oscan form is sigmatic,
and that's another circumstance that makes one expect vowel lengthening rather
than reduplication (as in Latin ve:xi, Slavic aorist *vesU <
*we:g^H-s-).
(b) "Attic" reduplication. The idea that
these forms could be explained with the help of the laryngeal theory was first
proposed by Kurylowicz, but there were so many problems with it, that Kurylowicz
eventually abandoned his own proposal. First of all, the normal IE
reduplication pattern is *h1e-h1l(o)udH- rather than *h1le-h1l(o)udh-, and
definitely not *h1de-h1d- for *h1ed- or *h2ke-h2kou- for *h2kou- (Greek ake:koua
: akouo:). Other languages support *e:d- as the perfect of *ed- (see above),
and Greek ede:da apparently represents an attempt to restore transparent
reduplication in the perfect. Similarly, ele:lutha represents "refreshed"
*e:l(o)udH-.
The productiveness of this formation is
evident from its application no matter if the initial vowel is inherited
(*h1ed-), prothetic (*h1leudH-) or secondary (as in egre:gora from egeiro:).
Since Attic reduplication is a secondary phenomenon which arose in all
likelihood after the vocalisation of initial laryngeals in Greek, it doesn't
provide any independent evidence for the "colour" of those laryngeals; the vowel
it shows is predictably the same as in the base form.
Piotr
P.S. I've got nothing personal against *h3;
I'm only playing the devil's advocate in order to show that the evidence for
that particular laryngeal is far less substantial than for *h2 or even for
*h1.
P.
----- Original Message -----
Sent: Sunday, November 26, 2000 8:27 PM
Subject: Re: [tied] How many laryngeals?
(a)
Latin o:di < ?h3e-h3d-ai. This is difficult to explain as
h2o-h2d-ai.
If we do not allow o< eh3, and insist on o< oh2, then we
have a different
pattern here. It would be a unique example of
early reduplication with an
o vowel (later ones such as momordi are merely
vowel harmony within the
perfect, where the perfect vocalism depends on the
present). And it's no
use running to o-grade
perfects. There are none in Latin with
reduplication.
There is also the parallel example of Oscan uupsens =
/o:psens/ < h3e-h3p-
(meaning they did, Latin fe:ce:runt; stem found in
Latin
opus).
o < eh3 allows us to keep a single
pattern for verb formation, while
o< oh2 makes us postulate different
patterns for these two verbs.
(b) Greek "Attic reduplication" (which is
not limited to Attic at all). I
mean perfects of this
pattern:
opo:pa, odo:da, olo:la, omo:moka etc.
If we do
not allow o< eh3, and insist on o< oh2, then we have a
different
pattern in these verbs from those in -a- and -e-, such
as:
aka:koa (Attic ake:koa), ele:laka,
ele:luTa, and so on.
o < eh3 allows us to keep a single
pattern for verb formation, whil o< oh2
makes us postulate different
patterns for different verbs.