I still think the explanation of the accent of the mobile
3rd. person plural ending by Dybo's law (*-ónti > -oNtI'),
as suggested by Thomas Olander, is too good to not be true.
It creates a few problems, though, within the context of the
theory of Balto-Slavic accentuation as I had set it up.
Besides the stem-stressed type (a.p. a) and the mobile type
(a.p. c), I have suggested the existence of a theme-stressed
(a.p. b) type in Proto-Balto-Slavic. This applies to three
categories of words:
1) neuter o-stem oxytones (peró, vêdró, etc.; originally
also mêNso, jâje).
2) compound words with a stressed suffix (stress may be on
the suffix itself, e.g. *-áh2tos > -átU [the term
"theme-stress" is inappropriate here], or on the desinence,
e.g. *-ikós > *-IcI')
3) thematic theme-stressed verbs (-jé-, -né-, -í:-).
If Dybo's law can act on mobile forms with medial stress,
then it should also work on such a.p. b forms as existed
before the law came into being, and indeed in the case of
non-acute non-recessive nominal suffixes, it seems likely
that Dybo's law *did* work (even if in Dybo's S.A., which is
the only MAS source I have available at the moment, no
example is given of a non-acute suffix which has acquired
theme-stress through Dybo's law: the reduced-vowel suffixes
*-ikó-, *-ijó- and probably *-istwóm had theme-stress to
begin with, and the recessive suffix -ota is used as a
place-holder on p. 172 for the missing category of
non-recessive suffixes with a short vowel). In any case,
the distribution of stress in the non-recessive suffixes is
in accordance with the rule as it exists for non-mobile
uncompounded nouns: stress on the suffix if acute, otherwise
stress on the theme-vowel/desinence (in theory, the stress
of -a"t-, -i"n- and such could also be derived from formerly
theme-stressed forms (**-ah2tós, **-iHnós), with retraction
either through Hirt's law, or through "-Dybo").
In the case of categories (1) and (3), however, we have a
problem. If Dybo's law were allowed to work on them, we
would get:
NA pero'
G pera'
D peru'
L perê'
I pero``mI => peromI' (=>peròmI)
NA pera'
G perU'
D pero``mU => peromU' (=> peròmU)
L perê~xU => perêxU' (=> perê'xU)
I pery'
NA perê'
GL peru'
DI pero``ma => peroma'
1 pisjoN'
2 pisje``s^I => pisjes^I' (=> pisje`s^I)
3 pisje``tI => pisjetI' (=> pisje`tI)
1 pisje``mU => pisjemU' (=> pisje`mU)
2 pisje``te => pisjete'
3 pisjoN~tI => pisjoNtI' (=> pisjoN'tI)
But this is not in fact what we find at all. Instead of
end-stressed forms (as a result of Dybo's law), what we find
is the exact opposite: stress retraction (marked by #).
NA pero'
G pera'
D peru'
L perê'
I pero``mI (or peròmI?)
NA #pe`ra
G #pe`rU
D #pe`romU
L #pe`rê~xU
I #pe`ry
NA perê'
GL peru'
DI pero``ma
1 pisjoN'
2 #pi'sjes^I
3 #pi'sjetI
1 #pi'sjemU
2 #pi'sjete
3 #pi'sjoNtI
The stress retraction in the a.p. b verbs has been explained
in different ways: analogy after the i-stems (assuming
circumflex i~ and Stang's law), the result of van Wijk's law
and analogical extension of lengthened (circumflex) thematic
vowel e~, etc.
It's funny that retraction of the accent should occur in
both the o-stem neuters and the theme-stressed verbs, the
categories that should have retained theme-stress in PBS
after Pedersen's law. As we saw earlier, another strange
phenomenon is that the "-Dybo" retraction law also doesn't
work in the oxytone neuters (e.g. vêdró) [within the
traditional framework: "another strange phenomenon is that
Dybo's law works in oxytone neuters with acute root (e.g.
vêdró)"].
Now what if these retractions are related?
We can postulate a kind of "neo-mobility" that applied to
a.p. b stressed forms in Proto-Slavic, _before_ Dybo's law.
The essence of it would have been the creation of a mobile
paradigm with reversed polarity in relation to the old
mobile paradigm. In the verb:
a.p. a a.p. c a.p. b
bo``doN be``roN pisjoN'
bo``detI be``retI pisje``tI
bo``demU beremU' pisje``mu => pi'sjemU
bo``doNtI beroN~tI pisjoN~tI => pi'sjoNtI
In the noun:
a.p. a a.p. c a.p. b
dvo``ro ber~gU pero'
dvo``ra ber~ga pera'
dvo``ra bergi' pera' => pe`ra
dvo``rU bergU' perU' => pe`rU
A.p. c had (in thematic forms) root-stress in the singular,
end-stress in the plural (except the nominal acc. pl.). The
new mobiles of a.p. b had theme-stress in the singular,
retracted stress in the plural.
Neither +Dybo nor -Dybo affected the forms of a.p. b with
(neo-acute) retracted stress, so in the peró/vêdró words,
nothing changes. The old non-acute a.p. a neuters do not
merge with the "neo-mobiles", but become a.p. b masculines
(dvorU', dvora' / dvory' [dvora'], dvorU').
In the verb, Dybo's law *did* affect the theme-stressed
forms of the singular, as well as the 3pl. of mobile forms,
so the paradigms became:
a.p. c a.p. b (old) a.p. b (new)
'beroN pisjoN' bodoN'
'beretI pisjetI' bode`tI
beremU' pi'sjemU bode`mU
beroNtI' pi'sjoNtI bodoN'tI
To get from here to the CS forms a.p. c 2/3 sg. bere`s^I,
bere`tI, and a.p. b pi'sjes^I, pi'sjetI requires an
analogical development. In both cases, the 2/3 sg. acquired
the accentuation of the plural (and of the *other*
paradigm). The old barytonic verbs affected by Dybo's law
meanwhile became mobile (#'bodoN, bode`s^I, bode`tI,
bode`mU, #bodete', bodoN'tI), which may also have been a
factor.
=======================
Miguel Carrasquer Vidal
mcv@...