Gilles Authier .1Introduction
The Great Caucasus range is famous for its linguistic diversity, and perhaps most deservedly so in its Eastern corner, a region covering Southern Daghestan and the North of Azerbaijan. Three linguistic families meet there: East-Caucasian, with at least thirty languages; Turkic, represented by Azerbaijani and Kumyk; and Indo-European, with Tati, (like Persian) a South-West-Iranian Indo-European language. This diversity makes the region an open laboratory for the study of language-contact: genetically related languages will diverge according to which unrelated languages they are in contact with, while genetically unrelated languages will, after some period of bilingualism, show convergence phenomena in their typological features. It is well-known, for instance, that Azerbaijani, although very close genetically to Turkey Turkish, shows features in its syntax which can be ascribed to the influence of Persian as a prestige language in the area. Conversely, although Tati dialects share the overwhelming majority of their lexical items with Persian (more than 80 % of Tati words have Persian cognates), they show a considerable number of idiosyncratic typological features in their syntax, which, along with a new phonological system and a very different TAM system, set them apart as a separate language, comprizing several “dialects” that are unintelligible to speakers of Persian.
Most Tati varieties are poorly documented, and no comprehensive overall description of the language is available. Northern varieties are mostly spoken by Jewish communities in Southern Daghestan and cities of the Northern Caucasus, while the Southern varieties are spoken in the Republic of Azerbaijan, mostly by Muslim communities. Tati has thus been in intimate contact for a long time with Azerbaijani, and also with various Caucasian languages. According to Grjunberg (1963) Tati is a string of dialects extending from the Apsheron peninsula as far as the town of Nalchik in Kabardino-Balkharia. But in fact, the easternmost varieties spoken by Jewish communities in Groznyj, Nalchik and Mozdok appear to be part of the Kaitag dialect, spoken just North of Derbent.2 In the Old town of Derbent itself, the Tati speaking population is Jewish, while Tati speaking villages surrounding Derbent, like Djalgan of Nügedi, are Muslim today. In Azerbaijan, Tati is the language of many thousands of Muslim inhabitants of the Quba, Ismailli, Apsheron districts and other adjacent regions. Jewish Tati-speaking communities are flourishing in Quba, but moribund in Oghuz (formerly Vartashen). While variation between adjacent dialects may allow intercomprehension, Tat speakers from the Apsheron peninsula near Baku (all Muslim) cannot understand even the written version of a text in the ‘dialect’ spoken by the Jewish community of Derbent in Daghestan. Northern (Derbent Jewish) Tati is rather well documented3 (though there is still no comprehensive grammar of the language), but other varieties like the one here called “Apsheron Tati”, which is spoken near Baku, are much more in need of investigation. This paper is intended as a contribution to the description of this dialect.
As an instance of relevant discrepancies, the type of head-internal relative clauses under analysis in this paper seems to be a feature specific to this Apsheron dialect, which was not recognized as a valid strategy by Tati speakers in Daghestan. I have not yet been able to test other Southern dialects. Relative clauses in Tati have not been studied in the scanty literature available on this language. This paper is thus the first publication concerned with this topic, dealing with only one strategy, the head-internal one, which is found only in varieties of Azerbaijan.
Three strategies for the formation of relative clauses can be identified across Tati dialects:
-
the ‘pluralised-finite’ form strategy is found in at least two independant dialects, Lahiç and Derbent4;
-
the participial strategy, found in all varieties, has been inherited from middle Iranian, but has witnessed a remarkable extension due to its prevalence in surrounding Turkic and Caucasian languages;
-
the head-internal strategy, found, as far as I know, only in the Apsheron dialect, constitutes the topic of this paper.
There is no evidence of a Persian-type relativization strategy (i.e. a finite postposed relative clause, cf. Section 2.2) in any dialect of Tati, nor in Azerbaijani. In the latter language as well as in Apsheron Tati, relative clauses are of two types, with some complementarity and overlap: ‘participial’ vs. ‘head-internal’ (or ‘correlative’). In the following, we will describe these types and examine the criteria for their distribution
The paper is structured as follows: Section 2 provides an overview of the participle strategy in Azerbaijani and Apsheron Tati. Section 3 (‘Non-participial relative clauses’) presents the head-internal strategy in Azerbaijani and Apsheron Tati. Section 4 summarizes the results.
.2Participial relative clauses
The participle strategy is dominant5 in the great majority of Turkic languages. Standard Azerbaijani, like Turkish, makes use of mainly two types of participles to relativize either subjects or other functions: (i) the –(y)An participle (glossed ‘subjpart’) is used to relativize subjects with no past-time reference; (ii) the participle in –mIş (glossed ‘perfpart’) also has S/A (subject) orientation like the one in –(y)An, but it is used with past time reference (cf. [1]).
(1)
|
Azerbaijani
|
|
sən-in
|
qardaş-ın-ı
|
dö-yən /
|
döy-müş
|
|
2-gen
|
brother-pos2-acc
|
beat-subjpart
|
beat-perfpart
|
|
|
|
|
|
|
adam
|
bu
|
kənd-dən
|
deyil.
|
|
person
|
this
|
village-abl
|
negcop3
|
|
|
|
|
|
|
‘The person who beats / has beaten up your brother is not from this village.’ (fn)
|
In order to relativise other positions than subject in a participial clause, Azerbaijani, like Turkish, mostly makes use of the DIQ-participle (cf. [2]–[4]).
(2)
|
sən-in
|
qardaş-ın-ın
|
döy-düy-ü
|
adam
|
bu
|
deyil.
|
|
2-gen
|
brother-pos2-gen
|
beat-nonsubjpart-pos3
|
person
|
this
|
negcop3
|
|
‘The person your brother has beaten up is not this one.’ (fn)
|
|
|
(3)
|
mən-i
|
istə-yən
|
oğlan
|
qoy-duğ-um
|
şərt-ə
|
|
1-acc
|
want-subjpart
|
boy
|
put-nonsubjpart-pos1
|
condition-dat
|
|
|
|
|
|
|
|
əməl
|
et-məlidir.
|
|
|
|
|
practice
|
do-deb3
|
|
|
|
|
|
|
|
|
|
|
The boy who wants [to have] me must fulfill the condition I put.
(Səhrli nağıllar,6 şahzadə Bəndali)
|
|
|
(4)
|
alma-nı
|
zəhərlə-yən
|
o-nun
|
inan-dığ-ı
|
vəzir
|
idi
|
|
apple-acc
|
poison-subj
|
anaph-gen
|
trust-nonsubj-pos3
|
vizier
|
cop.pst3
|
|
The one who had poisoned the apple was the minister in whom he had trust.
(SN, Quru kəllə)
|
In (4), the constituent relativised is a dative argument, which is required by the verb inanmaq ‘believe’. When relativising a dative argument corresponding to the Recipient (the third argument of a three-place verb), this position may be occupied in the relative clause by a pronoun marked for the corresponding case (ona in [5]).7
(5)
|
pişiyin
|
(ona)
|
siçan
|
ver-diyi
|
uşağ-a
|
bax
|
|
cat.gen
|
dist-dat
|
mouse
|
give-nonsubjpart.pos3
|
child-dat
|
look.imp
|
|
‘Look at the child to whom the cat has given a mouse!’ (fn)
|
Possessors can also be relativized. As a rule, a cataphoric possessive suffix appears on the possessum within the relative clause in this case, pointing to the postposed head noun as the possessor. This is not rare, sometimes found instead of Recipient relativising constructions, as in (6):
(6)
|
ata-sı-nın
|
(ona)
|
paltar
|
al-dığ-ı
|
qız
|
sevin-ir.
|
|
father-pos3-gen
|
dist-dat
|
dress
|
take-nonsubj-pos3
|
girl
|
rejoice-prs3
|
|
‘The girl whose father has bought a dress for her is happy.’ (fn)
|
If the possessum functions as a subject in the RC, the subject-oriented participles are used (cf. [7]); otherwise a participle in -DIQ- is employed (cf. [8]):
(7)
|
ata-sı
|
öl-müş /
|
öl-ən
|
oğlan
|
ağla-yır.
|
|
father-pos3
|
die-perfpart
|
die-subjpart
|
boy
|
cry-prs3
|
|
‘The boy whose father has died / is dying, is crying.’ (fn)
|
|
|
(8)
|
tilsim-i-ni
|
sındır-dığ-ın
|
otaq
|
səmt-i-ndə
|
ged-əcək-sən
|
|
curse-pos3-acc
|
break-nonsubj-pos2
|
room
|
direction-pos3-loc
|
go-fut-2
|
|
‘You will go in the direction of the room whose curse you have broken.’
(SN, Ayğır Həsən)
|
The participial strategy is dominant in Azerbaijani because the DIQ-participle takes (possessive) indexation of the subject within the relative clause, a feature which is of considerable help in recovering the other function relativised. By contrast, Indo-european participles do not bear such personal indexes, and they are usually specialized and restricted to the relativisation of subject and direct objet functions.
Headless participial relative clauses are much less frequent than headed ones, even for the relativization of subjects (unlike in Turkish, where this is very frequent). One example could be observed in (4) above. Usually headless participial relatives are in the plural form (cf. [9]), or based on a passive verb, as in (10):
(9)
|
gel-ən-lər
|
|
come-subjpart-pl
|
|
‘the arriving ones’ (fn)
|
|
|
(10)
|
de-yil-ən-ə
|
gör-ə
|
|
say-pass-subjpart-dat
|
see-cv
|
|
‘according to what is said’ (fn)
|
Note that in current Azerbaijani (not influenced by Turkish) DIQ-participles may head only relatively short RCs, as in (11) and (12):
(11)
|
axtar-dığ-ım-ı
|
tap-mış-am
|
|
look_for-nonsubjpart-1-acc
|
find-perf-1
|
|
‘I found what I was looking for.’ (SN, Tapdıq)
|
|
|
(12)
|
biz-im
|
bugün
|
kerosin
|
adlandır-dığ-ımız
|
neft-dir
|
|
1pl-gen
|
today
|
kerosene
|
call-nonsubj-pos-1pl
|
oil-cop3
|
|
‘What we call today ‚kerosene’ is just oil.’
(Fərid Ələkbərov, Şair Mikayıl Müşfiq, Azerbaijan International jurnalı, 2002)
|
..2Participial vs. finite strategies in Persian
In Persian, relative clauses using the participial strategy are always short, postposed to their head, linked by an ezafe as in any other complex NP, and restricted to relativizing the subject of intransitive verbs (cf. [13]) and the object of transitive verbs (cf. [14]; in the latter case, the subject cannot be expressed). Such constructions are not much different from NPs with adjectives.
(13)
|
Persian
|
|
nâme-ye
|
rasi-de
|
az
|
Tehrân
|
|
letter-ez
|
arrive-perfpart
|
from
|
Teheran
|
|
‘a letter arrived from Tehran’ (fn)
|
|
|
(14)
|
sag-e
|
kotak
|
za-de
|
|
|
dog-ez
|
kick
|
beat-perfpart
|
|
|
‘a beaten dog’ (fn)
|
The dominant RC-forming strategy in Persian is a postposed finite clause introduced by the general subordinator ki. The antecedent noun takes a suffix -i, which is often considered to be a variant of the ezafe found in other noun-complementing (adjectival, genitival) constructions.8 If the subject-position is relativized, there is a gap in the relative clause:
(15)
|
zan-i
|
[ke
|
Ø
|
tanhâ
|
yek
|
gusfand
|
dâş-t]
|
|
woman-ez
|
rel
|
(nom)
|
alone
|
one
|
sheep
|
have-pst
|
|
‘a woman who had only one sheep’ (fn)
|
|
|
(16)
|
sag-i
|
[ke
|
mardom-râ
|
gaz migir-ad]
|
be-u
|
|
dog-def
|
rel
|
people-acc
|
biting prs.strike-3
|
dat-3
|
|
|
|
|
|
|
|
ostoxân
|
nade!
|
|
|
|
|
bone
|
proh.give
|
|
|
|
|
|
|
|
|
|
|
‘Do not give a bone to the dog which bites people.’ (fn)
|
This gapping strategy is also common with relativized objects, the gapped function being easily inferable from the presence of an unmarked nominative argument. (cf. [17]).
(17)
|
gusfand-i
|
ke
|
gorg
|
bor-d
|
|
sheep-def
|
rel
|
wolf(nom)
|
carry-pst3
|
|
‘The sheep which the wolf has taken away.’ (fn)
|
Other, oblique syntactic positions of the head-noun are occupied by a resumptive pronoun or clitic inside the relative clause (u in [18] and -ş in [19]).
(18)
|
sag-i
|
[ke
|
be-u
|
ostoxân
|
midim]
|
u-râ
|
nakuf !
|
|
dog- def
|
rel
|
dat-dist
|
bone
|
prs.give.1pl
|
3-acc
|
proh.beat
|
|
‘Do not beat the dog to which we give a bone.’ (fn)
|
|
|
(19)
|
zan-i
|
[ke
|
gorg
|
gusfand-eş
|
bor-d]
|
|
woman-def
|
rel
|
wolf(nom)
|
sheep-pos3
|
carry-pst3
|
|
‘the woman whose sheep a wolf had taken away’ (fn)
|
Note that this strategy for forming restrictive relative clauses is probably derived from an ‘adjoined relative clause’ type9 in which ke is used as a general linker and no affix is found to mark definiteness on the head-noun, and in which a resumptive pronoun (including one functioning as a direct object) is grammatical, though always optional (cf. [20]).
(20)
|
zan
|
yek
|
gusfand
|
dâşt
|
|
woman
|
one
|
sheep
|
have.pst3
|
|
|
|
|
|
|
ke
|
(ân-râ)
|
gorg
|
(az-u)
|
bor-d
|
|
rel
|
(dist-acc)
|
wolf
|
(abl-3)
|
carry-pst3
|
|
|
|
|
|
|
|
‘The woman had a sheep, which a wolf took away.’ (fn)
|
..3The participial strategy in Tati
All Tati dialects make use of the participial suffix inherited from (old) Persian –ta-ka- (derived of Indo-European *-to-), with the two allomorphs -de (after consonant-stems) and -re (after vowel-stems), hereafter subsumed under -De. However, Apsheron Tati,10 which has been in contact with Azerbaijani for many centuries, makes much wider use of participles than Persian. We will first describe how Apsheron Tati uses the inherited participle without the orientation constraint usually associated with the related forms in other Indo-European languages, and second, how the rather restricted Persian agent noun forms ending in –degor has become a productive participle formation.
..1.The inherited participial strategy in Tati
In contrast to Persian, the participial strategy is dominant in Apsheron Tati, certainly as an imitation of the surrounding Azerbaijani. The relative clause headed by such De-participles continues to behave like its Indo-European counterparts in that it permits relativisation of intransitive subjects (cf. [21]) and direct objects, but also transitive subjects and some oblique positions in any factual – either past or non-past – context.
(21)
|
Apsheron Tati
|
|
e-piyer-i
|
men-de
|
ye-to
|
baq
|
doşd
|
|
abl-father-pos3
|
remain-part
|
one
|
garden
|
have.pst3
|
|
‘He had a garden which remained from his father.’ (fn)
|
The exact temporal interpretation of such participles depends on the context (cf. [22]/present and [23]/past).
(22)
|
rous-de
|
seg
|
dendu
|
ne-bzeren
|
|
bark-part
|
dog
|
tooth
|
neg-evt.strike.3
|
|
‘A dog who barks does not bite.’ (fn)
|
|
|
(23)
|
rous-de
|
seg
|
kuf-de
|
bü
|
|
bark-part
|
dog
|
beat-part
|
be.pst.3
|
|
‘The dog who barked was beaten.’ (fn)
|
When the direct object is relativised, the subject can be omitted, as in (24).
(24)
|
nug
|
voçi-re
|
gendüm-yo
|
|
new
|
collect-part
|
corn-pl
|
|
‘the corn newly reaped’ (fn)
|
If, however the subject is expressed in the RC, it remains in the nominative case (a feature sometimes found in informal spoken Azerbaijani as well), and an attributive marker is usually added to the participle ending (in (25), but not in (26)):
(25)
|
[imu
|
kuf-de-yi]
|
seg
|
|
1pl.nom
|
beat-part-attr
|
dog
|
|
‘the dog which we have beaten’ (fn)
|
|
|
(26)
|
merd-i
|
üsde-re
|
xune
|
be-xoş
|
zen
|
ne-ma-ren.
|
|
|
man-pos3
|
buy-part
|
house
|
dat-pleasure
|
woman
|
neg-come-prs3
|
|
|
‘The house that her husband has bought does not please his wife.’ (fn)
|
|
This construction with an overt subject within the relative clause is the one normally found when the position relativised is the Recipient of a ditransitive verb (cf. [27]).
(27)
|
imu
|
sugum
|
de-re-yi
|
seg
|
|
1pl
|
bone
|
give-part-attr
|
dog
|
|
‘the dog to whom we gave a bone’ (fn)
|
Note that the attributive ending should be carefully distinguished from the third person possessive suffix (a morphological borrowing from Azerbaijani), which appears in those RCs where the head-noun is a possessor, as in (28).
(28)
|
imu
|
guş-yeyi-re
|
keş-re-yi
|
seg
|
|
we
|
ears-pl.pos3-acc
|
pull-part-attr
|
dog
|
|
‘the dog whose ears we pull(-ed)’ (fn)
|
If the head of the RC is a noun expressing space (cf. [29]) or time (cf. [30]), a syntactic gap in the RC is not leading to similar ambiguities. Its probable function being obvious and easily recoverable, the Tati De-participle can be used in very much the same way as the DIQ-participle of Azerbaijani (cf. Section 2.1):
(29)
|
be
|
xisi-re
|
cege
|
ure
|
tike-tike
|
enci-r-ünd,
|
|
dat
|
sleep-part
|
place
|
dist-acc
|
piece-piece
|
cut-pst3pl
|
|
‘at the place where he was sleeping, they cut him to pieces’ (fn)
|
|
|
(30)
|
ser-e
|
na-re
|
saat
|
xo
|
ber-d
|
u-re.
|
|
head-acc
|
put-part
|
hour
|
sleep
|
carry-pst3
|
dist-acc
|
|
‘As soon as he put his head down, sleep took him.’ (fn)
|
The use of the participial strategy in attributive relative clauses is thus not restricted to the relativization of the single argument of intransitives and the Patient of transitives. This strategy allows the relativization of most syntactic positions, including that of a transitive subject: this is an innovation, and a major deviant feature from the inherent S/P orientation of the corresponding Persian participle.
..2.Voice neutrality of the S/P oriented participle in Tati
Since Tati, unlike Azerbaijani, has only one participial attributive verb-form, the participial strategy alone cannot disambiguate between object and subject relativization by way of the inherent ‘orientation’ of the participle employed. The De-participle had an inherited, ergatively patterned orientation, permitting relativisation of intransitive subjects and direct objects, while for the relativisation of an agent another participle, lost in Tati, was used. Nevertheless, the once S/P oriented participle is also used in agentive function.
In some idiomatic expressions (e.g. [31] and [32]), the object cannot be marked for the definite-accusative case and the predicate is not prototypically transitive:
(31)
|
dünya
|
di-re
|
adam-yo
|
|
world
|
see-part
|
person-pl
|
|
‘people who have seen the world’ (fn)
|
|
|
(32)
|
Molla
|
yeto
|
xob-e
|
şir
|
de-re
|
go
|
doş-de-s.
|
|
Molla
|
one
|
good-attr
|
milk
|
give-part
|
cow
|
have-perf-3
|
|
‘Mulla had a cow giving good milk.’ ( = ‘good milk-giving cow’) (fn)
|
However, relativised subjects can have a definiteobject (accusative-marked) in the participial RC:
(33)
|
şir-e
|
xar-de
|
nozu
|
ez-i
|
xune
|
nisdü.
|
|
milk-acc
|
eat-part
|
cat
|
abl-prox
|
house
|
negcop3
|
|
‘The cat who has drunk the milk is not from this house.’ (fn)
|
Participial RCs may also relativise possessors of possessa with subject function in the RC (cf. [34]), as well as possessors of nominals with non-subject function (cf. [35]).
(34)
|
piyer-i
|
mür-de
|
gede
|
herey
|
ze-ren.
|
|
father-pos3
|
die-part
|
boy
|
shout
|
strike-prs3
|
|
‘The boy whose father has died is crying.’ (fn)
|
|
|
(35)
|
xuniyi-re
|
tü
|
di-re
|
kile
|
e-mü
|
xuvar-mü-nü.
|
|
|
house.pos3-acc
|
2
|
see-part
|
girl
|
abl-1
|
sister-1-cop3
|
|
|
‘The girl whose house you see is my sister.’ (fn)
|
|
‘Voice neutral participles are also found in neighbouring East-Caucasian languages with ergative case-marking (cf. for instance Authier 2009 for Kryz, spoken in Azerbaijan where it neighbours with the main Tati speaking area). However, unlike these languages, Tati has no ergative marking of core arguments, retaining as it does Indo-European nominative-accusative alignment. This is sufficient to disambiguate Agent and Patient roles, one of these being usually expressed and duely flagged within the relative clause.
..3.Headless participial RCs in Tati
Headless RCs are only rarely derived from attributive participial constructions in Apsheron Tati. In some instances the participle can be nominalized if it relativises a direct object. In this case, the subject is treated as the possessor of a possessive NP in which the participle is the possessum, and as such marked by the ezâfe suffix11 (this seems to be the most probable origin of the attributive suffix illustrated above, although the syntactic reordering is difficult to explain):
(36)
|
di-re-yi
|
şumu
|
des-yo-mü
|
nistü
|
|
see-part-ez
|
you
|
arm-pl-pos1
|
neg.cop.3
|
|
‘What you see are not my arms.’ (fn)
|
Unlike in Azerbaijani, Tati participles relativizing oblique positions cannot be nominalized: a generic head-noun is added, like zen ‘woman’ in the translation of the Azerbaidjani example in (37), which is given in (38).
(37)
|
Azerbaijani
|
|
sən-in
|
danış-dığ-ın
|
mən-im
|
ana-m
|
ol-ub.
|
|
2-gen
|
talk-nonsubjpart-pos2
|
1-gen
|
mother-pos1
|
be-perf3
|
|
‘The woman you were talking to was my mother.’
(SN, Tapdıq)
|
|
|
(38)
|
Aspheron Tati
|
|
tü
|
hadi
|
sax-de-yi
|
zən
|
ə-mü
|
moy-mü
|
bi-res
|
|
2
|
talk
|
do-perfpart-ez
|
woman
|
abl-1
|
mother-pos1
|
be-perf.3
|
|
‘The woman you were talking to was my mother.’ (fn)
|
If the position relativized is the one of a subject (intransitive or transitive/Agent), or the possessor of the RC’s subject, either a derived form in -gar (cf. [39a]) or a generic noun (cf. [39b]) or has to be used, or both (cf. [40]).
(39)
|
a.
|
burvar-tü-ne
|
küş-degar
|
ez-i
|
di
|
nisdü.
|
|
|
|
|
brother-pos2-acc
|
kill-partsubst
|
abl-prox
|
village
|
negcop3
|
|
|
|
b.
|
|
küş-de
|
adam
|
|
|
|
kill-part
|
person
|
|
‘The one / person who killed your brother is not from this village.’ (fn)
|
|
|
(40)
|
burvar-tü-ne
|
küş-de-gar
|
adam12
|
di-imu
|
nisdü.
|
|
brother-pos2-acc
|
kill-partsubst
|
person
|
village-1pl
|
negcop3
|
|
‘The one who killed your brother is not a person from this village.’ (fn)
|
|
|
Further examples of gar-participles are given in (41)–(43). Note that this form is also found in Persian, where it derives nomina agentis.
(41)
|
ay
|
mü-ne
|
xas-degar-yo!
|
|
Oh
|
1-acc
|
want-subjpart-pl
|
|
‘Oh, you who love me!’ (fn)
|
|
|
(42)
|
be-möçüd
|
bi-regar-yo
|
hamme
|
xis-ren
|
|
dat-mosque
|
be-subjpart-pl
|
all
|
sleep-prs.3
|
|
‘All those who are in the mosque are sleeping.’ (fn)
|
|
|
(43)
|
mür-degar,
|
des-i
|
xurd
|
bi-regar
|
|
die-subjpart
|
hand-pos3
|
broken
|
be-subjpart
|
|
‘deads, people whose hand was cut.’ (fn)
|
|