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

From: Miguel Carrasquer
Message: 32846
Date: 2004-05-21

On Fri, 21 May 2004 13:27:14 +0000, elmeras2000
<jer@...> wrote:

>--- In cybalist@yahoogroups.com, "Rob" <magwich78@...> wrote:
>
>> Is the genitive *dem-s reconstructed solely on the basis of Gk.
>> despóte:s, or is there other evidence?
>
>The gen. form is seen in Avestan d&:n.g paitis^ and Vedic pátir dán,
>both reflecting IIr. */dans pati-s./ or the reverse. The short vowel
>is safe for the Indic form, but only Greek shows it was an /e/.
>Still, that is generally the way reconstruction works, so it seems
>to be *déms and very definitely is *dVNs (with a short vowel
>followed by a nasal consonant).

Agreed.

>> It seems that the 'house' root is tricky indeed. How do you
>explain
>> the contrast between e- and o-vocalism?
>
>It has become a classic in the literature on acrostatic paradigms
>and rootnouns alike. The doctrine says that acrostatic paradigms
>have é in their weak forms, as 3sg middle Ved. stáve 'is invoked' as
>opposed to a long vowel in strong forms, as 3sg act.
>stáuti 'invokes'. This is "Narten ablaut",

Except, as I said, that's not quite the full story. The
weak forms (pres.pl. stumási, stuvánti, and middle forms
such as stus.e: (2sg.)) have zero grade, as expected.

>and there are nouns doing
>the same, as Gk. hêpar, Avest. ya:kar& vs. Skt. yákr.t, yaknás, Lat.
>jecur 'liver', said to reflect *yé:kW-r(t), *yekW-n- (originally,
>the story goes, gen. *yékW-n.-s, then *yekW-n-ós by normalization).
>Some nouns

Almost all of them, in fact.

>, however, show /o/ instead of /é:/ in the strong forms.
>The nominative to go with *dem-s definitely has an o-vowel, cf. Arm.
>nom.-acc.sg. tun, and 'night' has *nókWt- all over except in Hitt.
>nekuz which is diagnozed as a genitive, this leading to *nókWt-s,
>gen. *nékWt-s, and the model of which one may also take the
>alternant of 'foot' to have been originally distributed as *pó:d-s,
>gen. *péd-s (-> normalized gen. *ped-ós).
>
>I believe we can get all of this regular by positing underlying root-
>vowel length in the acrostatic paradigms. The 3sg act. *sté:w-ti
>needs no further explanation if the stem was *sté:w- all along. It
>weak variant *stéw- may then be a reduction of *pretonic* *ste:w-´
>not unlike the process that led to zero-grade with short-vowel
>roots. There is the difficulty, of course, that the accent is not on
>the ending, but I assume it once was. That assumption, i.e. a
>prestage such as gen. *pe:d-ós, offers a reason for the shortening
>of the root vowel, and I do not think it is as bold as it might
>look. For, when the reduction of unaccented vowels has changed *pe:d-
>ós to *ped-ós, it has changed structures with short vowels into
>monosyllables, as gen. *H2ner-ós (*H2ner- 'man', Gk. ané:r) > PIE
>*H2n.r-ós. At that stage words of normal structure had become
>phonological monosyllables with only one full vowel.
>The "acrostatic" type however had only shortened its long pretonic
>vowel to a short pretonic vowel. Now we need a rule that puts the
>accents right. I think it's free of charge: pretonic vowels are
>accented. That is the same as saying initial accent, accent on the
>first full vowel of every word. Since all pretonic short vowels have
>been lost, there remain only the reduced products of originally long
>vowels to operate on, so if "initial accent" is introduced, we get
>the accent right. That gives us *péd-os. One might like to leave it
>at that, given the Greek and Sanskrit genitive podós, padás, but the
>existence of genitives like *dem-s, *nékWt-s, *gWéw-s (or *gWów-s),
>and 3pl forms like Ved. táks.ati, s'á:sati reflecting *-nti without
>the vowel of the desinence /-ent/ show that the desinential segments
>were subsequently reduced. So we have to assume just that: the stage
>*péd-os was reduced to *péd-s, which is the structure seen in some
>archaic forms, while others adjusted the form to the normal type
>with accented endings, so that *ped-ós is in all probability the
>correct PIE reconstruction for this particular lexeme.

Agreed so far.

>The strong paradigm forms should then be based on structures with a
>long vowel in the root segment. But we often find an o-vowel, as
>in 'house', 'foot' and 'night'. There are no verbs of this kind

Ahem. There's plenty of them. A sample from Jasanoff pp.
74/75:

*molh2-/*melh2-
*bhodh(h1)-/*bhedh(h1)-
*bhorH-/*bherH-
*dhou-/*dheu-
*g^hongh-/*g^hemgh-
*ghrobh-/*ghrebh-
*h2wos-/*h2wes-
*sor-/*ser-
*h2wog-s-/*h2weg-s-
*gWol-s-/*gWel-s-

>, so it looks like a thing that could arise only in nouns. There are no
>neuter root-nouns like this either

What does it matter whether they are root nouns or not? The
type is common enough in neuters (e.g. *wód-r *wéd-n-
"water").

>(not many of any other kind,
>however, I practically know only 'heart', but that's a footnote), so
>I would suggest that the o-timbre has been caused by the nominative
>lengthening. If we expect the product of *pé:d-s in the nominative
>and find *pó:d-s, is it then not the most sensible thing to ask if
>*pó:d-s could be the product of *pe:d-s (*pé:d-z if we need a
>special sibilant)? I guess it is, and the necessary assumption is
>quite easy: *pé:d-s underwent lengthening, and the additional length
>in the already-long vowel caused it to assume o-timbre.

This is of course a sub-optimal explanation, as it fails to
explain the other strong case forms, such as the accusative.

My suggestion is to accept the facts as they are: /ó/ is the
regular reflex of a pre-PIE lengthened vowel **/a:/. The
evolution is trivial and ubiquitous (a: > o: is a normal
development in the history of many, or even most languages:
OE sta:n > ModE stone, PIE a: > Gmc o:, Lith. o:, OPol. /á/
> dial. Pol. (Wielkopolska, etc.) /o/, Dutch /a:/ > /A:/,
/O:/ in many dialects (e.g. Amsterdam), etc. etc.).

We now have *pá:d- > *pód-, *da:m- > *dom- regularly, no
matter what follows, *-s, *-m or plural *-es. The
e/o-variation of the thematic vowel is now also naturally
explained as lengthening before a voiced sound (/a/ > /a:/ >
/o/). The length of /o/ in open syllables, as preserved in
Indo-Iranian (Brugmann's law), the facts of Tocharian
phonology (where short vowels are reduced to /ä/, but /o/
becomes /e/, like PIE */e:/), the fact that *o, like *e:, is
not coloured by a neighbouring laryngeal, all become
understandable.

The few remaining irregularities can easily be explained by
the fact that pre-PIE had two (x2) additional vowels besides
*a(:), namely *i(:) and *u(:), as typologically required in
any case. The short variants *i and *u merged with *a
(probably by way of *&), which gives é when stressed, and 0
when unstressed. The only exception is when *i and *u stood
at the beginning of a morpheme, when the developments are
*-y&, *-w& and further stressed *yé/*í, *wé/*ú, unstressed
*y, *w.

The long or lengthened variants give *i: > *é:, *u: > *ó
when stressed (morpheme-initially *yé:, *wó), but when
unstressed they were reduced early to *i and *u and further
behave like original unstressed *i and *u. This is the
fundamental difference with pretonic unstressed a:, which is
shortened to a, which attracts the accent (resulting in
stressed *é and an acrostatic paradigm). [Posttonic a: is
not reduced at all, and gives /o/].

This explains the Narten ablaut e: ~ 0 (also in nouns such
as *k^í:rd, *k^i:rd-ás > *k^e:r, *kr.dés "heart",
*í:kW-an(t), *i:kW-án(t)-âs > *yé:kWr(t), *(y)ikWnós (>
Slavic jIkra, jIkno "fish rod") "liver", etc.), and cases of
Ablaut *o ~ 0 such as *pú:nt-ah2- > *pónt-(o)h2-,
*pu:nt-áh2-âs > *pn.th2ós).

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