Summarizing.
What I gather is that the developments that led to Slavic
accentuation are chronologically the following:
1) Reshuffling of PIE accentual paradigms:
- (some) thematic nouns and verbs acquire mobility
- mobility becomes polarized (barytone/oxytone)
For convenience's sake I'll call this whole ensemble of
analogical changes "Pedersen's law".
2) Hirt's law. A non-vocalic laryngeal in the first
syllable attracts the stress. Raises the number of
barytones.
[3) Winter's law. Causes acute tone, but does not (usually)
result in retraction of the stress, so must come after
Hirt's]
The above were Balto-Slavic. Now for the Slavic laws:
4) Dybo's law. Non-acute barytones become mesotonic
(AP(b)).
5) Meillet's law. Mobile paradigms lose their original
stress (acute, circumflex or short) in the non-oxytone forms
and become enclinomenic ["unstressed"] vs. oxytone.
6) Stang's law. Stress is retracted from weak yers and
(circumflex) long medial vowels, resulting in neo-acute
accent on the preceding syllable.
7) Neo-circumflex and other accentual changes are
post-Common Slavic.
About "Pedersen's law". This is perhaps the most mysterious
and underspecified one. First, it's not a phonetic law, but
an analogical one. Second, it's the most sensitive one to
everyone's opinions on the reconstructed PIE accent. The
main enigma is why the thematic nouns and verbs, which had
no mobility in PIE, acquired mobility in Balto-Slavic on the
model of the mobile athematics, when the number of such
mobile athematic nouns and verbs is so small in both Baltic
and Slavic. But perhaps that is precisely why. The old
athematic nouns and verbs almost all became thematized in
Balto-Slavic, but in doing so, they _retained_ their
athematic mobile stress, thus introducing mobility in what
used to be columnar paradigms.
In general the cases that became barytone in nominal
paradigms are the accusative, (vocative), locative + dative
(and ablative = o-stem genitive?) singular, and the
nominative and accusative plural/dual.
The generalization was presumably:
barytone <= case forms with barytone variants (in the PD and
AD paradigms):
acc.sg. *h2ák^monm.
loc.sg. *h2ák^moni
(abl.sg. *h2ák^monod)
nom.pl. *h2ákmones
acc.pl. *h2ákmonm.s
oxytone <= case forms with oxytone variants (in the HD and
AD paradigms):
nom.sg. *p&2té:r
gen.sg. *p&2terés
ins.sg. *p&2teréh1
gen.pl. *p&2tróm
dat.pl. *p&2tr.mós
loc.pl. *p&2tr.sú
ins.pl. *p&2tr.mí:s
An exception are the (masculine) mobile o-stems, where there
was a tendency to make the whole singular (except the
instrumental?) barytone and the whole plural (except the
accusative) oxytone.
Oxytone neuters remained oxytone (or rather mesotone), as is
the case of *pteróm (peró, ins.pl perê'xU, not *perêxÚ),
presumably because they lacked barytonized accusative forms.
The same goes for barytone neuters, which remained barytone
(until Dybo's law, when the non-acute ones became masculine
mesotones).
In the verb, mobility has been all but lost in Lithuanian
(where it only shows itself in the retraction of the stress
towards a preverb in prefixed forms of what used to be "AP
4" verbs). In Slavic, all that remains is the enclinomenic
1sg. form of thematic AP(c) verbs. Originally, the barytone
forms must have been those of the 1/2/3 sg. present
(*h1és-mi, *h1és-si, *h1és-ti), but the 2sg. ending was
replaced by a mysterious acute suffix *-eí, and the other
singular forms (except thematic *-o:) followed suit.
Oxytone (mesotone) thematic verbs, including those in -né-,
-yé- and iteratives in -éie- (and denominatives in -iyé-)
did not become mobile, but remained mesotonic (AP(b)).
=======================
Miguel Carrasquer Vidal
mcv@...