Something occurred to me the other day, and I thought I'd
run it by you guys, just in case I'm overlooking something
important.
So far, my views on (Balto-)Slavic accentology can be
summarized as follows:
PIE had thematic (vowel-stems) and athematic
(consonant-stems) (pro)nouns and verbs. The athematic
categories had an inherited system of acrostatic versus
mobile accent (actually, different kinds of mobile
paradigms, but they had all merged in Balto-Slavic into a
laterally mobile [amphidynamic] pattern). The thematic
classes had an inherited distinction between root-stressed
(barytone) and theme-stressed (oxytone) paradigms.
In Balto-Slavic, mobility was transferred from the consonant
stems to the vowel stems. This is Pedersen's law. Oxytone
vowel-stem nouns retracted the accent in the accusative (as
well as in the oblique forms with root-stress in the
corresponding athematic words). This did not apply to
oxytone thematic neuters (which have no accusative).
Barytone thematic verbs advanced the stress in the dual and
plural, on the model of the mobile athematic verbs. This
gave three accent paradigms in Proto-Balto-Slavic:
nouns:
I static ath. + barytone them.
II oxytone them. (neuters)
III mobile ath. + oxytone them. (non-neuter)
verbs:
I static ath.
II oxytone them.
III mobile ath. + barytone them.
The system underwent at least two changes within the
Balto-Slavic period:
1) Hirt's law: a non-vocalized postvocalic laryngeal
attracts the stress from a following stressed syllable.
This affected nouns and verbs of accent class II, nouns of
accent class III, and _athematic_ verbs of class III. (Not
thematic verbs, because they were already root-stressed in
the singular, and in the plural/dual there was always a
syllable (to wit, the thematic vowel) between the laryngeal
in the root syllable and the stressed ending).
Hirt's law fails if the laryngeal was vocalized, as in the
sequences VrH, VlH, VNH, VwH [but not VyH?], and possibly
iX, uX (if the laryngeal was h2 or h3: Francis/Normier's
law. That would explain the accentuation Slav. bylá, z^ilá;
Latv. bût, dzît, etc.)
2) Winter's law: vowels and diphthongs become long/acute
before PIE *b, *d, *g, *g^, *gW [_not_ *bh, *dh, *gh, *g^h,
*gWh]. This did not immediately affect the accent, but it
was to have consequences at a later stage, both in Baltic
and in Slavic, when the accent was rearranged according to
the presence or absence of acute syllables in the word. In
Shintani's reformulation of Winter's law, which seems to be
correct, the lengthening only takes place in syllables
immediately preceding the stress (accent classes II (always
thematic) and _athematic_ III).
In Slavic, as far as I had worked it out, the system was
affected by three more accentological laws:
1) Meillet's law (affects accent class III): non-finally
stressed forms in mobile paradigms become stress-less
(enclinomenic) if possible [i.e. if there is a preposition
or other element which they can be enclitics of].
Otherwise, when independent, long syllables get "circumflex"
(long falling) intonation [â] (even if historically acute),
short syllables get short falling stress [o\\]. The
resulting mobile paradigm, with initial ^ ~ \\ ~ (no stress)
alternating with final stress becomes Slavic a.p. c.
2) Dybo's law (affects accent class I): if the stressed
syllable is not acute, the stress shifts to the following
syllable. These forms make up Slavic a.p. b. The unshifted
forms with acute root syllable make up Slavic a.p. a.
3) The "jabloko law" (affects accent class II): if there is
an acute syllable earlier in the word, stress is retracted
to that syllable (âblUkó > a"blUko). The resulting forms
become part of Slavic a.p. a. Unshifted forms become part
of Slavic a.p. b.
There are a few other laws, affecting mainly the output of
Dybo's law, e.g. Illich-Svitych's law: neuter barytones
affected by Dybo's law become a.p. b masculines, while
masculine barytones affected by Dybo's law become a.p. _c_
masculines, with traces of end-stress/theme-stress in
certain dialects (so-called "a.p. d"). A similar phenomenon
occurs in verbs: barytone verbs affected by Dybo's law
become a.p. c if they have a root ending in a stop (bo\\doN,
bodètI, bostí), but they become a.p. b je-verbs if ending in
a vowel or resonant (borjóN, bòrjetI, bo"rti).
This is more or less what I had up til now. There was only
one major law which worried me, didn't really fit into the
scheme layed out above, and which is therefore not mentioned
in the above summary: Stang's law.
I think I have now found a way to integrate Stang's law into
the above account in a natural way.
In the first place, I need to split Stang's law up into two:
(1) weak yers pass the stress onto a preceding syllable if
there is one, otherwise to a next syllable.
(2) the problematical part of Stang's law (Stang's law
proper).
The major aspect in which my view of Slavic accentology
deviates from that of the "Moscow school" is that I posit
_three_, not two, Balto-Slavic accent paradigms. The
solution to the problem of Stang's law, now that I've
realized what it is, is dead simple: it follows naturally
from the existence of _three_ accent paradigms in
Balto-Slavic. We can formulate the law as follows:
Stang's law: the accent is retracted from a stressed medial
syllable.
The law works _before_ Dybo's law and the "jabloko" law, at
a time when there were still three accent classes: I
(root-stressed), II (theme-stressed) and III (mobile). I
cannot tell whether it was before or after Meillet's law.
The law affects only words in accent class II, in other
words:
(1) neuter o-stems of accent class II (the "peró-group")
(2) jé-, né- and í:-verbs (iteratives) of accent class II.
There were no other forms with medial accent at the time.
Working this out:
before after (analogy:)
NAV peró
G perá
D perú
L perê'
I *perá
(perómI péromI) perómI ~ peromÍ
NAV perá (péra)
G perÚ pérU
D perómU péromU
L perê'xU pérêxU
I perý péry
NAV perê'
GL perú péru
DI peróma péroma
In the NAV plural, as noted by Illich-Svitych, root-stress
may have been in part already inherited from PIE.
In any case, after Stang's law, the pattern develops into a
neo-mobile paradigm, which [except possibly for the NAV pl]
has end-stress where a.p. c has begin-stress (or no stress),
and begin-stress where a.p. c has end-stress.
Verbs:
1 steljóN steljóN
2 steljés^I stéljes^I
3 steljétI stéljetI
1 steljémU stéljemU
2 steljéte stéljete
3 steljóNtI stéljoNtI
1 steljévê stéljevê
2 steljéta stéljeta
3 steljéte stéljete
After Stang's law, it's the a.p. c verbs which are
analogically transformed to become the mirror image of the
neo-mobile class II paradigm:
1 béroN
2 béres^I > beres^Í
3 béretI > beretÍ
1 beremU'
2 bereté
3 beroNtÍ
1 berevê'
2 beretá
3 bereté
It's possible that something like Stang's law also worked in
Lithuanian, but without the restriction that the stress was
not retracted from final syllables (i.e. in Lithuanian,
accent class II merged with accent class I).
=======================
Miguel Carrasquer Vidal
mcv@...