I'm still trying to figure out what happened to the o-stems
in Slavic, accentually speaking.
After Pedersen's, Hirt's and Winter's laws, but before the
Great Acute Reshuffling (Meillet/+Dybo/-Dybo), there must
have been four groups:
1) barytonic masculines (a.p. a, e.g. rakU, zoNbU)
2) mobile masculines (a.p. c, e.g. bergU, bêgU)
3) barytonic neuters (a.p. a, e.g. sito, dvorU)
4) oxytonic neuters (a.p. b, e.g. pero, vêdro, meNso)
Originally, the mobile o-stems would have been scarcely
different from the other mobile paradigms, having an
oxytonic nom.sg. and ins.sg. (but no gen.sg., replaced in
Balto-Slavic by the barytonic ablative), and possibly even
an oxytonic nom.pl. But when the ins.sg. was replaced by a
form in *-mI, and the nom.sg. either became barytonic too,
or was replaced by the acc.sg., the whole singular became
barytonic, which made the mobile o-stems significantly
different from other mobile paradigms.
Incidentally, speaking of the ins.sg. in *-mI: I do not
recall ever having encountered a detailed discussion of its
exact make-up or origin. That the form is old in
Balto-Slavic is shown by Lithuanian, which has i- and u-stem
ins.sg. forms -imi and -umi, and by the Balto-Slavic
accentuation of these endings in the mobile i- and u-stems
(*-imí, *-umí), which shows they are older there than
Pedersen's law (unlike Slavic o-stem *-omI/*-Umi). The
Slavic o-stem form, however, even if recent, may tell us
something about the structure of the ending. There is
general agreement that *-mI is the PIE adposition *bhi, but
what was it added to? I would suggest the answer is the
accusative singular: i-stem *-im-mi > *-imi, u-stem *-um-mi
> *-umi, and later in Slavic *-om-mi > *-omi but also in
part already *-um-mi > *-umi. For neuters other than o-stem
neuters, the ending *-imi would of course have to be
analogical (not **nebos-mi but *nebes-imi), but the
accusative solution also works for C-stem masculines
(*ka:menim-mi > *ka:menimi).
In any case, the rise of the ins.sg. ending *-omi ~ *-umi
sealed the fate of the oxytonic forms in the o-stem
masculine sg.. All the more remarkable, then, that the
barytonic forms which should have been affected by Dybo's
law (*zóNbU => *zoNbÚ) instead merged with the mobile type
(zôNbU, gen. zôNba), as if Dybo's law had been blocked in at
least the NA sg. and NA pl./du. (=> a.p. d) or in the whole
singular and the NA pl./du. (=> a.p. c). In the non-acute
neuter barytones (the dvorU-group), Dybo's law was not
blocked, but the words became masculines, and did not merge
with the already existing a.p. b neuters (peró, vêdró). And
in *that* category, there was no retraction of the accent in
the acute-root forms (vêdro), even though the forms with a
circumflex root (mêNso, jâje) became mobile.
Perhaps the merger of the masc. a.p. b (zoNbU) and a.p. c
forms can be made somewhat more understandable if we assume
that at the time Dybo's law worked, the nom.sg. (and pl.) of
the masculines was still oxytonic. We then have:
*zambás *bergás
*zambám *bérgam
*zambá: *bérga:
*zambaí *bergaí
*zambó:s *bérgo:s
*zambú:m *bergú:m
When *bergás, *bergaí became *bérgas, *bérgai, the
nominatives *zambás, *zambái were pulled along, as well as
the accusatives *zambám, *zambó:s (an oxytonic accusative
was something of an odd thing anyway). That's a.p. d.
Further barytonesis of the singular forms gives total merger
with a.p. c.
The neuter NA's (after Dybo) *dvarám, *dvará: were obviously
not touched by this development. But how on earth can it be
explained that *dvarám and friends did not merge with *perám
and friends? I was reading Stang's Slavic Accentology, and
apparently all Slavic languages that can show the difference
have a barytonic plural in the peró-group (not just Russian,
as I had assumed). Stang gives Russ. vinó, pl. vína, vín,
vínam etc., C^akavian (Novi) kri:lo``, pl. kríla, kríl,
Slovene ókno (< *oknó), pl. ó.kna, ó.k&n, ó.knom etc., Bulg.
(Razlog dial.) pl. séla, vlákna, zrÚna, etc. Stang's
explanation is that this is regular in the loc.pl. *perê~xU
> *pèrêxU (Stang's law), and that it spread from the loc.pl.
to all plural forms, already in Common Slavic. The
intonation indeed appears to be neo-acute on the stressed
syllable, so that makes sense. On the other hand, if there
were a way to explain away the neo-acute and have these
forms retract the ictus already _before_ Dybo's law, that
would explain a lot. We would have an answer to the
question why the dvorU-group remained distinct from the
pero-group, and why acute-root words like vêdró did not
retract the accent (because that would destroy the "new
mobility" in these o-stem neuters). If Slavic ustá "mouth"
(pl. tantum) is accentologically identical to Skt. ó:s.t.ha-
(*Háust-), then it's a word that belonged originally to the
dvorU-group, and not the pero-group. That the words in the
dvorU-group that *did* have a singular became masculines is
easier to understand: there was a gap left by the
zoNbU-group becoming mobile, and it was filled by the new
a.p. b neuters (as well as by some a.p. b u-stems).
Something completely different: Winter's law. If Shintani
is right and Winter's law worked only in pretonic position
(at least for plain vowels), then Lith. ès^ (às^) "I" is
regular (PIE *h1ég^). In the Slavic form, however, Winter's
law did work: (j)a(zU). Following Shintani's rules, this
could only have happened in a form ezÚ > e:zÚ > jazÚ. Dybo
doesn't give a reconstruction of the accent of the 1st.
person sg. pronoun (or I have missed it). SCr jâ and Slov.
jàz suggest either monosyllabic *jâ or disyllabic *jazÚ. In
principle, PIE *h1eg^ can be expected to give Slavic forms
ending in *-g^ => -0 and -zU (like *-t gives -0 or -tU in
the 3rd. person aorist and present/(con/in)junctive), but
the fact that Winter's law worked does rather point to
*h1egóm, does it not? But then the form <ja> becomes
incomprehensible. For Balto-Slavic, a doublet *h1ég^ ~
*h1eg^óm is necessary anyway. In Slavic, perhaps
lautgesetzlich *jè(s) ~ *jazÚ was normalized to *jâ(z) ~
*jazÚ?
On pp. 64-67 of Stang's accentology, there is a complicated
argument concerning the stress of the Lithuanian illatives
and allatives, used by some as evidence that the Lith.
mobile paradigms had originally had final stress everywhere.
Looking at the whole thing from my point of view, I don't
see any problem. Unlike the postposition *-mi discussed
above, which already had given a grammatical case in the
instrumental sg. of the i- and u-stems before Pedersen's
law, the illative and allative forms were built later by
adding postpositions *-na, and *-pi to the accusative or
genitive. When added to mobile words, these postpositions
apparently behaved like stressed suffixes (like e.g. Lith.
-ìkas), attracting the stress and rendering the resulting
forms immobile (oxytonic). When the accent was retracted
from non-barytonic fixed-stress forms (*-ikàs > -ìkas etc.),
it was also retracted in these illative and allative forms,
so we get -õn (< *-a:n-ná), -ósna (*-a:s-ná), -an~
(*-an-ná), -úosna (*-o:s-ná), -ópi (*-o:-pí), etc. It would
be confusing to call this retraction law Nieminen's law (in
its original formulation, Nieminen's law explains the
unrelated retraction of the accent from -ás in _mobile_
o-stem nom.sg. forms, e.g. obuolás > óbuolas, but not in
obuoly~s). Whatever we call it, its net effect was to
eliminate a.p. b in Lithuanian, leaving only (a) (1/2) and
(c) (3/4) (in non-compound nouns, the loss of the neuter
would already mostly have taken care of that, but in the
verb the effect was significant), and in compound words, to
eliminate the oxytone suffixes, leaving only suffix-stressed
compounds as a special category (e.g. -ìnis, -iny~kas,
-ó:tas in Table 85, p. 191 of Dybo's Accentology).
It seems that the retraction had the effect of turning an
acute receiving the stress into a circumflex (métatonie
douce), but not being an expert on Lithuanian, I may well be
overlooking something.
=======================
Miguel Carrasquer Vidal
mcv@...