The synchronic accent rules for Slavic compound words (root
+ suffix) are relatively simple. Some suffixes have no
influence on the stress of the compound ("Class I suffixes"
in Dybo '81; (-) suffixes in Zaliznjak). When combined with
a.p. a, b and c roots (+, -> and -), we get:
(a) + - => /--
(b) -> - => -/-
(c) - - => /-- ~ --/
Examples:
a b c
(yer) bra"t-IskU z^en-I'skU moN~z^-IskU, moNz^-Iska'
(short) ma"k-ovU dvor-o'vU do``m-ovU, dom-ova'
(circ.) i"st-i:na c^Irn-i:'na poln-i:na', po~ln-i:noN
(acute) ko"rst-avU loNk-a"vU krU''v-avU, krUv-ava'
Theoretically, the last form (krUvavá) should have been
subject to Hirt's law (the ictus was on the syllable right
after the laryngeal), but this didn't happen. Mobility was
restored or maintained.
Dybo's second class of suffixes are suffixes which have a
stress of their own, either on the suffix itself (e.g.
-á~rjI, later -a:rjI' [apparently a borrowing from Gothic])
[Zaliznjak's (+)-suffixes], or on the ending (e.g. *-ikós >
-IcI') [Zaliznjak's (->)-suffixes]. This distribution was
inherited from PIE, and was slightly rearranged due to
Hirt's law (*-iHkáh2 > -í:ka: > -íca). If we take as
examples the suffixes *-ikós (short vowel), *-i~nos
(circumflex vowel) and *-í:ka: (acute vowel), the Common
Slavic forms before Meillet/+Dybo/-Dybo would have been:
*bra:t-ikúh *dvar-ikúh *vaz-ikúh
*ba:b-i~nuh *gen-i~nuh *dukter-i~nU
*sta:r-í:ka: *kirn-í:ka: *mald-í:ka:
Note that Hirt's law cannot have worked in the a.p. a words
(with the possible exception of *bá:bi:nuh), because the
ictus was nowhere placed immediately following the syllable
containing the laryngeal. Note also that there is no
mobility: the stressed suffixes fix the stress on themselves
or on the desinence.
Dybo's law pulls the accent forward, making all non-acute
suffixes oxytonic (*ba:b-i:núh(?), *gen-i:núh,
*dukter-i:núh). This justifies grouping the + and ->
suffixes together.
The "-Dybo" retraction law works in the following cases
(a.p. a roots):
*bra:tikúh > *brá:tikuh
*ba:bi:núh > *bá:binuh [or already by Hirt's?]
*sta:rí:ka: > *stá:ri:ka:
The result is a system with only two accent paradigms:
bra"tIcI // dvorIcI' vozIcI'
ba"bi:na // z^eni:nU' dUkteri:nU'
sta"rica // c^Irni"ca moldi"ca
We can conclude that a.p. b and c roots initially behaved
identically when combined with stressed suffixes. Later,
there was a tendency for the pattern of compound words with
unstressed suffixes to be carried over to those with
stressed suffixes (especially when such an unstressed suffix
was itself involved in a compound suffix together with a
stressed one, e.g. vol-óv-ina, etc.), leading to the
synchronic rule sketched above, where a -> root (a.p. b)
always causes the stress to be on the suffix no matter what
type it belongs to (except when the suffix has type [-/+
Min], in Zaliznjak's terminology, which are precisely the
cases in which the original distribution was maintained).
The rules also work for synchronically intransparent
compounds, such as the word for "apple" (j)a"blUko (a.p. a).
The root is itself mobile, coming from a PIE athematic
mobile paradigm *h2ábo:l(s), *h2(a)búlos. In Balto-Slavic
that became *abó:l(s), *ábolim, *abulés (with lateral
mobility), thematized as Lith. óbuolas, Latv. âbols, with
Winter's lengthening (in Shintani's formulation) originating
in the end-stressed nominative (as also shown by the Latvian
Brechton). Hirt's law of course never applied, and Winter's
law applied because the word was _not_ (exclusively)
barytonic. In Slavic the stressed suffix *-kóm was added
(to the oblique?), and the word became immobile (*a:bul-ká
-> *a:blUkó). Retraction of the ictus to the acute root
syllable by "-Dybo" finally resulted in Slavic ja"blUko.
=======================
Miguel Carrasquer Vidal
mcv@...