On Wed, 11 Feb 2004, P&G wrote:
> (a) You now say that you see both the -n- infix and the -o- infix as
> metathesised, one a metathesised suffix, the other a metathesised
> prefix. I
> don't believe you had said that before - hence my search for an
> explanation.
> (b) You also repeated what I had missed in reading, that you believe the
> -o-
> grade in the perfect is different from the -o- grade in nouns and
> causatives.
>
> So what I understand your suggestion to be now is this:
> For causatives, the sequence:
> ****men-éye-ti > *** mn-éye-ti > ** o-mn-éye-ti > *mon-éyeti.
> and for nouns, something similar, with accent on the thematic vowel.
> Or have I misunderstood?
Yes, a little bit. That means simple questions are necessary - sorry I
reacted to what I wrongly sensed as a condescending note. This needs all
the redundancy we can find:
The four-star prestage must also have had (the prestage of) the
prefix/infix. If we write it as /O/ (meaning consonantal prestage of
later o of this kind), the full word was *O-men-éye-ti (reservations for
the exact shape of the thematic vowel -e/o- and the primary marker -i).
Now, by metathesis this became *mOen-éye-ti, by ablaut reduction then
*mOn-éye-ti, and finally by infix vocalization it ended up IE *mon-éye-ti.
I hope this clears away a misinterpretation. It should make it clear that
the following questions are beside the point. What you address in the
following, are indeed debatable issues, and I also have an opinion on most
things about them, but they are in my view not really connected with the
prefix-turned-infix. You simply bring in some cases of PIE /o/ from other
sources.
> This, I believe raises other problems. (I've pinched the examples from
> Sszemerenyi.)
> (a) What of forms without thematic vowel, such as (in Greek) eupáto:r,
> dysáno:r, áfro:n, dó:to:r, daímo:n, génos, and the forms with -eh2 such
> as
> forá?
The words in *-o:n and *-o:r are regular unaccented variants of accented
*-é:n, *-é:r. I basically accept Szemerényi's derivation of, say, *-é:r
from original *-ér-s (that is stem *-ér- + nom.sg. morpheme *-s) which
underwent a lengthening caused by the nominative marker (I am not keen on
an intermediate "*-érr, but that is a minor point). I assume that the loss
of unaccented /e/ went via an intermediate stage with o-timbre, i.e. e > o
> zero. Then if the lengthening caused by the nom.sg. marker operated on
the intermediate form, we get what we find: *'-er-s > *'-or-s > *'-o:rs >
PIE *'-o:r. That explains the first five of your examples.
génos is harder. Schindler made the analysis of *mén-os (Vedic mánas, Gk.
ménos 'thought') that it was a secondary reshaping of a regular form
*mén-s. The word is a neuter, so we simply expect the unaccented suffix to
go into the zero-grade without any ado. And he actually pointed to a
reflex of *mén-s in Avestan maNs dada:ti 'makes a thought, thinks', also
m%:ng ... dada:ti with tmesis. Thus the general story is that *méns is the
old an regular form, while *ménos is due to a younger analogy.
I have had some trouble coming to grips with the exact nature of the
presumed analogy involed in *ménos. It would work with adjectives:
end-stressed type non-neuter *-é:s, neuter *-és, barytone variant
non-neuter *'-o:s, neuter *'-s; reshaped to barytone non-neuter *'-o:s,
neuter *'-os. Fine, but the s-stems in *-os are not adjectives. They have
opposing adjectives, which may be good enough to bring about the
analogical pressure: Vedic adj. apás- 'active' has masc. apá:s, ntr. apás,
so its substantival counterpart is áp-as.
There may be an additional source, or an additional possibility of the
only source, for that matter: Schindler observed, but did not explain,
that laryngeal-final stems most often have s-stems of the type *-&-s, such
as *kréw&2-s (Gk. kréas, Ved. kravís). Perhaps that holds the key: If
reduction of *kréwH2-es via *kréwH2-os to *kréwH2-s changed faster than
the development of *mén-es > *mén-os > *mén-s, there would have been a
period with *kréwH2-s, but *mén-os. Now if at this point the laryngeal
came to be realized with a syllabic peak (or a prop-vowel), then both
types would have become disyllabic (again). And then it is just
conceivable that the disyllabic form exerted such an influence on the
laryngealless type that it kept it from being reduced to a monosyllable.
Then *mén-os would be a case of retention motivated by morphological
pressure. And if the same pressure was not felt in the set phrase *mén-os
*dheH1- 'make a thought, think', that may have been free to develop
regularly into PIE *mén-s *dheH1-. One way or the other, *mén-os does not
necessarily violate truths we hold to be self-evident.
The type *bhoráH2 from older *-é-H2 *is* the o-infix formation. That would
be, in four-star time, *O-bher-é-H2, whence *bhOer-é-H2 > *bhOr-é-H2 > PIE
*bhoráH2. If *pórnaH2 'prostitute' is derived from a men-derivative
*pér&2-mn. 'trade', then *bhoráH2 is derived from a root noun
*bher-/*bhr-. Give the relations I would like that root noun to have the
meaning of an action noun, but then comes the nasty question: Where are
the examples? Still, we do get quite a bit of coherence in the chaos, and
perhaps we should not expect to find the original point of departure in
pristine form anyway.
> (b) What of those where the -o- grade takes the accent, such as fóros,
> fó:r,
> aidó:s, gónu?
> (c) Why should the initial combination gn- be a problem in nouns (leading
> to
> o-gn) when it is not in verbs?
The type *bhór-o-s 'act of carrying' must have been created by simple
accent variation of *bhor-ó-s 'carrying (adj.)'. The latter is the o-infix
formation of which *bhor-á-H2 is the collective (or the feminine with a
noun understood, as 'carrying [activity]'?). The "contrastive accent",
which can switch a noun into an adjective or vice versa, is active at all
stages of the linguistic history, so *bhór-o-s was formed after the infix
had been vocalized in *bhor-ó-s.
Gk. phó:r 'thief' is the regular form of a root noun with underlying long
/-e:-/. It has a synonym kló:ps, and both roots *bher- and *klep- have
been found to form occasional derivatives with an additional mora in the
root vowel ("Narten ablaut", or even "Narten roots"). Therefore, it is not
unlikely that phó:r is based on a stem *bhe:r-. Then the nom.sg. would
have been originally *bhé:r-s, whence, with the lengthening caused by the
nominative marker, first *bhé::rs with a superlong vowel. If superlong
vowels are realized with two syllabic peaks, one of them must be
unaccented, and if unaccented /e/ becomes [o], the result will be
something like either *bhéors or *bhoérs, which would both produce *bhó:rs
by contraction and end up being PIE *bhó:r.
Gk. aidó:s, telamó:n and many others show accent on the second syllable at
variance with the full-grade of the root. The original paradigm must have
been mobile, as such structures are, so the result is just a fortuitous
encounter of the ablaut of the strong cases and the accent of the weak
cases. If mobility was given up this is one of the forms its ruins would
be expected to have.
*g^ón-u is hard, but perhaps not desperate. The structure is the same as
in *wód-r 'water'. The existence of e-forms like Lat. genu, Hitt. ge:nu,
which in the type of Hitt. gen. wetenas are confined to weak cases,
indicates that we have to depart from lengthened-grade formations. That
can lead to o-timbre in the collective, which then is what I assume form
*g^ónu, *wódr. In roots ending in a single consonant we may have vowelless
suffixes (as Szemerényi has shown for i- and u-stems), so we can depart
from *g^é:n-w-H2 'knees' > *g^é::nwH2 > *g^éonwH2 > *g^ó:nwH2 > *g^ónwH2 >
IE coll. *g^ón-u-H2, whence sg. *g^ón-u. The individual steps are the
usual ones: The collective marker *-H2 triggers the same lengthening as
the nominative sg. marker *-s, hence the superlong vowel in the next
stage. At least nominatives with lengthened vowels shorten that vowel
again of the nominative ending is preceded by a cluster (*nókWt-s, but
*pó:d-s; *-o:n, but *-ont-s, etc.), so, if collectives work the same, the
intermediate form *g^ó:nwH2 should shorten to *g^ónwH2. Back-formations
from collectives are very common, as is seen from the many vacillations
with regard to gender. Just one example: Gk. misthós : Slavic mIzda
'reward' must reflect a pair with sg. *misdh-ó-s, coll. *misdh-á-H2, but
Sanskrit mi:d.há-m is a neuter, obviously backformed from the collective.
Thus, IE *g^ón-u 'knee' does not contain infixal /o/.
Jens