On Thu, 05 Jun 2003 18:30:47 +0000, Jens Elmegård Rasmussen
<
jer@...> wrote:
I agree with most of what you say, but I can't resist commenting (again) on
the differences...
>Oh, not again ... "Foot" has *pe:d- in the loc.pl. *pe:dsu reflected
>in OIr. ís 'down', Alb. për-posh 'down' (Pedersen). The basis from
>which Slavic pe^s^I 'going on foot' is formed could conceivable have
>acquired its long vowel by Winter's law, but the others cannot. A
>word with the same o/e alternation is 'house' which forms the
>loc.sg. daNm in Avestan which must reflect IIr. *da:m with a long
>vowel (since *-am gives Avest. -&m). That is part one of the
>evidence showing that paradigms with full grade in the weak forms
>are based on structures with long-vowel root segments.
But *e: has no business being in weak forms like the locative (plural). We
expect loc. sg. *pV:dé(i) > *pédi, and loc.pl. *pV:dsú > *pédsu (or *pedsú
by analogy). Oir. ís (and perhaps Alb. -posh) must be the result of
secondary lengthenings of *e (e.g. OIr tír < *ter-sa). The word for
"house" should have given loc. sg. *dém(i), like gen. sg. *déms (= Av.
<d&:ng>, Skt. <dan>). If the word was m. (f.) and not n., the length in
Av. <da,m> is analogical after the length in both the nom. *do:(m) (*dom-z)
and the acc. *do:m (< *dom-m). I see no reason for original */e:/ in
either word.
>[...]
>There is in a way a third act in this, to wit, suffixed nominal
>paradigms with full root vowel in the weak cases, e.g. *yekW-n- in
>the weak cases of 'liver' (Ved. gen. yakn-ás); these also have
>strong forms with -é:-, IE *ié:kW-r. (Gk. hé:~par, Avest. ya:kar&).
>Some of these forms have -o- instead of -e:-
Almost all of these, in fact.
>, and there is massive
>influence from the normal types. An example with -o- is *wód-
>r. 'water', which has various weak forms: -e- in Hitt. watar,
>wetenas, zero in Ved. gen. udnás. Typical events of normalizations
>included introduction of é/zero ablaut, mobile accent, and full-
>grade inflectional endings, all simply copied from the bulk of
>productive words and needing no explanation.
The Hittite forms are regularly from *wódr, *wédnos. The collective
*udó:r, *udéns is seen in Hitt. witar (pl.), Grk. hudó:r, Skt. udán (loc.),
of which udnás is a normalized G. (*udnós for *udéns).
>[...]
>I can therefore easily account for genitives like *dém-s, *péd-s
>(whence presumably PIE *ped-ós by ending normalization), *nékWt-s by
>positing the root segments with long vowels, i.e. *pe:d-ós > *ped-ós
>> *péd-s (etc.). The o-alternants are harder, but if one takes stock
>of their distribution, it is not diffciult to come up with a
>suggestion. The /o(:)/ belong in the strong paradigm forms. There
>are no neuters among the o/e root nouns,
There are few neuter root nouns...
>so the o-timbre may well be
>associated with the nominative which could do things you would not
>find in the verb. What the nominative sibilant does is lengthen (I
>guess even we agree about that), so if we tag on an *-s (of the
>right kind in case there are more) to the stem *pé:d- and then find
>the product *pó:d-s, the natural question would be: Could it be the
>lengthening, i.e. an additional quantum of length, imparted on the /-
>e:-/ already long that caused it to take on o-timbre? I would say,
>it seem so, and leave it at that. I'd certainly better leave it at
>this here, for the phonetic justification I can now offer for it has
>proved to be too much for the reader's blood pressure. There is no
>need to call for mud that has already been slung. I simply observe
>that where unbridled use of the rules would have produced **pe::d-s
>I find *po:d-s instead, and therefor I assume that /e:/ in case of
>lengthening changes to /o:/.
>
>Now, that takes care of *pó:d-s and *dó:m (m-stems lose the sibilant
>in the nom., Gk. chthó:n, chío:n), and in a way also of *nókWt-s,
>given the existence of the additional rule that the vowel of the
>nom. is shortened if the stem ends in two consonants (as seen from
>the shortness of Sanskrit nominatives in -an, -an* from *-ants, *-
>anks, which are retained in Avestan). The accusatives are a bit
>tricky, but not much: Acc.sg. forms like *pód-m., *nókWt-m. are
>analogical on the "normal type" which has lengthening in the
>nominative and short vowel in the acc.: Gk. ané:r, acc. anér-
>a 'man'. So the acc.sg. is formed like this: Take the vowel of the
>nominative, only short, and add /-m/.
>
>There is another morpheme that causes lengthening just like the
>sibilant of the nominative, namely the laryngeal of the collective.
>Already Johannes Schmidt pointed out that ntr. forms like Gk.
>húdo:r 'water' were in origin collectives (a pair retaining both is
>Gk. tékmar, tékmo:r 'sign, omen' from *kWék^-mr./-mo:r), since the
>same lengthening appears in a number of ntr.pl. types, such as the s-
>stems in Avestan and Old English. Accepting all of that, I can now
>utilize it to explain the o-type of suffixed neuters, such
>as 'water'. If 'water' can have a young collective in húdo:r, could
>*wód-r perhaps be based on the old type of collective? It would look
>like this: *wé:d-r-h2 > *wé::drh2 > *wó:drh2 > *wódrh2.
But the lengthening worked _before_ zero grade, when the word was still (in
your notation) *wé:d-or-h2...
>I have
>assumed, again, that a lengthened /e:/ yields a long /o:/, and that
>the ntr.pl. marker acts like the nom.sg. marker also in the point
>that a long root vowel is shortened if the stem ends in two
>consonants. That would lead us to *wódrh2 with the collective marked
>still sitting on the form. But if *-h2 was still synchronically
>identifiable as a plural marker in neuters, and this was now just
>used as the word in the singular, it would not be surprising if the
>*-h2 was simply taken away, since its message was not meant. And
>that leaves the form *wód-r, Hitt. wa:tar, to go with weak cases
>with *wéd-, Hitt. gen. wetenas.
Does this explain all the neuters with *o/*e, such as:
*wodr, *wednos "water"
*smok^wr, *smek^nos "beard"
*g^onu(r), *g^enwos "knee"
*h2ost(Hi), *h2est(H)nos "bone"
*pok^u(r), *pek^wos "livestock"
*mostr(g), *mestnos "brain, marrow"
*h1oudhr, *h1udhnos "udder"
*k^ouh1r, *kuh1nos "hole"
*h2ongl, *h2englos "charcoal"
*k^okWr, *k^ekWnos "excrement" (also *sok^r, *sek^nos)
*sókWt(Hi), *sekWtHnos "upper leg"
*stomn, *stemnos "mouth"
*wosr, *wesnos "spring"
*h1osr, *h1esnos "autumn"
*doru, *derwos "tree"
*woh1r, *wehros "water" ?
The problems I have with your explanation are the following.
- the accusatives are explained away as analogical
- the neuters are explained away as "old" collectives
All *o(:)'s are then the result of lengthening of *e: by *-s or *-h2
(Szemerényi lengthening).
On the other hand, if we take a fresh look at the data, we see that the
lengthening caused by *-s or *-h2 is already accounted for (nom. *po:ds
with long *o: vs. acc. *podm with short *o, collective *udo:r with long *o
vs. *wodr with short *o). It seems unnatural to apply the lengthening
twice. Moreover, why was it not applied twice in the HD type (*p&2tér-z
(1)-> *p&té:rz (2)-> *p&2té::r = p&2tór ?).
A much simpler solution, in accordance with the fact that *o(:)/*e is many
times more frequent than *e:/*e, is to reformulate your otherwise splendid
theory using instead of */e:/ a long vowel that became /o/ when stressed,
/e/ when unstressed. The natural choice is **a:, of course: /á:/ > /ó:/ is
trivial, and so is shortening of unstressed long vowels (a: > a, and then
*e, just like short **a).
The nominatives with /o:/, accusatives with /o/ and weak forms with /e/ now
become fully regular, as do the collectives with /o:/ and neuters with /o/:
*pá:d-z > *po:ds *wá:d-an > *wódr *wad-á:n-h2 > *udó:r
*pá:d-m > *podm.
*pa:d-ás > *péds *wa:d-án-âs > *wédnos *wad-a:n-ás > *udéns
*pa:d-á > *péd(i) *wa:d-án-a > *wédn(i) *wad-a:n-á > *udén(i)
Of course, this simplification does force us now to look for a different
solution in the handful of cases where we _do_ have /e:/ ~ /e/ (e.g.
*yé:kWr.t, *yékWnos), as well as for the cases where the accusative _does_
deviate from the nominative (*pontoh2s, *ponth2m.). While we're at it, we
can also try incorporate some of the cases of nominal Ablaut o ~ 0 (e.g.
*póntoHs, *pn.tHós).
Well, having recovered **a(:) out of PIE *e/*o, the next most important
"typological" problem is that of **i(:) and **u(:) as full vowels (not
merely zero grade variants of *ey, *ye etc.). I have formulated the theory
that *i: and *u: became *(y)e: and *(w)o:. If this is so, then the
overabundance of /o/ over /e/ in the strong forms simply follows from the
fact that /a(:)/ is more frequent than /i(:)/ and /u(:)/, and that **a: and
**u: together must be much more frequent than **i:. In the weak cases,
**u: and **i: were weakened to *(y)& and *(w)& and then fell prey to zero
grade. This explains why strong *pónt-oh2-s (if from **pú:nt-ah2-z) has
weak forms with *pn.th2-. In the case of *ye:kWr.(t), we would expect weak
forms with *ikWn-, which is exactly what we have in the derived meaning
"fish roe": MIr. i(u)chair (*ikuri-), OCS ikra (besides /i/ also in Pers.
j^igar, Afghan yi(:)na, ON (Germanic in general) lifr etc.). *yékWnos
would then be a normalized genitive.
Returning now to the accusative of *pont-oh2-s (*ponth2m.), we see another
characteristic of nouns with a long vowel (be it **a:, **i: or **u:). As
is also the case with the collective of *wódr (**wad-á:n- > *udó:r not
*wa:d-á:n > *wédo:r [we _do_ have *wódo:r, but that's analogical]), there
was a rule prohibiting two consecutive long vowels. The normal "svarita"
lengthening of the unstressed post-tonic syllable does not take place if
the stressed vowel was long (in fact it does not take place either if the
preceding root syllable is heavy [ends in -CC] or if the suffix itself is
heavy [e.g. *-ent]). We have:
*h2ák-mo:n-z > *ák^mo:n but: *qrawh2-as > *kréu&2s
*h2ák-mo:n-m > *ák^monm.
*h2ak-mán-a:s > *ak^ménos *qrawh2-ás-a:s > *kruh2sós
Note that, because the suffix is unlengthened in the strong cases, it also
loses the stress to a long vowel in the desinence (*kruh2sós not
*kruh2ásos, but loc. (with short vowel in the desinence): *kruh2ás(i))).
This is the origin of the AD (amphidynamic) type.
In the case of a masculine stem like *pont-Vh2-, we have (just before
zer-grade and Szemerényi-lengthening):
**pú:nt-ah2-z > *pónt-&h2-z
**pú:nt-ah2-m > *pónt-&h2-m
**pu:nt-áh2-a:s > *p&nt-&h2-ós
Now Szemerényi lengthening lengthens *& to *&:, which eventually gives *o:
*pónt-&:h2-z > *pónto(:)h2s
*pónt-&h2-z > *pónth2m.
*p&nt-&h2-ós > *pn.th2ós
So originally all PD [proterodynamic] suffixed stems with heavy or long
root syllable (= AD [amphidynamic] stems and ST (static) stems,
respectively). In many such cases, nominative and accusative were indeed
ausgeglichen (I agree with Jens in that), but the Ausgleich went the other
way, at least in the case at hand (OCS *poNtI, Lat. pons, pontem, etc.).
Another interesting (but more complex) case is the word for "lord". I
reconstruct:
**pá:t-in-z
**pá:t-in-m
**pa:t-ín-a:s
The suffix in the strong forms was not lengthened (because of long *a:),
but stressed *í palatalized the following *n, before the stress was
transferred to the desinence (*i, being morpheme initial, did go to *y&
here, not *&, except before *y as in the G.):
**pót-y&n-z
**pót-y&n-m
**pat-&y-ós
With Szemerényi-lengthening and zero grade:
**pót-yo:n(s)
**pót-yn-m
**pét-y-os
Tocharian still has nom. petso ( = *potyo:n), elsewhere the word mainly
became an i-stem (*pót-i-s, *pót-i-m, *pét-y-os). Indo-Iranian does not
have /a:/ (pati, patyur), because /o/ was not in an open syllable in the
strong forms, and it was of course /e/ in the weak forms (analogical /o/ in
Greek, Latin etc.)
But I digress....
=======================
Miguel Carrasquer Vidal
mcv@...