On possessed relative clauses in Kyrgyz

Abstract

I argue that Dependent-Case Theory approaches cannot account for the subject case alternation in Kyrgyz relative clauses, supporting previous accounts arguing against a DCT analysis (Kornfilt 2009, Gribanova ms).

Full text

ONPOSSESSED RELATIVE CLAUSES IN KYRGYZ *
SABINE LASZAKOVITS
University of Connecticut
1 Introduction
This paper is concerned with case and agreement in Kyrgyz (Turkic) object relative clauses (RCs).
These clauses show an alternation between nominative (NOM) and genitive (GEN) case on the
subject of the relative clause — similar to the -ga/-no alternations in Japanese (Hiraiwa, 2001,
Miyagawa, 1993, 2008, Ochi, 2001, Watanabe, 1996, i.a.). The subject’s case is correlated with
the absence or presence of possessive agreement (POSSAGR) with the subject, which appears on
the RC’s head noun (Kornfilt, 2004, 2008, 2009). This paradigm is illustrated in (1) and (2).1
(1a) shows a RC with a NOM-marked subject and no POSSAGR. (2b) illustrates a RC with GEN
on the subject and with POSSAGR. It is impossible to have GEN without POSSAGR, (2a). The
combination of NOM with POSSAGR is attested, but must be interpreted as agreement with a silent
(pro-dropped) possessor rather than with the NOM-marked subject, (1b).
(1) a. aytmatıf
Aitmatov.NOM
caz-gan
write-PCPL
kitep
book
‘the book that Aitmatov wrote’
b. aytmatıf
Aitmatov.NOM
caz-gan
write-PCPL
kiteb
book
-i
-3SG
only: ‘her/his book that Aitmatov wrote’
(2) a. * aytmatıf
Aitmatov
-tın
-GEN
caz-gan
write-PCPL
kitep
book
*MANY THANKS TO MY KYRGYZ CONSULTANTS,ESPECIA LLY TO TOLGONAJ B., VENERA MAMBAE VA,
AND NURSULTAN ZHANYBEKOV. FOR COMMENTS ON THIS OR AN EAR LI ER V ER SION, I THA NK
JONATHAN BOBALJ IK , ŽE LJ KO BOŠKOV I ´
C, HOSSEP DOLATI AN, MIA GONG, VERA GRIBAN OVA, ESZTER
ÓTOTT-KOVÁCS, NAZILA SHA FIE I, SUSI WURMBRAND,AN D THE AUD IE NC ES O F A UCON N LINGLUNCH,
CONCALL 3, AND WAFL 14. THE FIELD WORK WAS SUPPORTED BY THE UCONN RESEARCH
EXCELLENCE PROGRAM (PI: BOBALJ IK )AN D EX EC UTED BY KSENIA BOGOMOLETS, CHRISTO S
CHRISTOPOULOS, PASHA KOVAL,AND MYSELF.
1I use Turkish orthography for the Kyrgyz examples. Deviations from IPA are as follows: hci=[Ã], hçi=[Ù], hgi=[g]
or [G], h˘
gi=[j] or [:], hıi=[W], hji=[Z], hki=[k] or [q], höi=[œ], h¸si=[S], hüi=[y], hyi=[j].
1

2Laszakovits
b. aytmatıf
Aitmatov
-tın
-GEN
caz-gan
write-PCPL
kiteb
book
-i
-3SG
‘the book that Aitmatov wrote’ (not: ‘her/his book that Aitmatov wrote’)
These constructions have been used to argue for a Head-Licensing approach to case assignment
(Chomsky, 2000, 2001), where genitive case must be licensed by a D-head carrying possessive
agreement (Baker and Vinokurova, 2010, Gribanova, 2017, Kornfilt, 2004, 2006, 2008, 2009).
They pose a problem for configurational approaches to case assignment (Baker, 2015, Baker and
Vinokurova, 2010, Bobaljik, 2008, Levin and Preminger, 2015, Marantz, 1992, Yip et al., 1987),
where case is assigned independently of functional heads, based on two observations:
1. Genitive never appears without an agreeing possessive suffix.2
2. Each possessive suffix can only agree with one genitive-marked NP.
Similar data can be found in simple possession constructions (3) and argument clauses (4):
(3) a. * nursultan
Nursultan
-dın
-GEN
kitep
book
b. nursultan
Nursultan
-dın
-GEN
kiteb
book
-i
-3SG
‘Nursultan’s book’
(4) a. * bakıt
Bakit.NOM
aytmatıf
Aitmatov
-tın
-GEN
kitep-ti
book-ACC
caz-gan
write-PCPL
-dı
-ACC
ayt-tı.
say-PST
b. bakıt
Bakit.NOM
aytmatıf
Aitmatov
-tın
-GEN
kitep-ti
book-ACC
caz-gan
write-PCPL
-ı
-3SG
-n
-ACC
ayt-tı.
say-PST
‘Bakit said that Aitmatov wrote the book.’
Examples (2a), (3a), and (4a) show that it is not possible to have GEN without POS SAGR.
This paper examines these observations in light of constructions where object relative clauses
modify head nouns with a GEN-marked possessor.
1.1 RCs and possessors
When adding a possessor to a noun that is modified by an object RC, we observe an asymmetry
between NOM-RCs and GEN-RCs. NOM-RCs as in (1a) above can appear with a possessor, as
illustrated in (5a), while GEN-RCs as in (2b) above cannot, (5b).
(5) a. nursultan
Nursultan
-dın
-GEN
aytmatıf
Aitmatov.NOM
caz-gan
write-PCPL
kiteb
book
-i
-3SG
‘Nursultan’s book that Aitmatov wrote’
b. * nursultan
Nursultan
-dın
-GEN
aytmatıf-tın
Aitmatov-GEN
caz-gan
write-PCPL
kiteb
book
-i
-3SG
2Öztürk and Erguvanlı Taylan (2016) show that in Turkish, genitives can sometimes appear in the absence of an
agreeing possessive suffix as modifiers and in predicative position under special semantic restrictions of the possessee.
As I do not have the parallel data for Kyrgyz, I will leave these constructions aside.

On Possessed Relative Clauses in Kyrgyz 3
Turkish differs from Kyrgyz in a relevant respect: possessive agreement with GEN-marked
subjects is not placed on the head noun, but on the participle. This is illustrated in (6a). Kyrgyz
does not allow this position of POSSAGR, (6b).
(6) a. aytmatov
Aitmatov
-un
-GEN
yaz-dı˘
g
write-PCPL
-ı
- 3SG
kitap
book
‘the book that Aitmatov wrote’ (Tu)
b. * aytmatıf
Aitmatov
-tın
-GEN
caz-gan
write-PCPL
-ı
-3SG
kitep
book
(Ky)
The Head-Licensing approach to case makes a prediction (Kornfilt, 2009): in Turkish, GEN-
marked subjects of RCs are “possessors” of the participle, thus the head noun should be able to
have an independent possessor, licensed by an independent possessive suffix. This is borne out:
(7) nursultan
Nursultan
-ın
-GEN
aytmatov
Aitmatov
-un
-GEN
yaz-dı˘
g
write-PCPL
-ı
- 3SG
kitab
book
-ı
- 3SG
‘Nursultan’s book that Aitmatov wrote’ (Tu)
In Kyrgyz, both suffixes would have to occur on the head noun, which seems to be prohibited
cross-Turkically (Kornfilt’s, 1986, “Stuttering Prohibition”), and can be interpreted as a selectional
restriction stating that the phrase projected by possessive agreement cannot select another phrase
of the same type.
(8) * nursultan
Nursultan
-dın
-GEN
aytmatıf
Aitmatov
-tın
-GEN
caz-gan
write-PCPL
kiteb
book
-i
- 3SG
-si
- 3SG
intended: ‘Nursultan’s book that Aitmatov wrote’ (Ky)
Note that there is no apparent semantic reason why this should not be available. The possessive
construction can be interpreted with a variety of relations between possessor and possessee:
(9) a. nursultan
Nursultan
-dın
-GEN
kiteb
book
-i
-3SG
‘Nursultan’s book (e.g., that he owns)’
b. nursultan
Nursultan
-dın
-GEN
kiteb
book
-i
-3SG
‘Nursultan’s book (e.g., that he wrote)’
The Stuttering Prohibition and the Head-Licensing approach to genitive can thus explain the
unacceptability of (5b) as well as the contrast with Turkish (7).
2 Case assignment
2.1 Dependent Case Theory
In Dependent Case Theory (DCT) (Baker, 2015, Baker and Vinokurova, 2010, Bobaljik, 2008,
Levin and Preminger, 2015, Marantz, 1992, Yip et al., 1987), the distribution of case is solely

4Laszakovits
dependent on the presence or absence of other NPs and on case assignment domain boundaries.
The following rules are adapted from Marantz (1992) and Baker (2015):
(10) a. ACC is assigned to an NP xiff xis in the CP phase and there is another NP yin the
same phase that c-commands x(“dependent-down case”).
b. DAT is assigned to an NP xiff xis in the VP phase and there is another NP zin the
same phase that xc-commands (“dependent-up case”).
c. NOM is assigned to an NP xin the CP phase if xis not assigned any other case
(“unmarked case”).
d. GEN is assigned to an NP xin the DP phase if xis not assigned any other case
(“unmarked case”).
All NPs will always be assigned some case, so there is no way for the derivation to crash due to
lack of case assignment (Case Filter).
The dependency between case and
φ
-agreement is reverted compared to the Head-Licensing
model (Bobaljik, 2008, Levin and Preminger, 2015, Preminger, 2014): Rather than
φ
-agreement
licensing case, case licenses
φ
-agreement. Probes on the agreeing head (T for finite verbs, D for
possessive agreement) look in their search space for NPs to agree with, and they are only sensitive
to NPs with particular cases. For Sakha (Turkic), Levin and Preminger (2015) suggest that the
probe on D can only agree with NOM- or GEN-marked NPs. If D does not find any NPs, it spells
out as default, which in Sakha is zero-marking.
2.2 DCT in Turkic
Baker and Vinokurova (2010) have shown that in Sakha (Turkic), the distribution of accusative
case can be explained by DCT, but not by the Head-Licensing theory. Accusative appears on the
subjects of embedded clauses, and is dependent on the presence of a matrix subject with which it
enters into case competition according to rule (10a). Crucially, accusative cannot be explained by
making reference to the accusative-assigning capabilities of the verb in the embedded clause nor
of the verb in the matrix clause.
Baker and Vinokurova (2010) go on to argue that DCT is not sufficient to describe the Sakha
case pattern and argue for a hybrid case-assignment system where some cases (ACC,DAT) are
determined by DCT, and other cases (NOM,GEN) are determined by the Head-Licensing model.
The necessity for a hybrid approach has been refuted by Levin and Preminger (2015), who argue
for a pure DCT model. Recently, Gribanova (2017) has taken up this issue for Uzbek and provided
evidence for a hybrid approach based on adjunct clauses. Additionally to adjunct clauses in several
Turkic languages, relative clauses pose a challenge for DCT as well, as this paper points out.
2.3 Problems with DCT
The difficulty for DCT is to explain the ungrammaticality of examples like (5b), repeated here:
(11) * nursultan
Nursultan
-dın
-GEN
aytmatıf-tın
Aitmatov-GEN
caz-gan
write-PCPL
kiteb
book
-i
-3SG
intended: ‘Nursultan’s book that Aitmatov wrote’ = (5b)
The case assignment domain is nominal, thus making GEN the unmarked case. If no dependent
cases are defined for this domain, or if the two NPs do not stand in a c-command relation, we

On Possessed Relative Clauses in Kyrgyz 5
predict unmarked case to appear on both NPs. The probe on D looks for a GEN-marked NP, and
finds a target, and spells out the target’s features.
What could DCT say to rule out such examples? Perhaps case assignment proceeds as predicted
but
φ
-agreement fails and this crashes the derivation. If the structural configuration is such that the
probe cannot decide with which NP to agree, it might try to agree with both targets, spelling out the
features of both. In this case, the derivation could crash due to incompatible features that the probe
cannot spell out. However, in (5b), both targets are specified for 3SG, thus no incompatibility can
arise.
Instead of agreeing with both targets, the probe could be specified in such a way that it is not
able to agree with either, thus remaining unsatisfied, and this could crash the derivation. However,
by common assumption, the possessor and the subject of the RC are not equidistant from the probe.
If the probe is on D, and is specified to look up or down (first), and the possessor is in Spec,DP,
and the subject is inside the RC in the complement of D, then the probe will be able to determine
a single target.
Perhaps the problem is with case assignment and we should analyze GEN not as unmarked
case in the nominal domain, but as a dependent-up case like dative or ergative. It seems to hold
that in languages with ERG as dependent-up case, we do not find CPs with 3 arguments where
more than one argument is marked ERG (Baker and Bobaljik, 2017, Nie, 2017). The reason that
we do not find more than one GEN-marked argument could be parallel. However, GEN cannot be
a dependent-up case in Kyrgyz (and Turkic more generally) because we find GEN on subjects of
intransitives (assuming no cognate object):
(12) kristos
Christos
-tun
-GEN
kül-gön
laugh-PCPL
kün
day
-ü
-3SG
‘the day on which Christos laughed’ (Ky)
DCT could be saved if GEN-subjects are possessors, and if there is a unique position for
possessors. Then, the ungrammaticality of (5b) results from two NPs trying to occupy the same
structural position. However, data from NPI-licensing show that GEN-subjects are in a different
structural position from possessors: Possessors can be NPIs licensed by matrix negation (13a), but
GEN-subjects cannot (13b).
(13) a. bakıt
Bakit
eç
any
kim
who
-din
-GEN
kiteb
book
-i
-3SG
-n
-ACC
satıp al-ba-dı.
buy-NEG-PST
‘Bakit didn’t buy anybody’s book.’
b. * bakıt
Bakit
eç
any
kim
who
-din
-GEN
oku-gan
read-PCPL
kiteb
book
-i
-3SG
-n
-ACC
satıp al-ba-dı.
buy-NEG-PST
lit.: ‘Bakit didn’t buy the book that anyone had read.’
To conclude, in the absence of an independent explanation, DCT cannot capture the
ungrammaticality of (5b), while the Head-Licensing theory, which requires that each GEN-marked
NP have a designated possessive agreement head, can.
3 Analysis
In this section, I propose an analysis that captures the data from possessive agreement and NPI
licensing and is compatible with a Head-Licensing approach to the distribution of genitive case.

6Laszakovits
Recall the NPI data in (13b) where matrix negation is not able to license an NPI as the GE N-
marked subject of a RC. Together with the pattern for possessive agreement, this yields a 3-way
classification:
(14) possessor subject-GEN subject-NOM
poss. agreement with ✓ ✓ ✗
NPI lic. by matrix neg. on ✓ ✗ ✗
Further data from NPI licensing consolidate the split between GEN-marked subjects and NOM-
marked subjects. Negation inside the RC can license NOM-marked subjects (15a), but not GEN-
marked subjects (15b).
(15) a. bakıt
Bakit
eç
any
kim
who.NOM
oku-ba-gan
read-NEG-P CPL
kitep
book
-ti
-ACC
satıp aldı.
bought
‘Bakit bought the book that noone had read.’
(lit.: ‘Bakit bought the book that anyone hadn’t read.’)
b. * bakıt
Bakit
eç
any
kim
who
-din
-GEN
oku-ba-gan
read-NEG-P CPL
kiteb
book
-i
-3SG
-n
-ACC
satıp aldı.
bought
It follows thus that GEN-marked subjects must be in a position where the
φ
-probe on D can reach
them, Neg in the matrix clause cannot reach them, and Neg in the embedded clause also cannot
reach them.
I assume a variable theory of phases (Bobaljik and Wurmbrand, 2005, Boškovi´
c, 2014): in
each spine, the highest projection constitutes the phase. Following Hale (2002), Krause (2001), the
highest projection in the participial RC is AspP; and calling the head hosting possessive agreement
D, this is the phase of the nominal spine. A phase head can see into the spell-out domain it closes
off, i.e., into its complement, but also anything in the same spell-out domain as itself.
I propose that GEN-marked subjects are outside the spell-out domain closed off by Agr, but not
as high as possessors in Spec,DP. 3This is illustrated in (16), together with possessive
φ
-agreement.
The box indicates the spell-out domain introduced by Asp, continuous arrows indicate accessible
targets, and dashed arrows unaccessible targets. The probe prefers agreement with its specifier
over agreement with the GEN-marked NP in its complement.
3For concreteness, I assume that GE N-marked subjects are adjuncts to AspP, but nothing hinges on this.
Alternatively they could reside in an intermediate projection. Whether such a projection is motivated depends on
a more precise characterization of the difference between NOM-RCs and GEN-RCs. Tests that I have run that have not
shown a difference include topicalization with the topicalizer bolso, quantificational subjects, and matrix interpretation
of subject- and non-subject wh-words. I have not yet been able to derive results from the ‘probability with which’-tests
from Miyagawa (1993, 2013).

On Possessed Relative Clauses in Kyrgyz 7
(16) DP
PossGEN D’
NP
AspP
SubjGEN AspP
vP
SubjNOM v’
VP v
Asp
N
D
This structure also derives the observed NPI licensing patterns. In (17), matrix negation is able
to license an NPI as possessor, but not inside the complement of D, which is a closed-off spell-out
domain. In (18), embedded negation can reach the NOM-marked subject in Spec,vP, but not the
GEN-marked subject in a higher spell-out domain (nor the possessor even higher up).
(17) . . .
DP
PossGEN D’
NP
AspP
SubjGEN AspP
N
D
Neg
(18) AspP
SubjGEN AspP
NegP
vP
SubjNOM v’
VP v
Neg
Asp
!
This position of Neg is motivated by the morpheme order in the embedded verb.4
4I remain agnostic as to the exact licensing requirements of NPIs. Do Kyrgyz NPIs need to be c-commanded by
Neg? Then NOM-subjects are c-commanded and GEN-subjects aren’t, so GEN-subjects are higher than NOM-subjects.
Do Kyrgyz NPIs need to be in the same clause/phase as Neg? Then NOM-subjects are and G EN-subjects aren’t, so
GEN-subjects are higher than NOM-subjects.

8Laszakovits
3.1 Argument clauses
Additional evidence for this structure comes from the subject-case-alternations in Kyrgyz non-
finite argument clauses: the subject can receive NOM or GEN, possessive agreement is present on
the participle, and possessive agreement can agree with both NOM- and GEN-subjects, unlike in
RCs.
(19) a. bakıt
Bakit.NOM
men
me. NOM
üy-gö
house-DAT
kel-gen
come-PCPL
-im
- 1SG
-di
-ACC
ayt-tı.
say-PST
‘Bakit said that I had come home.’
b. bakıt
Bakit.NOM
men
me
-in
-GEN
üy-gö
house-DAT
kel-gen
come-PCPL
-im
- 1SG
-di
-ACC
ayt-tı.
say-PST
‘Bakit said that I had come home.’
I propose that this agreement paradigm arises because in argument clauses, the N layer hosting the
head noun of the RC is absent, whereby D extends the phase introduced by Asp. Thus D can see
inside the complement of Asp and agree with a low NOM-marked subject, (20).
(20) (argument clause)
DP
AspP
SubjGEN AspP
vP
SubjNOM v’
VP v
Asp
D
We now make a prediction. If Asp is not a phase head in argument clauses, NPI licensing should
pattern differently as well: embedded negation should be able to license GEN-marked subjects
outside the complement of Asp. This is borne out, (21).
(21) bakıt
Bakit.NOM
eç
any
kim
who
-din
-GEN
üy-gö
house-DAT
kel
come
-be
-NEG
-gen
-PCPL
-i
-3SG
-n
-ACC
ayt-tı.
say-PST
‘Bakit said that noone came home.’ (lit.: ‘Bakit said that anyone hadn’t come home.’)
3.2 Case assignment
So far we have only seen the derivation of possessive agreement and NPI licensing, but not of case
assignment.
There are (at least) the following possibilities to capture the fact that every GEN needs its own
POSSAGR, but not every POSSAGR needs a GEN.

On Possessed Relative Clauses in Kyrgyz 9
The first possibility is that D carries a probe looking for a target for
φ
-agreement, and D
optionally assigns GEN to the NP it agrees with. The contexts for this optionality need to be
made more explicit, but are necessary to capture that in argument clauses (19a) (and perhaps
compounds), we observe PO SSAGR with a NOM-marked NP.
The second possibility is that D carries two probes, as Gribanova (2017) suggests for Uzbek:
one looking for a target for
φ
-agreement, and one assigning GEN. Gribanova shows that in Uzbek
argument clauses that are predicated by a copula verb, the
φ
-probe on D obeys a person hierarchy
effect, yielding examples where the
φ
-target and the GEN-target are distinct.
The third option is that GEN is assigned configurationally with DCT and needs additional
licensing. To capture the dependency of GEN on D we would need to put a probe on the
GEN-marked NP that looks for a D head that has not yet been targeted by a similar probe. This
approach does not seem promising as the presence of such a probe would need to depend on the
case the NP receives.
A fourth option, as mentioned before, arises if an independent reason can be found that rules
out the presence of both a GEN-marked subject of a relative clause and a GEN-marked possessor.
Then, DCT would be a viable strategy for case marking in Kyrgyz RCs.
In the absence of such an indepent reason, this paper concludes that DCT cannot deal with the
present data and that D must be responsible for licensing GEN. This is parallel to the conclusion
reached in Baker and Vinokurova (2010) and Gribanova (2017), albeit for a different Turkic
language and based on a different construction. I leave it for the future to work out the details
regarding which of the above options is the correct one for Kyrgyz.
4 Conclusions
I have re-examined data showing that in Kyrgyz (as in many other Turkic languages), object relative
clauses with a GEN-marked subject cannot be combined with a GEN-marked possessor on the head
noun of the relative clause (Kornfilt, 2008, 2009). This construction bears on the question of
whether a Dependent Case Theoretical approach to case assignment is better suited for Turkic
languages than a Head-Licensing approach (Baker and Vinokurova, 2010, Gribanova, 2017, Levin
and Preminger, 2015).
I have argued that DCT cannot capture the present data without making reference to an external
factor, such as competition for the same structural position. In the absence of such an external
factor, these data are another example of where Turkic languages must license case via a functional
head, in addition to the arguments presented in Gribanova (2017).
References
Baker, Mark. 2015. Case: Its principles and its parameters. Cambridge University Press.
Baker, Mark, and Jonathan Bobaljik. 2017. On inherent and dependent theories of ergative case. In
The Oxford Handbook of Ergativity, ed. Jessica Coon, Diane Massam, and Lisa deMena Travis.
Oxford University Press.
Baker, Mark, and Nadya Vinokurova. 2010. Two modalities of case assignment: case in Sakha.
Natural Language & Linguistic Theory 28:593–642.

10 Laszakovits
Bobaljik, J.D., and Susi Wurmbrand. 2005. The Domain of Agreement. Natural Language and
Linguistic Theory 23:809–865.
Bobaljik, Jonathan D. 2008. Where’s phi? Agreement as a post-syntactic operation. In Phi theory:
Phi-features across interfaces and modules, ed. Daniel Harbour, David Adger, and Susana Béjar,
295–328. Oxford: Oxford University Press.
Boškovi´
c, Željko. 2014. Now I’m a Phase, Now I’m Not a Phase: On the Variability of Phases
with Extraction and Ellipsis. Linguistic Inquiry 45:27–89.
Chomsky, Noam. 2000. Minimalist inquiries: the framework. In Step by step: Essays on
Minimalist Syntax in Honor of Howard Lasnik, ed. Roger Martin, David Michaels, and Juan
Uriagereka, 89–155. MIT Press.
Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: a life in language, ed. Michael
Kenstowicz, 1–52. MIT Press.
Gribanova, Vera. 2017. Case, agreement, and differential subject marking in Uzbek nominalized
clauses. Ms. Stanford, July 4 2017.
Hale, Ken. 2002. On the Dagur Object Relative: Some Comparative Notes. Journal of East Asian
Linguistics 11:109–122.
Hiraiwa, Ken. 2001. On nominative-genitive conversion. In A few from Building E39, ed.
Elena Guerzoni and Ora Matushansky, Vol. 39 of MIT Working Papers in Linguistics, 66–125.
Cambridge, MA: MIT Press.
Kornfilt, Jaklin. 1986. The stuttering prohibition and morpheme deletion in Turkish. In
Proceedings of the Second International Conference in Turkish Linguistics, ed. A. Aksu-Koç
and Eser Erguvanlı-Taylan, Vol. 400, 59–83. Istanbul: Bo˘
gaziçi University Publications.
Kornfilt, Jaklin. 2004. Agreement and its placement in Turkic non-subject relative clauses.
In Handbook of Comparative Syntax, ed. Guglielmo Cinque and Richard Kayne, 513–541.
Oxford/New York: Oxford University Press.
Kornfilt, Jaklin. 2006. Agreement: The (unique and local) syntactic and morphological licenser of
subject case. In Studies on agreement, ed. João Costa and Maria Christina Figueiredo, 141–171.
John Benjamins.
Kornfilt, Jaklin. 2008. Subject case and Agr in two types of Turkic RCs. In Uluta¸s and Boeckx
(2008), 145–168.
Kornfilt, Jaklin. 2009. A Constraint on Certain Relative Clauses in Turkic. International Journal
of Central Asian Studies 13:373–398.
Krause, Cornelia. 2001. On reduced relatives with genitive subjects. Doctoral Dissertation, MIT.
Levin, Ted, and Omer Preminger. 2015. Case in Sakha: are two modalities really necessary?
Natural Language & Linguistic Theory 33:231–250.

On Possessed Relative Clauses in Kyrgyz 11
Marantz, Alec. 1992. Case and licensing. In Proceedings of the Eastern States Conference on
Linguistics (ESCOL) 1991, ed. German Westphal, Benjamin Ao, and Hee-Rahk Chae, 234–253.
Ithaca, NY: Cornell University, CLC Publications.
Miyagawa, Shigeru. 1993. LF Case-checking and Minimal Link Condition. In Case and Agreement
II, Vol. 19 of MIT Working Papers in Linguistics, 213–254. MIT Press.
Miyagawa, Shigeru. 2008. Genitive subjects in Altaic. In Uluta¸s and Boeckx (2008), 181–198.
Miyagawa, Shigeru. 2013. Strong Uniformity and ga/no conversion. English Linguistics 30:1–24.
Nie, Yining. 2017. Why is there NOM-NOM and ACC-ACC but no ERG-ERG? In Proceedings
of NELS 47, ed. A. Lamont and K. Tetzloff, Vol. 2, 315–328.
Ochi, Masao. 2001. Move F and Ga/No Conversion in Japanese. Journal of East Asian Linguistics
10:247–286.
Öztürk, Balkız, and Eser Erguvanlı Taylan. 2016. Possessive constructions in Turkish. Lingua
182:88–108.
Preminger, Omer. 2014. Agreement and its failures, Vol. 68 of Linguistic Inquiry Monographs.
MIT Press.
Uluta¸s, Süleyman, and Cedric Boeckx, ed. 2008. Proceedings of wafl 4, Vol. 56 of MITWPL.
Watanabe, Akira. 1996. Nominative-genitive conversion and agreement in Japanese: A cross-
linguistic perspective. Journal of East Asian Linguistics 5:373–410.
Yip, Moira, Joan Maling, and Ray Jackendoff. 1987. Case in tiers. Language 63:217–250.