Vowel-Zero Alternations in Albanian and
Morphophonological ContactAndrew Dombrowski
Introduction
• Slavic + Geg Albanian both have vowel-zero alternations in inflection, due to independent processes of syncope.
• Some Geg dialects in contact with Slavic extend vowel-zero alternations to include nouns ending in –ull, -ur, -urr.
• In some instances, the alternating vowel in Geg is shifted to match corresponding Slavic jer reflex.
Introduction
• Goals of this paper:– argue that the extension of vowel-zero
alternations in Geg is due to Slavic influence– demonstrate that this cannot be accounted for
in terms of direct Slavic > Albanian grammatical transfer
– explore ramifications of this for modeling phonological contact
Introduction
• Outline:– Vowel-zero alternations in Geg– Vowel-zero alternations in Slavic– Extension of alternations in Geg– Analysis– Repercussions
Vowel-zero alternations in Geg
• /ə/ > Ø except when conflicts with phonotactics– note: schwa is always unstressed
• Can be accounted for phonologically
• Sample and sketch account taken from Luznia e Dibrës, a central Geg dialect near Debar along Albania-Macedonia border
Vowel-zero alternations in Geg
• Schwa deleted in Luznia e Dibrës• See handout; key examples below
Luznia e Dibrës
Original Gloss
prrallz përrallës fairy tale-gen.sg.def.
kpuc këpucë shoe-nom.sg.indef.
shnre *shëndre December-nom.sg.indef.
Vowel-zero alternations in Geg• Schwa preservation in Luznia e Dibrës• See handout; key examples below
Luznia e Dibrës Original Gloss
e kërmashme e kërmashme red and white (of sheep)-fem.sg.indef.
përjashta përjashta outdoors
i vokël i vogël small-masc.indef.
pullën pullën button-acc.def.
Vowel-zero alternations in Geg
• Descriptive generalizations– Complex onsets are tolerated except for
CRCV syllables; CNCV is permitted.– Rising sonority codas are not permitted.– Codas of two sonorants are not permitted.
• Sketch OT account– Constraints: Sonority, OCP-son, *CrC, *ə– See handout for details
Vowel-zero alternations in Geg
• Sketch OT account is not complete– Luznia e Dibrës dialect description does not
have a complete lexicon; above account is consistent with the lexicon given.
• Vowel-zero alternations in Luznia e Dibrës can be captured straightforwardly in an OT model.– With the exception of morphemes like për, the
OT model is agnostic as to whether schwa is present in the UR.
Vowel-zero alternations in Slavic
• Slavic vowel-zero alternations are older and much more complicated than Geg.
• See handout for outline of standard Macedonian vowel-zero alternations.
• Fairly representative of Slavic dialects with which Geg is in contact.
• Much lexical variation.
Vowel-zero alternations in Slavic
• Key examples from standard Macedonian:• Adjectives in –en:
– gladen ‘hungry’ ~ gladniot ~ gladna– zelen ‘green’ ~ zeleniot ~ zelena
• Nouns in -ok:– dobitok ‘livestock’ ~ dobici– početok ‘start’ ~ početoci
• Nouns in –ol:– jazol ‘knot’ ~ jazli– sokol ‘falcon’ ~ sokoli / sokli
Vowel-zero alternations in Slavic• Analysis of Geg does not extend.• Several possible approaches (cf. study of vowel-
zero alternations in Russian):– Abstract jer vowels with rules for deletion (Lightner
1965, Rubach 1986); requires lexical specifcation– Government Phonology ‘translation’ of this (Scheer
2005)– Treat as synchronic vowel insertion with
morphological conditioning of resulting alternations (Darden 1989)
– Treat jer vowels as morphological constituents (Chew 2000)
Vowel-zero alternations in Slavic
• Cannot be treated in terms of ‘pure’ phonology
• Reference must be made to the lexicon– Classical generative approach involves lexical
specification (/dobit+ъk-ъ/ vs. /počet+ok-ъ/; /jazъl-ъ/ vs. /sokol-ъ/)
– Alternative approaches involve morphological specification
Extension of vowel-zero alternations in Geg
• Extension to nouns ending in (idiosyncratically) unstressed –ull, -ur, -urr
• Patterns of behavior:(1) Preservation without alternation
(2) /u/ > /ə/; introduction of alternation in paradigms
(3) Preservation of /u/, introduction of alternation in paradigms
Extension of vowel-zero alternations in Geg
• (1) - /u/ preserved, no alternations– Plava and Gucia in Montenegro, Kastrati, Hoti,
Kelmendi, Peshteri in the Sandžak region of southern Serbia, and Reç-e-Dardhës e Dibrës near Debar.
– Data from Kastrati dialect
Nom.sg.indef. Nom.sg.def.
vetull ‘eyebrow’ vetulla
kumull ‘plum’ kumulla
hekur ‘iron’ hekuri
Extension of vowel-zero alternations in Geg
• (2) - /u/ > /ə/; introduction of alternation in paradigms– Hasi, Qyteza e Kaçanikut, Shala e Bajgorës,
Gjakova, Tuhini i Kërçovës, Morava e Epërme, Vila-e-Kalisit të Lumës.
– Data from Hasi dialect
Nom.sg.indef. Nom.pl.def.
vetëll ‘eyebrow’ vetlla
kumëll ‘plum’ kumlla
hekër ‘iron’ Hekra
Extension of vowel-zero alternations in Geg
• (3) - /u/ preserved, introduction of alternation in paradigms– Mirdita, in Gryka e Madhe e Dibrës, Ana e Malit,
the Debar city dialect, Luznia e Dibrës, Karadak, and Puka
– Data from Puka dialect
Nom.sg.indef. Nom.sg.def.
vetull ‘eyebrow’ vetlla
kumull ‘plum’ kumlla
hekur ‘iron’ hekri
Extension of vowel-zero alternations in Geg
• Fourth pattern: in Opoja, /u/ > /o/ in these nouns, mirroring jer reflexes in neighboring Gora.
Nom.sg.indef. Nom.pl.def.
vetoll ‘eyebrow’ vetlla
kumoll ‘plum’ kumlla
grumoll ‘pile’ grumlla
Opoja• Actually, in Opoja, [ə] > [o]…• Nominal declension:
Opoja Standard Albanian
Indef Def. Indef. Def.
Nom (ni) motor motra (një) motër motra
Acc (ni) motor motron (një) motër motrën
Gen (i, e ni) motros
(i, e) motros (i, e një) motre
(i, e) motrës
Dat (ni) motros motros (një) motre motrës
Alb (pi(j) ni) motros
(pi(j)) motros
(prej një) motre
(prej) motrës
Opoja
• Adjectival declension compared to general Geg:
Opoja General Geg
Masc.Sg. Fem.Sg. Masc.Sg. Fem.Sg.
‘clothed’ i veshom e veshme i veshun e veshun
‘dried’ i terom e terme i terun e terun
‘slow’ i kadalshom e kadalshme i ngadalshëm e ngadalshme
Opoja
• Changes in adjectival declension compared to other Geg dialects:– (1) /o/ corresponding to /u/
• Possible intermediate stage: /u/ > /ə/ > /o/, but this implies intermediate forms like *i terën, which are not attested
– (2) generalization of feminine ending –e
• Result: similar to template in Macedonian
Opoja
• Adjectival declension in Opoja compared to Macedonian
Masc.(Indef).
Sg.
Fem.(Indef).
Sg.
Opoja Macedonian Opoja Macedonian
Class A i vesh-om
‘clothed’
slad-ok
‘sweet’
e vesh-me
‘clothed’
slat-ka
‘sweet’
i kadal-sh-om
‘slow’
mrt-ov
‘dead’
e kadal-sh-me
‘slow’
mrt-va
‘dead’
Class B i ble-m
‘bought’
rod-en
‘born’
e ble-m-e
‘bought’
rod-en-a
‘born’
i shti-m
‘added’
zelen
‘green’
e shti-m-e
‘added’
zelen-a
‘green’
Analysis
• Degree of isomorphism between Opoja and neighboring Slavic strongly suggests contact-driven explanation
• On u > ə dialects:• All in Kosovo or vicinity (Hasi is between Kukës
and Kosovo; Vila-e-Kalisit të Lumës is in vicinity of Kukës, but economic ties have historically been with Kosovo)
• This correlates strongly with Slavic dialects where ъ, ь > ə, suggesting that this pattern is structurally very similar to Opoja
Analysis
• On dialects with preserved /u/ and innovated alternations:– Geographical position: on periphery of /u/ >
/ə/ zones, ranging from Montenegro in the NW (Ana e Malit) to Debar in the south to Karadaku in the E.
– Suggests that this is not under Slavic influence, but instead is diffusion within Albanian
Analysis
• Stages:– (1) Albanian dialects in and around southern
Kosovo shift /u/ in endings –ull, -ur, -urr to ə under influence from neighboring Prizren-Timok dialects of Serbian where jers > ə.
– (2) Opoja developments (can be seen as subset of stage (1) with subsequent shift due to neighboring Gora, except for participles).
– (3) Spread of vowel-zero alternations to neighboring dialects without /u/ > /ə/ shift
Analysis
• Things to account for…– (1) equation of (one) Slavic alternating vowel
with Albanian alternating vowel.• Opoja is clearest example of this as an overt
change, but is arguably implicit in u > ə dialects.
– (2) extension of alternations to nouns ending in –ull, -ur, -urr.
– (3) subsequent spread of alternations in neighboring Albanian dialects without u > ə shift
Analysis
• Can (1) and (2) be analyzed as direct borrowing of Slavic grammar by Albanian?– (1): probably not. If Slavic alternating vowels
are underlying, specification of quality is nowhere in the grammar.
– (2): also probably not. Slavic vowel-zero alternations involve lexical specification, and the relevant lexemes + morphemes are not borrowed.
Analysis
• Suggestion:– Some reorganization seems to be happening at an
intermediate interface stage between the two languages
• An interlanguage? Similar on first glance, but an interlanguage analysis might make overly strong claims re: sociolinguistic particulars. Also, this would only account for reanalysis of Slavic, not its impact in Albanian.
– Interface-based approach might be an interesting prism to look at questions structural compatibility in borrowing.
Analysis
• Sample implementation 1: the Opoja shift (ə > o)stage A: <o~Ø>[+Slavic], <ə~Ø>[+Albanian]change: <o~Ø>[+Slavic] [+Albanian]stage B: <o~Ø>[+Slavic, +Albanian]
• Elements in stage A reflect generalizations made by speakers of Albanian, and elements in [brackets] are metadata.
• Key point: a generalization <G>[+Slavic] does not have to actually be completely true of Slavic; it should be deducible from the Slavic evidence but can be a reanalysis.
Analysis
• Sample implementation 2: spread of alternations without u > ə between dialects D1 and D2– possibility (a): reanalysis of D1
stage A: <[-ull, -ur, -urr][-alternations]>[+D1] <[-ull, -ur, -urr][+alternations]>[+D2]
change: <[-ull, -ur, -urr][+alternations]>[+D2] [+D1] stage B: <[-ull, -ur, -urr][+alternations]>[+D1, +D2]
– In this analysis, D1 speakers reanalyze D2 such that the only salient feature of D2 is the presence of alternations in the marked nouns.
Analysis
• Sample implementation 2: spread of alternations without u > ə between dialects D1 and D2– possibility (b): partial implementation
stage A: <[-ull, -ur, -urr][-alternations]>[+D1] <[-ull, -ur, -urr][+alternations]>[+D2]
<[-ull, -ur, -urr][+ u > ə]>[+D2]change: <[-ull, -ur, -urr][+alternations]>[+D2] [+D1]stage B: <[-ull, -ur, -urr][+alternations]>[+D1, +D2]
<[-ull, -ur, -urr][+ u > ə]>[+D2]
– D1 only partially reassign tags from D2
Analysis
• The distinction made in sample implementation 2 between reanalysis and partial implementation of shift might be useful in other instances.
• How to characterize the mechanism of tag reassignment, and what constraints might be involved?
• Can the concept of grammatical interface be productively applied to other situations?