On Wed, 7 Nov 2001 18:42:01 +0100, "Piotr Gasiorowski"
<
gpiotr@...> wrote:
>In general, I don't accept the Beekesian reconstruction of PIE accentuation and its prehistory. In particular, I doubt the existence of *CéCo:C(s) as a distinct non-static type (though there are nouns which according to Greek, Germanic or Vedic evidence were stressed similarly to *népo:ts -- *swéso:r, *ték^so:n, *h2ák^mo:n, etc.). I regard the accentuation *C(e)Có:C(-s)/*C(e)CC-ós as historically primary and normal, *CéCo:C(s)/*CéCC-os as secondary and due to the appearance of lexical accent that enforced stress retraction (presupposing an earlier stage with no lexical accent and "normal" stress, except in transparent derivatives of late origin), and *CéCo:C(s)/*CeCóC-m/*CeCC-ós as invented by modern linguists.
Agreed, I think, on the latter. However, I don't see how the type
CéCo:Cs (mostly with analogical hysterodynamic weak forms C(e)CCós)
can be explained away so easily. It is, I would say, the most common
type of IE noun, in any case far more common than the "normal"
hysterodynamic type with final accented syllable (the words in
unaccented -o:n are far more common than those in -é:n, likewise for
-o:s vs. -é:s. While -o:r vs. -é:r and -ont vs. -ént are maybe both
about equally common, I have no evidence at all for the types -é:m,
-é:t [maybe > *-e:h1-stems?], -é:is [maybe Hitt. <utne>?]).
Not only that, but even in the proterodynamic type there are things
that cannot be explained independently from the CéCo:C(s)-type. I
have long suspected that the s-stems, with their typical -os, -ésos
vocalism are manifestations of essentially the same thing that we see
in the pseudo-hysterodynamics [amphidynamics] in CéCo:R(s), and now I
believe I am sure of it. I have given less thought to the
"collective" type, of the structure C(e)Co:C [accentuation CéCo:C or
CCó:C?], but it may be a related phenomenon as well.
If we restrict ourselves for now to the nom. sg., there are three
accentual and vocalic patterns that are regular: unaccented -Co:C(s),
accented -Cé:Cs and acrostatic -CC(s) [i.e. usually -CR.(s)]. For the
prehistory of the acrostatic type, I think Rasmussen's theories are
adequate: the acrostatic type represents words that had an original
long vowel in the first syllable, a long vowel that was unstressed in
the weak forms, became shortened, but finally attracted the accent to
itself (e.g. *yé:kWr, *ye:kWén(o)s --> *yé:kWr, *yékWn(o)s). This
also explains the "Narten forms" in the verbal system. Rasmussen is
left with the problem of unexpected */o/ for */e:/ in a large number
of strong forms of the acrostatic type (e.g. *wódr, *wédn(o)s, etc.),
but the problem is solved by simply assuming that this is the normal
development **/a:/ > */o/, which then also explains the vocalism in
the perfect (accented *a: > *o, unaccented *a: > e [in the stem in
Hittite, in the reduplication syllable elsewhere]).
We are left to explain the other two types: -Co:R(s) and -Cé:R(s).
The lengthening in both is simply the (late) result of Szemerényi's
rule (in my adaptation of it), i.e. V > V: before -C(C)s# or -C(C)h2#
(as in the 2/3sg s-aorist, or the o-stem ins.pl -o:ih1s, etc.). The
underlying forms are *'-CoR and *-CéR, i.e older **'-Ca:R and **-CáR.
We are dealing with a lengthening rule (for reasons that are not
relevant here, I think caused by the loss of a final short vowel) that
affects only unaccented syllables. The phonetic justification is
probably that accented vowels were already phonetically (but not
phonemically) long. In summary:
*h2ák^man(a) > *h2ák^ma:n > *h2ákmon-s > *h2ákmo:n, but:
*pah2tár(a) > *pah2tár > *ph2tér-s > *p&2té:r.
And so on for all hysterodynamic/amphidynamic words.
A weak form like the genitive is explained:
*h2ak^mán-as(i) > *h2ak^mána:s^ > *h2kménos > *h2akménos
*pah2tar-ás(i) > *pah2tarás^ > *ph2tr-és > *p&2trés ~ *p&2trós.
Note that the case endings themselves are also affected by the
lengthening rule (which is a good thing, otherwise the zero-grade rule
would have swept most of them away), explaining (in part) such
doublets as G. -os / -és / -s [acrostatic]; Dat/Loc. -o / -é / -0;
Ins/Abl -od / -ét > -éh1 / -t [the Ins. *-éh1 is always accented].
The lengthening also affected the nom. plural ending -es (< **-atu),
but note that this ending does not cause the accent to shift, and is
never accented itself. In the "amphidynamic" type, the plural seems
to be based on the "collective" type with lengthening of the second
syllable (formally identical to the Semitic "broken plural") [or
simply based on the singular]:
*h2ák^ma:n + at(u) > *h2ák^ma:n-a:sW > *h2ákmones
The length of */a:/ is not maintained in the unaccented, non-svarita
position, but the vowel does not yield to zero-grade, which is why we
get -es(W) (not -os(W) or -s(W)) [I have preliminarily called my
lengthening rule the "svarita lengthening rule"].
In the HD type, the ending -es must be analogical (I'm not happy with
that):
*pah2tár + at(u) > *pah2tár-a:sW > *ph2téros ~> *ph2téres
It gets more challenging to fit in the proterodynamic type. Let's
start with the obvious and easy case of the s-stems:
*k^láwas(a) > *k^láwa:s > *k^léwos
*k^lawásas(i) > *k^lawása:s > *k^lewésos
The neuter r/n-stems and men-stems have, AFAIK only static NAsg. forms
(*h1ésh2r(gW), *-mn [can't think just now of a neuter in -mn that
doesn't in fact have lengthened/o-grade vocalism, such as the standard
example *hxnomn "name"]), but I'm pretty confident that traces of
nom.sg.'s in *-or and *-mon are there to be found (in any case they
are there in the collective "plurals" *w[e]do:r < *wada:r-h2, etc.).
The interesting part are the i- and u-stems, which I've only figured
out today. There are three types:
A: N. -is (-i); G. -éi(o)s [normal type, *h2awis, *h2wéy(o)s "bird"]
B: N. -is (-i); G. -iós [e.g. *h3ówis, *h2wyós "sheep"]
C: N. -o:is ; G. -iós [Skt. -a:, Grk. -o:, Hitt. -ais]
(Type D, -é:is, -iós(?) doesn't seem to exist: maybe the Hitt. neuter
<utne>?)
If we assume two original types, one in *-i: (or *-ay-) and one in
*-i, we have:
N. '-i:(a) > '-ay(a) > '-a:y-s > '-o:is
G. -í:as(i) > -áya:s > -éios [ ~ -iós] (type C)
N. '-i(a) > -i: > -i:-s > -is
-ías(i) > -ia:s > -iós (type B)
For type A (the most common one in nAIE), the solution lies in
Hittite: there this type is the paradigm followed only by i-stem
*adjectives*, where the */i:/ may have originated out of the
adjectival formant */iy/. We can reconstruct:
N. '-iya > -i:-s > -is
G. -iy-as(i) > -í:-as(i) > -áyas(i) > -áya:s > -éios.
The u-stems (A. -u(s), -éwos; B. -u(s), -wós; C. -é:us, -wós and D.
-o:us, -wós) exactly like the i-stems.
I can't say I have all the details worked out yet, but I think what I
have fits most of the evidence. Let's take the "dog"-word:
superficially, it looks like a root with consonantal skeleton *k^-w-n,
but that would give us AD ++k^éwo:n, ++k^wénos or HD ++k^wé:n,
*k^unés, neither of which fit the evidence entirely. The better
solution is to depart from a proto-form **k^ú:n(a), **ku:nás(i):
*k^ú:n(a) > *k^áwn(a) > *k^a:wn-s > *k^ó:wns
*k^u:nás(i) > *k^awnás > *k^unés
The form *k^ó:wns (*k^ó::ns > *k^a~:ns (?)) eventually gives Latin
<canis>, but is elsewhere metathesized to *k^wo:n(s).
=======================
Miguel Carrasquer Vidal
mcv@...