On Fri, 10 Nov 2000 22:44:24 +0100, "Piotr Gasiorowski"
<
gpiotr@...> wrote:
>As for "eight", the exact reconstruction of the ending is difficult. Practically everyone agrees it's identical with the masculine dual ending (Brugmannian *-o:/*-o:u). It's by no means certain that it involved a laryngeal, and if it did, the laryngeal may have been *H1 ([h], [?] or whatever) added to a stem-final *-o, rather than *H3 ([xW]). The dual of root nouns (*-e < *-H1(e)?) lends some support to that interpretation. The [w] glide added to *-o: in prevocalic sandhi may be just an automatic onset-filler after a rounded vowel.
I'd say the Kartvelian word is a strong argument for dual *-(e)h3
(=*-(e)xw). That's not to say that *-h1 is not a dual ending (it is).
We can compare the two dual forms with the two plural forms *-es and
*-oi. I mean that quite literally, actually: the forms are built on
the same pattern, but it takes a fair amount of internal
reconstruction and explanation to see why, or to see how I think
why... Allow me to dive deep into the depths of
pre-Proto-Indo-European:
We'll start with the personal pronouns, and see if we can see some
regular patterns in the seemingly chaotic array of forms (the
following is largely inspired by Jens Elemgaard Rasmussen's analysis
of the PIE personal pronouns, although it differs substantially from
his conclusions). Apart from 1sg.nom. *(h1)eg(^)(h)-om, which is
suppletive, we can reconstruct:
1. (*mu(:)) obl. *mué > *mé(:)
2. *tu(:) obl. *tué(:)
Rasmussen derives the acc. ending from *-mé (*mu-mé > *mué; *tu-mé >
*tué), maybe an accented variant of normal *-m. This is confirmed by
the plural acc. forms *ns-mé and *(j)us-mé, and I accept it. For an
older stage (before e/o umlaut), I would posit *-má.
In the nom.pl., we have *wei- (Skt vayam, Goth weis, Hitt. we:s, Toch
*wesI) ~ *me(:)s (Lith. me~s, OCS my, Arm mek`); and *swei- (Hitt.
sume:s, Celt *swe:s) ~ *ju(:)s (Skt. yu:yam, Goth. ju:s, Lith. ju~s,
OCS vy, Toch *yesI, Arm [d]uk`, Alb ju). My proposal is to derive
these from:
1. *mwés < *mwésw < *mu-átu
*wéy < *mwéty < *mu-áti
2. *(y)ús < *swésw < *tu-átu
*swéy < *swéty < *tu-áti
There are some soundlaws involved: Cu > Cw (labialization), Ci > Cy
(palatalization), a > e (fronting), tw > sw (assibilation, cf. Greek
*tw > s), and finally depalatalization (ty > y) and delabialization
(mw > m, w). The development *swesw > *usw is peculiar, but accounts
nicely for certain forms of the numeral "six" as well (OPr. uschts
"6th", Arm. vec` "6" < *uswec`, from a variant *sweswe(-to)- of the
numeral borrowed as *swék^sw(e) from Semitic *s^eds^e(-tu-)).
The oblique forms can be derived from:
1. *mu-atu-má > *mwswmé > *nsmé ~ *nos-
2. *tu-atu-má > *swswmé > *usmé ~ *uos-
The labialized character of plural *-es(w) would explain why we find
-k` in Armenian (*sw > k` is regular in Armenian), maybe also Winter's
Tocharian proto-forms *wesI and *yesI (PToch *I < *u). In the noun,
more evidence of *sw can be seen in the loc.pl. variation Greek *-si
vs. *-su elsewhere, from *-0-sw + -i (cf. loc.sg. *-0 + -i).
Now the dual forms of the personal pronoun can be reconstructed using
the same model:
1. *mu-áku > *mwexw > *moh3 (not attested)
*mu-áki > *mwexy > *weh1 (Goth wi-t, OCS ve^, Toch *we-ne)
2. *tu-áku > *swexw > *(y)uh3 (Goth ju-t, Toch *ye-ne)
*tu-áki > *swexy > *sweh1 (not attested)
obl.
1. *mu-aku-má > *mwxwwé > *nh3wé ~ *noh3-
(Skt. a:va-/na:u, Grk. no:ï, Goth ugk-, OCS na-)
2. *tu-aku-má > *swxwwé > *(y)uh3wé ~ *uoh3-
(Skt. yuvam, Goth igq-(?), OCS va(-))
Greek 2du. spho:ï might be another development from *swxwwé.
There was also a 3rd.p. pronoun *su, attested only in the oblique as a
reflexive pronoun (*sue) and in the du. and pl. largely merged
phonetically with the 2nd, person pronoun (Hitt. -smas, Grk. sphe).
Turning now to the noun, we have in the plural (non-o and o-stems):
nom. *-es, *-o:s, *-oi
acc. *-ns, *-ons
gen. *-om, *-o:m
loc. *-su, *-oi-su
dat. *-bhi-os, *-oi-os
ins. *-bhi-:s, *-oi-:s
These (except the genitive) can now be analyzed as:
nom. *-a:tu > *-asw > *-es (but -k` in Armenian)
*-a:ti > *-aty > *-oi
(*a: shortened to *a when unstressed; *a/*@ lengthened to *a: > *o
before voiced/vowel)
acc. *-m + *-sw > *-ns (Arm. -s suggests a soundlaw *CCw > *CC)
loc. *-0 + *-sw + *-i > *-swi > *-su (but -si in Greek)
all. *-a: \
abl. *-a(:)d + *sw > *-os, with *-bhi-/*-oi- prefixed
dat. *-a /
ins. *-at > *-eh1 + -sw- > *-bhi-h1-s > *-bhi:s (OCS -mi, Av. -bi:s^),
or *-et * -sw- > *-bhi-t-s > *-bhis
For the dual, we can tentatively propose:
nom. *-a(:)ku > *-(a)xw > *-(o)h3
*-a(:)ki > *-(a)xy > *-(e)h1
acc. *-m + *-xw > *-nh3 (not attested)
loc. *-0 + *-xw + *-i > *-xwi > *-(oi-)h3u(m) (Grk. -oi-in, -oi-un)
all. *-a: \
abl. *-a(:)d + *xw > *-bhi-oh3(m), -oi-oh3(m)
dat. *-a /
ins. *-at > *-eh1 + -xw- > *-bhi-h1-h3 > *-bhi:(m)
As to the strange alternation between -atu/-ati and -aku/-aki in the
pl. and du. forms, they seem to be vestiges of ancient active/inactive
(ergative/absolutive) cases, predating PIE inflection as we know it.
This would explain why we find both forms in the nominative,
especially in the personal pronouns (the nominative is the union of
transitive erg. subject with intransitive abs. subject, and personal
pronouns have a strong tendency to move towards a nom/acc. pattern
before other word classes do ["split ergativity"]). It is interesting
to compare the Semitic pattern nom. *-u, acc. *-a, gen. *-i, pl. nom.
*-u: (or *-a:tu, *-a:nu), acc./gen. *-i: (or *-a:ti, *-a:ni). One
possible trace of singular erg. *-u, abs. *-a could be the
demonstrative pronoun:
erg. *tu
abs. *ta
gen. *ti
extended with a deictic element *a: (> *o), giving:
nom. *twa: > *so
acc. *ta: > *to
gen. *tya: > *yo
We also have the pf.ptc. (*-wot-/*-wos-) and other nouns showing
-t-/-s- alternations (*-s < *-tu, *-t < *-ta).
Finally, some words about the verbal inflection. The active (-mi)
forms follow from the personal pronouns postulated above:
*-mu > *-mw > *-m, *-mi
*-tu > *-sw > *-s, *-si
*-0 > *-0 + -t > *-t/*-:, *-ti
The plural was originally made with a suffix *-an-:
*-mu-an > *-mwen > *-men/*-wen, *-meni/*-weni
*-tu-an > *-twen > *-ter, *-teni
*-0-an > *-en + -t > *-e(:)r, *-enti
(*-n > *-r, except after *m ; 2pl. -ter only attested in Toch, -cer <
*-te:r). The morpheme -es was often substituted in the 1/2 pl. (-mes,
-tes)
The "stative" forms use a different set of endings:
1. *-h2 < *-k
2. *-th2 < *-tk
3. *-0 < *-0
1. *-mu (?) < *-mu
2. *-th3u (?) < *-tku
3. *-n(t) (?) *-u (?)
The parallel with Afro-Asiatic is fairly clear:
Sem. A.Eg pPIE
1. *-ku *-kj *-k
2. *-ka/*-ta *-tj *-tk
*-ki/*-ti
3. *-0 *-j *-0
*-at *-tj
1. *-na *-wjn *-mu
2. *-kanu/*-tanu *-twjn *-tku
*-kina/*-tina
3. *-u: *-wj *-u (?)
*-a: *-tj
Enough for now.
=======================
Miguel Carrasquer Vidal
mcv@...