Re: [tied] Re: Bader's article on *-os(y)o

On Sun, 23 May 2004 17:48:34 +0000, elmeras2000
<jer@...> wrote:

>--- In cybalist@yahoogroups.com, Miguel Carrasquer <mcv@...> wrote:
>
>> On Sun, 23 May 2004 14:54:15 +0000, elmeras2000
>> <jer@...> wrote:
>>
>> >Verbal sg. -o- : pl. -e- is found in two places: (1) Hitt. hi-
>> >conjugative sakki sekkweni, arhi erweni, etc. (2) Goth. prt.
>bar :
>> >be:rum.
>>
>> The Germanic case is not ablaut /o/ ~ /e/ at all: it's an
>> alternation /o/ ~ /e:/.
>
>That is not really important; who knows what quantity the Hittite e-
>forms originally had?

Exactly. It's premature to group them together with the
Germanic forms.

>Further down you derive the Germanic /e:/
>yourself from the vowel of the reduplication which you posit long
>but is generally found as short /e/. If you're right about the
>length it's even easier to avoid ó/e.

The problem I have is that the ablaut of the perfect
*should* have been /o/ ~ /e/. And I believe this is indeed
what is seen in the Hittite hi-conjugation, and, as
convingly argued by Jasanoff, in other (present and aorist)
forms elsewhere. Now these are unreduplicated forms, while
the perfect has reduplication. Since the o-grade of the
perfect singular is in my opinion the result of lengthening
(*a -> *a: (o)), and since the reduplication vowel was *i
(lengthened *i: -> e:), the solution is to assume the
following:

In the pf. sg., the root vowel was lengthened (*wayd- =>
*wá:yd > *wóid-). In the plural, the reduplication vowel
was lengthened (*wi-wayd- => *wí:-wayd- > *wé:wid-). The
1/2pl. endings *-me, *-te subsequently attracted the accent
(after the 1/2pl. in all the other verbal forms), and /e:/
was shortened in unaccented position: *wé:-wid-me =>
*we-wid-mé). The 3pl. was more resistant to the accent
shift, and retained reflexes of both *wé:widr.s and
*wewidé:r.

Some related thoughts:
- The difference between sg. and pl. is perhaps due to the
fact that the pf. sg. originally had _no_ reduplication.
Cross-linguistically, reduplication is associated with
plurality (multiple actions -> iterative, multiple
objects/entities -> plural). The IE perfect, hi-conjugation
and middle all go back to a Nostratic stative, i.e. an
originally _nominal_ form ("I am X, you are X", etc.).
If so, the general principle is the same in both sg. and
pl.: lengthen the first syllable of the verbal form.
- We'd expect to see a minority of verbs showing é:-grade in
the singular (lengthened *a:/*u: give /o/, lengthened *i
gives /e:/).
- We'd expect to see dorms with long reduplication vowel:
some 30 Vedic verbs (10%) indeed have a long reduplicative
vowel according to Macdonell's VG for S.

>> >The weak preterite type be:rum, ge:bum, ne:mum is generally
>explained as
>> >analogical on root with initial *H1-. There are not many examples
>> >that can serve as a model: A candidate is *H1em- 'take', so that
>> >ne:mum is perhaps from *H1e-H1em-me >> *e:mum -> *n-e:mum in case
>> >the /n-/ can be credited to a fossilized preverb.
>>
>> I see no reason to assume that.
>
>I said "in case". The assumption is made by many, or at least
>comtemplated, for it's one of those things where one can't know. It
>can be this way or that way.

Is it assumed by the many who assume it for reasons
independent of explaining the Germanic preterite?

>> >Another could be
>> >*H1ed- 'eat', giving *H1e-H1d-me >> *e:tum, OHG âzum, ON átum,
>where
>> >the structure /e:t-/ has been generalized to the whole verb (sg.
>âz,
>> >át).
>>
>> This is a genuine verb with initial *h1-, and the fact that
>> we have /e:/ throughout the preterite shows that it has
>> nothing to do with the type of be:rum, ge:bum, ne:mum etc.
>
>It shows nothing of the kind unless there is such a principle, and
>you cannot elevate the case story of a single verb to a model to be
>followed by all. The generalization of /e:/ in 'ate' does not
>exclude retention of /e:/ in weak forms in other verbs, even if the
>origin is the same.

We would at least expect it in a putative *n-e:m-.

>> The Germanic verbs with /e:/ in the plural preterite are
>> those verbs which have a root structure ending in a single
>> resonant (ne:m-, be:r-) or obstruent (le:g-, se:t-, le:s-).
>> They would have been expected to show zero grade in the
>> plural (as the verbs in classes I (*-oi ~ *-i), II (*-ou ~
>> *-u) and III (*-oRC ~ *-R.C)). Instead they show /e:/,
>> which I would explain as due to the fact that the
>> reduplication vowel in the plural (at least the 3pl.) was
>> originally stressed /é:/ (causing the 3pl. ending to be
>> reduced from -érs to -r.s). Haplological reduction of the
>> root-initial consonsant produces the attested forms:
>> *né:-nm-r.s > *né:m-r.s, *bhé:-bhr-r.s > *bhé:r-r.s,
>> *ghé:-ghb-r.s > *ghé:bh-r.s, etc.
>
>Yes, something of that sort. The point is, Germanic may, why mustn't
>Hittite?

Hittite doesn't have reduplication in the hi-conjugation
forms, so this explanation is not available. (And, given
that ó/é Ablaut is the thing that doesn't need to be
explained in my scheme of things, an explanation is not
required).

>> >We do not find *wednós at all
>>
>> Hitt. wetenas reflects it quite accurately.
>
>In that case, so much the better, but the middle vowel seems to be
>accented. It's based on a single attestation (plene ú-i-te-e-ni),
>though, but the alternative is based on none (no plene -na-a-as, -ni-
>i).

I assume witéni is the locative? In that case, the locative
may very well have been witéni (normalized for *udéni). But
the genitive may still be *wednós > wetenás.

>> >, but it may well be precisely the form
>> >we should posit. The full picture contains *wód-r. and, for the
>weak
>> >cases, the presence of full-grade in the root and of gradation in
>> >the suffix. Rieken posits "*wód-r., gen. *wéd-n.-s (-> *ud-n-
>és)/*ud-
>> >én", assuming a change of the genitive to Hitt. witenas on the
>> >pattern of the antonym pahhur, gen. pahhuenas 'fire', but with
>> >preservation of the full-grade /wed-/ in the root. The collective
>> >has wida:r reflecting *wéd-o:r
>>
>> No, *wed-ó:r.
>
>Yes, directly, sure, but you might read the rest of the sentence
>before blasting out your disgust.
>
>> >, gen. *ud-n-ós (Rieken *ud-n-és), in
>> >Hittite with accent levelling on the second syllable, but with a
>> >unique retention of the full-grade of the root which used to be
>> >accented (and still is in Greek húdo:r).
>>
>> And isn't in Skt. udá:, Hitt. witá:r, Lith. vanduõ.
>
>Yes, the alternations have been reduced. Sanskrit has neither *vá(:)
>d- nor *úd-, but only ud-. Greek has only accented (h)úd-. Hittite
>has no *ud-, and outside wátar only accent on the second syllable.
>Lots of things that would be expected are not found, because the
>variation has been reduced. There may be many ways to envisage the
>course of this reduction, but it appears reasonable to me to combine
>full grade with accent and zero-grade with lack of it where it is
>possible.
>
>> The
>> Greek form obviously has retracted the accent, as shown by
>> the zero grade of *(h)ud-.
>
>That is not at all obvious. If accent and ablaut are in conflict any
>of them can be unoriginal. Then húdo:r can reflect *wéd-o:r with
>subsequent introduction of zero-grade from the weak cases. And Hitt.
>widá:r can reflect the same *wéd-o:r with introduction of second-
>syllable accent from the weak cases.

Considering the fact that Greek is alone in showing initial
accent, and that Greek retracts the accent in other
instances (e.g. kúo:n), I don't think the Greek accent is
decisive here. That being said, *wédo:r > údo:r cannot be
excluded (I think there was a soundlaw *wé > *ú (*yé > *í),
that affected the stressed reflexes of pre-PIE **u and **i,
but sporadically extends to clear cases of pre-PIE *wa and
*ya). I have claimed the same for Ved. s'únas < *k^wénos,
where I also opt for teh edic accent over the Greek one.

A form *wédo:r is unacceptable (unless as a late analogical
innovation) for several reasons: it goes against the way I
think collectives were made [with lengthening _and_ stress
shift], and it would produce a genitive form *udénos, which
is unattested. There are three different genitives to match
(wédnos, udnós and udéns), so failure to produce a proper
genitive (one would say: shooting fish in a barrel) surely
means that *wédo:r can't be original.

The way I see it, there are four different prototypes which
can each explain a subset of the attested archaic forms of
the "water"-word.

nom. *wódr. can be explained as *wá:d-an or *ú:d-an
nom. *udó:r can be explained as *wad-á:n or *ud-á:n

gen. *wédnos can be explained as *wa:d-án-a:s
gen. *udnós can be explained as *u:d-án-a:s
gen. *udéns can be explained as *wad-a:n-ás or *ud-a:n-ás

loc. *udén(i) can be explained as *u:d-án-a, *wad-a:n-á or
*ud-a:n-á. *wa:d-án-a > *wédn(i) (?) seems to fail here.

>> >> Similarly,
>> >> the genitive of *nokWts would be *nekWtós while *nekWts must be
>a
>> >> later contracted form which coincidentally is only attested in
>> >> Hittite in the phrase /nekuz mehhur/.
>> >> Does it exist outside this phrase?
>> >
>> >No, nekuz is still expressly reported to exist only in this
>> >phrase: "das allein in der Verbindung /nekuz mehur/ '(zur)
>> >Nachtzeit' auftritt" (Elisabeth Rieken, Untersuchungen zur
>nominalen
>> >Stammbildung des Hethitischen, Wiesbaden 1999, p. 128). The
>> >monosyllabic form of the old genitive is confirmed by Old Latin
>> >nox 'at night'.
>>
>> Despite Hitt. nekuz and Slavic nekWto- in the "bat" word,
>> the weak form of this word seems to have been *n.kWt-: Ved.
>> aktá: "night", aktú- "dark, night", Germanic *unhtu-
>> "morning".
>
>That is not that paradigm. It's like using stutá- to deny the
>existence of stáve again.

I'm no more denying the existence of stáve than you are
denying the existence of stuvánti.

>> Since we need an **u here anyway to explain the
>> labiovelar, I would reconstruct *nú(:)kt-z, *nú(:)kt-m,
>> *nu(:)kt-ás -> *nókWts, *nókWtm (*nékWtm/*núkWtm), *n.kWtés
>> (*nukWtés), which also nicely accounts for Greek núks,
>> nuktós.
>
>None of this is based on principles I know. It is not nice of the
>presumed /u/ to turn up only in Greek where it cannot be
>distinguished from old /o/.

Please explain. I don't know of any Greek soundlaw that
would turn /o/ into /u/, either before *kW (osse) or after
*n.

>Or am I reading too much into your
>words? The reasoning does not appear to be logical. Could you be
>more specific: do you take Greek /nukt-/ to reflect a form with
>PIE /u/?

With Pre-PIE **u (or *u:), unstressed. This should have
given PIE zero, and it's possible that Greek /u/ just
reflects a schwa secundum-like prop-vowel, coloured to /u/
by the labiovelar (cf. *kW-kWl-ós > kuklos).
On the other hand, I think Greek does have a tendency to
preserve /u/ as the reflex of zero-grade **u(:) in other
cases, such as ónoma/énuma "name" (*h1nú:h3-man-), ónuks
(*h3nú:ghW-).


=======================
Miguel Carrasquer Vidal
mcv@...