I'll try to recapitulate my thoughts.
Balto-Slavic accentuation continues a late PIE simplified
accentual system, which can be reconstructed as follows:
Nouns: C-stems are either static (barytone) or laterally
mobile. The laterally mobile paradigm represents the merger
of the three PIE mobile nominal accent paradigms
(proterodynamic, hysterodynamic and amphidynamic), where the
acc., loc.+dat. sg., the nom. and acc. pl. and the nom./acc.
du. are barytone, the other cases are oxytone (except the
nom. sg., which can be barytone or oxytone, depending on the
original type: *nébhos, *dhughté~). All "vowel" stems (-o-,
-a:-, -i-, -u-) have acquired columnal stress, and can be
either barytone or non-barytone.
Verbs (present system): C-stems are either acrostatic
(barytone) or mobile (originally the sg. was barytone, the
pl. and du. were oxytone). Thematic verbs are barytone
(*bhéreti) or non-barytone (*tudéti, and verbs in *-néti,
*-yéti, *-sk^éti, *-éyeti, *-iyéti).
There are therefore basically three accent paradigms:
I. fixed barytone
II. fixed non-barytone
III. laterally mobile
Like Greek, Armenian and Tocharian, Balto-Slavic is affected
by the Francis-Normier law, which vocalizes the laryngeals
h2/h3 after the resonants y/w (i/u). This explains why byti
(*bhuh2-), z^iti (*gWih3-), piti (*pih3-) etc. were not
affected by Hirt's law.
Hirt's law: non-vocalized laryngeals attract the stress from
a following syllable. This affects the sequences *eh1, *eh2,
*eh3, *oH, *r.H, *l.H, *m.H, *n.H, *ih1, *uh1, but not
sequences like *e/or&, *e/ol&, *e/oN& or *i&, *u&.
The effect of Hirt's law is to increase the number of
barytones at the expense of non-barytones and mobiles.
Hirt's law is (sometimes?) blocked by oxytonic (a.p. II)
suffixes, such as *-tlóm (e.g. stah2-tlóm > sta:tlá(m)) or
*-kó- (e.g. *-ah2kós > *-a:kás).
Winter's law lengthens (with acute intonation) short vowels
before PIE (*b), *d, *g, *g^, *gW, (*z). Diphthongs are
also acutized under the same circumstances. This reinforces
the opposition between acute and circumflex vowels and
syllables in Balto-Slavic. Short vowels (a, e, i, u) have
falling intonation, as do resonants in syllable final
position (-m, -n, -l, -r, -i, -u). Length (:) has rising
intonation, whether original or due to a laryngeal (h, &) or
to Winter's law (?d etc.), but superlong vowels (~) have
falling intonation. E.g. the o-stem nom. sg. -a\s is short
falling (-às), the acc.sg. -a\m\ has a short vowel, but the
syllable is circumflex (-am~). The dat. [PIE -o\e\i\ >]
-o\:/i\ is circumflex (-õi), as is the abl. -a\a\(d) > -a\~\
(-ã, uncontracted -a(h)à > Slav. -ogo). In the ah2-stems,
we have e.g. nom. sg. -a\h2/ (acute), acc.sg. -a\h2/m\
(acute vowel, circumflex syllable), acc.pl. -a\h2/ms (acute,
as the /m/ is not syllable final), gen.sg. -a\h2/o\s = -ãs
(circumflex). A special case are the sequences -a\i\(h1)
(circumflex -ai~ > Slav. -ê: o-stem loc.sg.; NA n. du.) vs.
-a\j(h1/) (o-stem nom.pl.; optative suffix. Notated as -aí >
Slav. -i, but not really acute [doesn't trigger Saussure's
law in Lith.])
Pedersen's law is the analogical extension of lateral
mobility to, initially, the accusatives of vowel stems. As a
result, masculine and feminine non-barytones become mobile
(as well as adjectives). Only the neuter o-stems nouns
retain fixed non-barytone stress.
In the verb, the only similar development (which needn't be
simultanous with Pedersen's law) is that some (most?)
causatives (e.g. *poih3-éje-ti > *paji~ti) become mobile for
some reason, while iteratives (e.g. *wod-éje-ti > *wadi~ti)
remain fixed non-barytone, as do the other types of verbs
mentioned above (*-né-, *-(i)jé-, *-sk^é- and denominatives
in *-ijé > -í:- [acute]).
In Lithuanian, the three accent paradigms are reduced to two
(barytonic vs. mobile) by metatonic processes such as
*de:tlóm -> de~:klas, and are subsequently split into two
each (a.p. 1/2, 3/4) by Saussure's law.
The Slavic accent laws are:
- Meillet's law. This deletes most prosodic features from
mobile paradigms. The barytone forms become enclitic (ná
golvoN) or proclitic (z^ivU-jÍ) if possible (otherwise they
have non-acute stress on the first syllable: gol~voN,
z^i~vU), the oxytone forms also lose all trace of acute
intonation in the pretonic syllables.
- Dybo's law. Non-acute barytones shift the ictus one
syllable to the right (a.p. b). In nominal forms, this
should have resulted in a merger of non-acute barytones with
old a.p. II words (in practice, only PIE oxytone o-stem
neuters). However, masculine o-stems affected by Dybo's law
become mobile (a.p. c, c.q. traces of a.p. "d") [this also
happens to a large extent in the i-stems], while the gap is
filled with neuter o-stems affected by Dybo's law, which
become a.p. b _masculines_ (Illich-Svitych's law). (The
merger of acute barytone (a.p. a) u-stems with mobile (a.p.
c) u-stems may be a later development).
In the verb, non-acute barytones also become a.p. (b),
except that old (non-acute) o-grade barytones ending in an
obstruent (bod-tei > bosti, etc.) become mobile (a.p. c).
O-grade verbs ending in a resonant (borjoN, koljoN, poljoN,
porjoN, meljoN) do not become mobile, but shift to jé-stems
(old a.p. II -> a.p. b verbs are typically jé-stems).
The existence of a.p. II = a.p. b verbs (-né-, -jé-, -sk^é-
verbs, iteratives and denominatives in -i~- and -í:-) and
nouns (o-stem neuters) _before_ Dybo's law is something of a
novelty, as far as I know. It raises the possibility of a
parallel "Obyd's law" (Dybo's law in reverse), retracting
the accent to the left in non-acute a.p. II verbs (this
certainly didn't happen in the o-stem neuter nouns, which
remain oxytone even if the root is acute: vêdró, stadló,
dêdló, etc.). The inverse Dybo's law certainly seems to
have worked in the infinitive (oxytone, therefore comparable
to a.p. II/b), where the stress is retracted in most (all?)
acute-root verbs (whether their present is a.p. a, b or c).
In any case, this needs further investigation. By another
retraction law, a long circumflex vowel in pretonic position
also causes retraction of the stress in oxytone forms, such
as for instance in the infinitive of i-verbs (*vodi~tí =>
vodi''ti). In this case, the retraction also applies to the
o-stem neuters (*me~msá(m) => *méNso, *o:ujá(m) => *jáje,
and subsequent merger with the mobile class (meN~so, ja~je).
- Stang's law. The accent is retracted from weak (usually
final, a.p. c) yers, and from medial long circumflex
(non-acute) vowels (e.g. iterative -i~-, lengthened thematic
vowel -é~- [what caused this?]). The syllable receiving the
ictus gets neo-acute intonation. An acute long vowel does
not cause the effect (denominative -í:-, a.p. "b2").
- Neo-circumflex-law. An acute changes into circumflex if
the vowel in the following syllable is long.
I'm sure I've forgotten something(s).
=======================
Miguel Carrasquer Vidal
mcv@...