Re: [tied] Re: Nominative: A hybrid view

From: Jens Elmegaard Rasmussen
Message: 22889
Date: 2003-06-09

On Mon, 9 Jun 2003, Miguel Carrasquer wrote:

> >
> >There is no such rule in either language. *rs yields /rr/ without any
> >lengthening in both.
>
> Oops, my mistake. The rule in Albanian is *VsR > *V:R (where R is any
> resonant, except that before /m/ there is no lengthening). The following
> examples from Beekes:
>
> jam "I am" < *h1esmi
> yll "star" < *h1uslo-
> dorë "hand" < *g^hesr-

The rule is not correct, nor, however, was my "rs > rr". Apparently it
should just be "rs > r". A safe example appears to be <ter> 'to dry
(trans.)' which must be a causative *tors-eye-. The vowel is the same as
in other causatives like <ndez> 'ignite', i.e. i-umlaut of a from o. The
lengthening is dore", pl. duar is not different from the one of dere", pl.
dyer 'door'. I do not know what has caused it, but in the latter example
it certainly was not an s-cluster; I have guess at the dual marker *-e
which would be appropriate in both, but what would the rules be like then?
The example <yll> is only an example of this if the etymology is correct;
the old connection with sun, sol is still lurking.

>
> Of course, this has no bearing on *peds-.
>
> For OIr. tír < *ters-, I see no other explanation than lengthening caused
> by the *-s-. Do you?

No, in a way not, but I *have* managed to integrate it. I have assumed a
reshaping of an s-stem of the type pointed out by Schindler, i.e.parallel
with *men-s -> *men-os. Then, if I depart from a root noun (a neuter like
heart) *ters, I can have that develop all by itself in to *te:rs
(supposing of course the sibilant is of the lengthening kind, but why
shouldn't it be?); by normalization this apparent s-stem could be reshaped
to *te:r-os, whence Celtic *ti:ros. Latin terra would be the collective of
the old word.


>
> >> >And then what *is* Avestan daNm ? You said it could be *do:m or *de:m.
> >> >Why then are you making a set of rules where it can't be either? Surely
> >> >the weight of Skt. padí is light since it shows the same stem as all
> >> >the weak cases.
> >>
> >> The locative *is* a weak case, so that is perfectly alright.
> >
> >
> >That is a simplistic view not matching the facts. The locative is a "weak"
> >case only in the sense that the stem is reduced if the accent can get off
> >it, which it only can if there is room for it to move. That is not
> >captured satisfactorily by calling it "weak".
>
> The accent _could_ get off it, and did, and in root nouns caused a long
> vowel in the root syllable to be shortened. The analysis is:
>
> *pa:d-á -> *pad-á (= in your notation: *pe:d-é -> *ped-é),
>
> and then by the inital accent rule and the zero-grade rule:
>
> *ped-é -> *péd-e > *péd + i
>
> I'm using your rules exactly, so what's the problem?

A big one, which I solved the first time around. That would make
*p&2-te'r-i have **-tr-e'. There is nothing to trigger the initial accent
rule in that type, nor does it operate. The accent of Vedic padi is on -i;
I take it to be thoroughly analogical.

>
> >We do not otherwise find that collectives have their own paradigms -
they
> >have their own collectives (the forms they are), but the system cannot be
> >seen to extend any further. I don't think the doctrine ever was anything
> >but an arrogant postulate on Schindler's part which has caught on because
> >it was so difficult to "understand". Wrong statements are. I hope this is
> >not being read by the really influential gurus of our field - I'll be
> >considered a traitor.
>
> My source is not Schindler.
>
> I find that collectives _do_ have their own paradigms, in some cases even
> in the absence of a non-collective (and secondarily "animatized").
>
> "water" *wód-r, *wédnos, COLL. *udó:r, *udéns
> "dawn" *h2áuso:s, *h2áussm., *h2ussós, COLL. *h2usó:s, *h2uséss
> "winter" *gh^éyo:m, *gh^yémos, COLL. *gh^yó:m, *gh^yéms
> "earth" *d(h)é:gh^m, *d(h)gh^mós, COLL. *d(h)gh^ó:m, *d(h)gh^éms
> "dog" (COLL.->animate) *k^wó:n, *k^wéns ~ *k^úns

How did you decide on the pairing of collective nom.-acc. and coll. weak
case forms? I see no evidence for this at all. Believe me, I have been
heavily indoctrinated with this, but I just can't see the basis of it. How
did two clever brains see structures here which cannot be conveyed to me?

>
> >>[...]
> >> Let me rephrase that: what is the collective marker *h2 doing _before_ the
> >> suffix -r/n-?
> >
> >It isn't. It follows the stem, so the coll. ended in *-r-h2. If you're
> >staking anything on Ved. asthnas, sakthnas, I'd say forget it: Indic
> >generalizes aspiration from where it originated to whole paradigms. Would
> >you also see an old aspiration "before" the root vowel in the aorist
> >astha:t? It's the same kind of question - why is one wise and the other
> >silly?
>
> I wasn't claiming silliness. I was interested in your opinion on the
> "bone" word (and the -i/-n- stems in general). There is a *h2 there, but
> it's hard to tell where it's supposed to be exactly, which is a question
> that must be answered before we can call it "collective *-h2" or anything
> else.
>
> Pokorny IEW gives the reconstruction as *ost(h); ost(h)i, ost(h)r.(g), obl.
> ost(h)-(e)n-. The relevant forms are Skt. ásthi, asthnás, Grk. ostéon
> (*osteyon), ostakós/astakós "crab", óstrakon "sherd", óstreon "oyster",
> óstrus "a kind of tree", astrágalos "knuckle". Wenat. (what's that?)
> ostüakon "ossuarium", Lat. os, ossis; ossu, ossua; ossum, OIr. asil
> "member", asna (*astonio-?) "rib", We. ass-en "rib", asgwrn; Hitt. hasta:i,
> Av. asc^a- "shin", Arm. oskr, Alb. asht, Av. ast-, asti-.
>
> EIEC reconstructs OIr. asna as *h2estniyo- and mentions Latin costa "rib"
> and Slavic kostI as problematical.
>
> In fact, every single part of the word is problematical. Latin costa and
> Slavic kostI point to initial *q- (*k-), while Hittite h- points to a
> laryngeal (*h2 or *h3) [and so do the other words, inclusing Latin os and,
> perhaps, Slavic *ostI = Pol. os'c' "fishbone"]. If the laryngeal is *h3,
> the vowel is automatically *o, but if we want to explain the Celtic (and
> Greek) forms with a-, it's better to assume *h2- and *o (obl. *e) vocalism
> of the root. The next two segments are *s and *t, that's more or less
> clear. But what comes next? Generalizing, there seems to be a velar
> element which mainfests itself as *h (aspiration) in Skt., *-k- (ostakos,
> ostrakos) and *-g- in Greek (astragalos) and Welsh (asgwrn) and perhaps
> Armenian (oskr). There is an *n (Skt. asth-n-, Greek *ost-n.-kos, Ir.
> asna-) alternating with *r (Grk. ostrakon, ostreon, ostrus, astragalos,
> Arm. oskr), and, apparently, with *(V)i (Skt. asth-i, Av. asti-, Greek
> ost-ey-on, ostr-ey-on, Hitt. hastai, Slav. (k)ostI). Finally, some forms
> show an *u (Grk. ostrus, Wenat. ostü-, Latin ossu, Arm. oskr if the k is
> from *w).

So we need some differentiation. Now, a neuter is eminently eligible for a
collective in *-h2 (the Avestan i does not go into the paradigm and so
must be from schwa). The vacillating initial in Lat. oss, costa is like
Slav. koza : Skt. aja-/aja:-, probably a sandhi hardening of *-h2#h2-. Gk.
osteon is commonly explained as an adj. of material, but I would like to
know what Hitt. hastai is. I would posit the root as *h2e:st- which would
give coll. *h2o'st-h2, weak cases *h2ast-, from where there is little that
cannot be explained. - The r/n(t) derivative has recently been given the
function of denoting "a single manifestation of a collective" (Melchert)
which seems in place here. The velar increment looks like Skr. asrk/g
'blood' which is equally obscure; Schindler equated it with the -t/d of
yakrt, s'akrt which is perhaps not much clearer (assuming complementary
distribution). The u-part is obscure to me; it looks a bit like the suffix
of Hitt. hasduer 'twigs' (which could also explain /ru/ by metathesis).

On the i/n-stems?? Ah, I actually do believe such exist. And of course, if
r/n(t), l/n(t) and i/n(t) are nothing but variants of a single type, then
hastai is on a par with wita:r 'waters', only with *-o:y, not *-o:r. If
the -r is a phenotype of -n-, then we should probably assume the same for
-l and -y. Sandhi variants?


> >> The Hittite Ablauting hi-conjugation verbs have precisely such an
> >> arrangement (-a- in the act.sg., -e- in the act.pl. and middle).
> >
> >That is news to me. It should be easy for you to substantiate it. Would
> >you please do that?
>
> I can only quote my Hittite syllabus by Theo van den Hout:
>
> "Naast de verba van de mi-conjugatie met Ablaut -e/a- (type app-/epp-
> "nemen" Ib) beschikt ook de hi-conjugatie over een vergelijkbare klasse
> verba (IIb). Hier is de verdeling echter anders; oorspronkelijk
> verschijnen hier de vormen met -e- in de stam in de plur., door
> analogiewerking echter is dit schema regelmatig doorbroken.
> Voorbeelden: sakk-/sekk- "weten", has-/hes- "openen", asas/ases- "doen
> zitten" (Fr. $170):
>
> Prasesens:
> Sing. 1 sa:khi asashi
> 2 sa:kti asasti
> 3 sa:kki ha:si, haszi, heszi asasi
> Plur. 1
> 2 sekteni, sakteni
> 3 [sekkanzi] hassanzi, hesanzi asesanzi
> Praeteritum:
> Sing. 1 hashun asashun
> 3 sakkis hasta asasta, asesta
> Plur. 1 hesuwen
> 3 sekker heser, haser aseser, asaser
> Medio-pass.praet 3sg. hestat
> Participium sakkant- hassant-, hesant- asesant
>

[...]

A form like <hestat> is news to me. The rest are not, nor do I think they
constitute the necessary evidence. I protested against deriving, say,
arhi/erweni from *o/e, since it has /ar-/ in all forms of the middle.
Still, that may be due to levelling, and an old alternation ar-/er- may
indeed reflect the reduplicated perfect type *h1e-h1or-/*h1e-h1r-. Other
verbs may be analogical on this, much like Lat. le:gi: may be analogical
on e:di:. Indeed <hes-> *must* be analogical since the vowel has not be
coloured. The verb asa:s-/ase:s- is plainly reduplicated; I go with the
crowd and analyze it as an intensive which, in my reading, would make it
*h1s-h1o's-/*h1e's-h1s- > asa:s-/*e:s-, levelled to asa:s-/ase:s-. The
function, however, is more like that of the reduplicated aorist, i.e.
causative. The reason for this functional surprise is obscure to me.

This in no way changes my opinion about the history of the hi-conjugation.
Most verbs with the same vocalism as the perfect were inflected like
perfects, i.e. using the old perfect as a preterite and forming a present
to go with it by pure analogy. The perfect stem was mostly dereduplicated,
but soundchange could also obscure the old reduplication, and analogical
forms could copy the resulting structures. This is not dramatically
different from what we find in Germanic or Latin.

Jens