180 THE ROLE OF THE GENITIVE SUFFIX IN RELATIVE CLAUSES

BLS 5. Linguistic Society of America
180
DOI: http://dx.doi.org/10.3765/bls
*
THE ROLE OF THE GENITIVE SUFFIX
DEDE
TO
REPLY
A
TURKISH:
IN
CLAUSES
IN RELATIVE
Laura Knecht
M.I.T.
1. Introduc tion and Backgrou nd
Turkish is an SOV language with a right-he aded relative clause
construc tion. The rule of relativi zation operates unbound edly leftward, effectin g deletion of the NP in the relative clause (RC) which
is corefere ntial with the head noun.
There are two morphol ogically distinct types of RC's in Turkish.
Compare the form of the RC in (1), where the target is a direct object, with (2), where the target is a subject. The morphem es of
interest are underlin ed. 1
(1) Direct object relativi zed
hali
kadin-in 0 al-dif-i_
buy-PART-POSS rug
woman-GEN
'the rug which the woman bought'
(2) Subject relativi zed
0
kadin
hali-yi al-~
rug-ACC buy-PART woman
'the woman who bought the rug'
In both cases, the verb of the RC appears in a particip ial form; the
2
particip le suffixes -DIK and -(y)En both encode non-futu re tense.
posand
genitive
the
of
The addition al morpholo gy in (1) consists
sessive suffixes . The former is affixed to the subject of the RC,
the latter to the particip le. The possessi ve agrees with the subject
in person and number. (1) exempli fies what I shall call an "object
particip le" (OP) RC; (2) exempli fies a "subject particip le" (SP)
RC. 3 (-DIK is thus an OP suffix and -(y)En a SP suffix.)
The first generati ve account of RC particip les to appear in
the literatu re was that of Underhi ll (1972). He makes the following observa tions. In the simple cases like (1) and (2), the subject of a RC relativi zes with the SP construc tion, non-sub jects
with the OP construc tion. There are two circumst ances, however, in
which relativi zation of somethin g other than the subject of the RC
requires the SP: when the target is (i) a genitive NP attached to
the subject of the RC or (ii) a genitive NP attached to an oblique
object-- provide d that the subject appears in immedia te pre-verb al
position and is interpre ted as indefin ite. Underhi ll notes that in
these cases and in the simple case, the SP is used for relativi zation of a clause- initial target. By assuming that the genitive
suffix is assigned by a transfor mationa l rule that applies sometime
after relativi zation, Underhi ll could claim that the SP is assigned
just when a clause- initial, caseless NP is the relative target.
BLS 5. Linguistic Society of America
181
DOI: http://dx.doi.org/10.3765/bls
Othe rwise , the OP is chose n. The relat
iviza tion trans form ation could
thus be writt en entir ely withi n the frame
work of Stand ard Theo ry.
Hanka mer and Knec ht (1976 )--he ncefo rth
HK-- argue that the linea r
posit ion and casem arkin g of a relat ive
targe t are irrel evan t to particip le choic e. Rath er, the prima ry princ
iple gover ning selec tion
is that the subje ct of the RC relat ivize
s with the SP and non- subje cts
relat ivize with the OP.
HK show that, contr ary to Unde rhill' s claim
, relat iviza tion of
a casem arked , claus e-me dial NP conta ined
in a sente ntial subje ct
requ ires the SP cons truct ion. In fact,
every thing in a sente ntial
subj ect, regar dless of its case or posit
ion, relat ivize s with the
SP and every thing in a sente ntial non- subje
ct relat ivize s with the
OP. HK propo se the Moth er Node Princ iple.
(3) Moth er Node Princ iple (MNP): If a
subc onsti tuent of a majo r
cons titue nt of the RC is relat ivize d, the
parti ciple is
chose n which would be appro priat e for relat
iviza tion of
the majo r cons titue nt itsel f. [That is,
for the simp le
cases , if the moth er node domi natin g the
targe t is the
subje ct of the RC, the SP is chose n; other
wise , the OP
is chose n.] 4
The MNP and Unde rhill' s propo sal make the
same pred ictio ns abou t relativi zatio n of subp arts of phra sal cons
titue nts: the SP is chose n
when a geni tive NP attac hed to the subje
ct of the RC is the targe t;
other wise , the OP is chose n. Both corre
ctly pred ict the OP when
subc onsti tuent s of sente ntial objec ts are
relat ivize d. What distingu ishes the two propo sals is relat iviza
tion of NP's in sente ntial
subje cts. For insta nce, the direc t objec
t in the sente ntial subject of a RC relat ivize s with the SP.
This is in accor dance with
the MNP and in viola tion of Unde rhill' s
prop osal.
HK propo se a secon d princ iple to accou
nt for relat iviza tion of
NP's in impe rsona l passi ves. Breck enrid
ge (1975 ) has argue d that
there is no NP in an impe rsona l passi ve
which funct ions as a subje ct. 5
HK obser ve that no matt er what is relat
ivize d in an impe rsona l passive, only the SP is poss ible. They propo
se the No Subje ct Prin ciple.
(4) No Subje ct Princ iple (NSP) : If there
is no subje ct in the
RC at the time of RC form ation , the OP
cons truct ion is imposs ible and only the SP cons truct ion is
chose n.
Reca ll what Unde rhill disco vered abou t
relati V;i.z ation into a sentence with an inde finit e, imme diate ly pre-v
erbal subje ct: a geni tive
NP attac hed to a non- subje ct relat ivize
s with the SP. HK show that
no matt er what the relat ive targe t is in
such a sente nce, the OP is
impo ssibl e. The princ iple that deter mine
s parti ciple selec tion here
looks susp iciou sly like the NSP. HK claim
that it is in fact the
NSP and that inde finit e-sub ject sente nces
, like impe rsona l passi ves,
are subje ctles s at the time of RC form
ation . They propo se a rule
of Subje ct Demo tion which strip s inde finit
e NP's of their subje cthood .
HK, then, argue that the SP is chose n in
a disju nctio n of cir-
BLS 5. Linguistic Society of America
182
DOI: http://dx.doi.org/10.3765/bls
the subj ect of the RC or part
cums tance s, i.e. , when the targ et is
at the time of RC form ation .
of it or when the RC is subj ectle ss
have the effe ct of requ iring
ose
The two prin ciple s that they prop
relat ions of the leas t deep ly
al
matic
gram
the
at
the gram mar to look
be deter mine d is whet her
must
What
embe dded cons titue nts in the RC.
subj ect relat ion and if so,
any of these cons titue nts bear s the
ect or part of it.
subj
whet her the rela tive targ et is the
2. Dede 's Prop osal
"obs cures what is reall y
Dede (1978 ) claim s that HK' s prop osal
faile d to notic e that
HK
68):
happ ening durin g rela tiviz atio n" (p.
disco very of the
is
t
nmen
assig
le
icip
the cent ral probl em in part
to the subj ect of
xed
is suffi
cond ition s unde r whic h the geni tive
mine d, it is a
deter
is
tive
geni
the
the RC. When assig nmen t of
suff ix and dele
icip
te part
simp le matt er to choo se the appr opria
argu es that the
Dede
e.
essiv
poss
the
cide whet her or not to attac h
a sing le func by
rned
) is gove
Geni tive Suff ix Attac hmen t Rule (GSAR
ned if
assig
is
tive
geni
the
that
tion al prin ciple whic h ensu res
from
RC
the
of
ect
the subj
and only if it is need ed to prev ent
nctio n
disju
a
s,
claim
HK's
to
rary
bein g misc onstr ued. Thus , cont
nmen t
assig
le
icip
part
unt for
of prin ciple s is not requ ired to acco
in Turk ish.
that attac h the part iciDede orde rs the GSAR befo re the rules
tial NP Dele tion
feren
Core
r
afte
and
ple and poss essiv e suffi xes
(p. 69):
te the NP whic h is core fI. Core feren tial NP Dele tion: Dele
.
noun
eren tial with the head
the subj ect, add the GEN-S
II. GSAR: If the dele ted NP is not
ition s for the appl icaCond
to the subj ect of the clau se.
r.
late
n
give
be
will
tion of this rule
(PSAR): If the GSAR apIII. Part icip le Suff ix Attac hmen t Rule
PS -Dik ; othe rwis e,
the
se
choo
plie s, then obli gato rily
~n.
-(y)E
PS
the
se
choo
(Poss -SAR ): If the PSAR
IV. Poss essiv e Suff ix Attac hmen t Rule
a Poss -S to the part iciadd
rily
gato
obli
choo ses -Dik , then
perso n with the subj ect
pial verb whic h agre es in numb er of
of the RC.
acco unt for the choic e of the SP
Noti ce that Dede can auto mati cally
ss at the time of RC form ation
cons truct ion when the RC is subj ectle
well as when the subj ect of
as
ive)
(e.g ., it is an impe rsona l pass
targ et. In both case s, by the
the rela tive cl'au se is the rela tive
there is no subj ect in the RC
y,
appl
time the GSAR has a chan ce to
So, PSAR choo ses the SP sufto whic h the geni tive can be attac hed.
essiv e. This is an attra cposs
the
fix and Poss- SAR does not attac h
tive resu lt.
part of the subj ect of a RC
Howe ver, assig nmen t of the SP when
mati c. When a subc onst ituen t
unde rgoes rela tiviz atio n is not auto
BLS 5. Linguistic Society of America
183
DOI: http://dx.doi.org/10.3765/bls
of the subje ct is delete d, there is somet hing left
over for the GSAR
to attach the geniti ve to; never theles s, the geniti
ve must not be
suffix ed. It was this fact that led HK to propo
se a disjun ction of
princ iples to accoun t for partic iple choice .
Dede claims to be able to explai n why the GSAR
does not apply
in this circum stance ; the explan ation makes refere
nce to the
"func tions" that the geniti ve serves in relati vizat
ion. Dede
states (p. 70):
[The geniti ve] has two impor tant functi ons in relati
vizat ion:
to distin guish the subje ct which remain s in the
RC from the
subje ct of the senten ce in which the RC appea rs;
to preve nt
any change in the functi on and defin ite and indef
inite featur e
(sic) which might be caused by deleti on of the
coref erenti al
NP in the proces s of relati vizat ion. In order
to fulfil l
these two functi ons the GSAR applie s oblig atoril
y
A. to the subjec t of a trans itive verb regard less
of
the defin ite or indef inite featur e of the subje
ct
B. to the defin ite subje ct of an intran sitive verb
and to the indef inite subje ct when it does not
occupy the [imme diate] preve rbal positi on.
In subseq uent sectio ns, the two functi ons mentio
ned here will be examined in some detai l. For now, it suffic es to
say that Dede's
claim appar ently is that the geniti ve suffix picks
out the subje ct
of a RC and preve nts it from being confus ed with
other nomin als, e.g.,
with the subje ct of a highe r clause or with other
NP's in the RC.
Notic e that it is not clear from the above passag
e exactl y what
the relati onshi p is suppo sed to be betwee n the
two functi ons of the
geniti ve and the two condi tions on the applic ation
of the GSAR.
Dede simply says that the GSAR applie s in A and
B "in order to fulfill" the functi ons. This would appea r to rule
out assign ing the
geniti ve in some RC's in accord ance with the condi
tions and in
others in accord ance with the functi ons. So, suppo
se that the genitive is always assign ed in compl iance with condi
tions A and B.
Is the result that the geniti ve is assign ed just
when it is functiona lly requir ed in RC's? Or is the result that
it is assign ed in
some cases where it has a functi on to serve and
in others where it
does not?
It turns out that under condi tions A and B, the
geniti ve will
be suffix ed to the subje ct of a RC when it is not
requir ed to serve
the functi ons which Dede ascrib es to it. Moreo
ver, assign ment in
accord ance with these condi tions leads to the wrong
result in some
cases . Simil arly, an analy sis of geniti ve assign
ment that makes
refere nce to the two functi ons the suffix serves
in RC's can be
shown to be inadeq uate.
3. Condi tions A and B of the GSAR
Under condi tion A, the GSAR incor rectly assign s
the geniti ve
suffix when the targe t of relati vizat ion is in
an indef inite- subje ct
BLS 5. Linguistic Society of America
184
DOI: http://dx.doi.org/10.3765/bls
vizes the subje ct
senten ce. Under condi tion B, it incor rectly geniti
ent of the
nstitu
subco
a
to
d
applie
has
on
vizati
relati
of a RC when
subje ct.
is obliga tory for
3.1 Dede states that under condi tion A, the GSAR
er it is defwheth
of
less
regard
verb,
itive
trans
a
of
ct
the subje
ces:
senten
ing
follow
the
inite or indef inite. Consi der
fsfrd f.
(S)a. Kopek adamfn kfzfn f
bite-P AST
S-ACC
er-POS
daught
N
man-GE
dog
'The dog bit the man's daugh ter.'
b. Adamfn kfzfn f kopek fsfrdf .
'A dog/do gs bit the man's daugh ter.'
(Sa) and (Sb) is
Assume that the targe t of relati vizat ion in both
ct of the transubje
ite
defin
the
vize
adamf n. The GSAR will geniti
subje ct of the
inite
indef
the
and
case
first
the
in
verb
sitive
then attach the
trans itive verb in the second case; the PSAR will
posse ssive.
the
add
will
AR
Poss-S
the
and
verbs
the
to
OP suffix
fact, adamfn
in
But
RC.
OP
an
be
The resul t in both cases should
(Sb).
in
SP
the
and
(Sa)
in
OP
the
with
relati vizes
adam
fsfrdf gf
(6)a. kopeg in kfzfn f
man
S
P-POS
bite-O
S-ACC
dog-GEN daught er-POS
'the man whose daugh ter the dog bit'
b. kfzfn f kopek fsfran adam
SP
'the man whose daugh ter a dog/do gs bit'
in (b):
HK can accoun t for the SP relati vizat ion of adamfn
due to the
tion
forma
RC
of
time
the
at
ctless
the senten ce is subje
the SP must
NSP,
the
to
ding
applic ation of Subje ct Demot ion. Accor
senten ce;
ctless
subje
a
as
(b)
e
analyz
not
does
be assign ed. Dede
subje ct.
inite
indef
the
nothin g preve nts the GSAR from geniti vizing
y to the defi3.2 The GSAR under condi tion B applie s oblig atoril inite subje ct
indef
the
to
and
verb
sitive
nite subje ct of an intran
positi on. This
when it does not appea r in immed iate pre-v erbal
cases.
some
in
t
leads to the wrong resul
ct of (a)
The follow ing senten ces are intran sitive ; the subje
. Both
phrase
ssive
posse
a
is
(b)
of
ct
is a clause while the subje
ite.
defin
are
cts
subje
~upheli.
yedig i
kabag f
(7)a. Yflan fn
doubt ful
SS
RT-PO
eat-PA
snake-GEN squash -ACC
'That the snake ate the squash is doubt ful.'
gitti.
pl~ja
arkada~f
b. Kadfn fn
go-PAST
AT
beach-D
-POSS
woman-GEN friend
'The woman 's friend went to the beach .'
BLS 5. Linguistic Society of America
1S5 DOI: http://dx.doi.org/10.3765/bls
As we know, subparts of the subject of a RC relativize with the SP.
For instance,
(S)a. kabag± yedigi
~iipheli
olan y±lan
SP
'the snake which it's doubtful that (it) ate the squash'
b. arkada~± plaja giden kad±n
SP
'the woman whose friend went to the beach'
But the GSAR under condition B requires that the definite subject
of an intransitive verb be genitivized. As a consequence, OP RC's
are predicted in (Sa) and (Sb).
Finally, consider the intransitive sentence in (9); the subject
NP is indefinite and it does not occupy immediate pre-verbal position.
bir ke~isi
benden ka~t±.
man-GEN a goat-POSS ls-ABL run away-PAST
'One of the man's goats ran away from me.'
(9) Adamin
In accordance with the GSAR, the genitive will be suffixed to the
subject NP when adam±n is relativized; the PSAR and Poss-SAR will
attach the OP suffix and the possessive, respectively. The result,
of course, is an OP RC. However, adam±n relativizes with the SP.
(10) bir
ke~isi
benden
ka~an
adam
SP
'the man, one of whose goats ran away from me'
I conclude that the genitive suffix is not in fact assigned to
the subject of a RC in accordance with conditions A and B. In the
next two sections, I investigate accounts of genitive assignment
that make reference to the "functions" served by the genitive in
RC's. I argue that these accounts are no more successful in predicting the distribution of the genitive in RC's than conditions
A and B are.
4.
The First Function of the Genitive Suffix
Dede states that one of the functions of the genitive is to distinguish the subject of the RC from the subject of the sentence in
which the RC·appears. It is not clear what this means; I therefore
assume that Dede has sentences like the following in mind. The subject of the RC is marked genitive and the higher subject is caseless.
(11) Mustafa Ersin'in optugii
k±z±
tan±yor.
GEN kiss-OP-POSS girl-ACC know-PRES
'M knows the girl who E kissed.'
BLS 5. Linguistic Society of America
186
DOI: http://dx.doi.org/10.3765/bls
Here it looks like the genitive suffix helps to sort out the clausemembership of subject NP's by establishing a morphological distinction between them.
But Dede's description of this "function"' is hard to interpret
as an empirical claim about assignment of the genitive. In this
section, I attempt to extract an empirical claim from the description. A number of possibilities are considered and each is shown
to be wrong.
4.1
The first claim to be investigated is given in (12).
(12) The genitive is assigned to the subject of a RC
if and only if suffixation would morphologically
distinguish this subject NP from the subject of
the clause which contains the RC.
For an immediate counterexample to (12), consider (13).
adam4::
Mahmut'un tanfdfgf
(13) Rahmi kfzfn
GEN know-OP-POSS man-ACC
girl-GEN
soyledi.
sevdigini
like-PART-POSS-ACC say-PAST
'R said that the girl likes the man who M knows.'
The subject of the RC (Mahmut) is marked genitive--and so is the subject of the clause which contains the RC (kfzfn). No speaker of
Turkish is confused about the clause-membership of these two NP's.
(13) clearly shows that the genitive is assigned to a RC subject
when the function of morphologically distinguishing it from the
next higher subject is not served by assigning it the genitive.
4.2 Let us restate the claim. Perhaps Dede intends the genitive
to be understood as the morpheme which signals that the NP to which
·
it is attached is an embedded subject.
(14) The genitive is assigned to the subject of a RC if and
only if suffixation would morphologically distinguish
this subject NP from the subject of the matrix clause.
Now (13) is no problem: genitivizing Mahmut does indeed morphologically distinguish it from Rahmi. But (14) makes the wrong prediction about the following sentence.
$iipheli olan
(15) Ahmet Murat'fn optligii
GEN kiss-PART-POSS doubtful be-SP
gordu.
kfzf
girl-ACC see-PAST
'A saw the girl who that M kissed (her) is doubtful.'
BLS 5. Linguistic Society of America
187
DOI: http://dx.doi.org/10.3765/bls
The subje ct of the RC (unde rlined ) is sente ntial;
one of its subco nstitue nts has been relati vized --not with the OP
as (14) predi cts,
but with the SP. That is, the geniti ve should
have been assign ed
to Murat'~n optligi i to distin guish it from Ahmet
, the subje ct of the
matrix . (15) shows that there are circum stance
s in which the genitive is not assign ed even though it has a functi
on to serve . 6
Perhap s the senten ce in (15) has some specia l
featur e that makes
it unnec essary to assign the geniti ve to the subje
ct of the RC.
Notice that the clause -memb ership of the two subje
ct NP's is readil y
determ inable : the verb gar- ('see' ) is not subca
tegori zed for sententia l subje cts so its subjec t must be Ahmet ;
on the other hand,
the predic ate ~upheli ol- ('be doubtful 1)~incompat
ible with
Ahmet so its subje ct must be the sente ntial subje
ct. Let us append
to (14) the follow ing statem ent: the geniti ve
is not assign ed if
the subca tegori zation featur es of the verbs make
it possib le to
determ ine what their subjec ts are.
But this will not work eithe r. Consi der (16) below
; a genitive NP attach ed to the subje ct of the RC has been
relati vized with
the SP const ructio n.
(16) Demet
karde~i
ka~an
kfzf
gordii .
sibling -POSS run away-S P girl-A CC see-PA ST
'Demet saw the girl whose siblin g ran away. '
The subca tegori zation featur es of the two verbs
tell us nothin g:
both ka~- ('run away') and gar- are compa tible
with Demet and
karde~i.
The geniti ve should have been suffix ed to the
subje ct
of the RC to morph ologic ally distin guish it from
the matrix subje ct.
But an OP RC in (16) is impos sible.
4.3 Final ly, consid er the follow ing senten ce.
is an imper sonal passiv e; embed ded in it is an
The matrix clause
OP RC.
(17) Sinan 'fn bina ettig i
camiye
gidili rdi.
GEN build-O P-POS S mosque-DAT go-PASS-AOR-PAST
lit: 'To the mosque that Sinan built used to be
gone. '
The geniti vizati on of Sinan is myste rious: the
matrix clause does
not contai n a subje ct, so what is there for Sinan
to be distin guish ed
from?
In sum, if we take Dede's descr iption of the "disti
nguis hing"
functi on of the geniti ve seriou sly, that descr
iption fails as an
accou nt of the distri butio n of the geniti ve in
RC's. In any event ,
Dede does not put this functi on to much use in
her paper ; most of
the burden of explan ation falls on the second
functi on which the
geniti ve is claime d to serve. Hence I will say
no more about the
"disti nguis hing" functi on and procee d to discus
s the other functi on.
5.
The Second Functi on of the Genit ive Suffix
The second functi on which Dede claims the geniti
ve serves in RC's
BLS 5. Linguistic Society of America
188
DOI: http://dx.doi.org/10.3765/bls
the relat ional statu s or the
is that of preve nting misco nstru al of
ct of the RC. Actua lly,
defin itene ss of the NP which is the subje
no reaso n to think
is
There
two.
but
ion,
this is not one funct
relati ons is relaatical
that preve ntion of misco nstru al of gramm
ss.
itene
defin
of
al
nstru
misco
ted to preve ntion of
Trans itive Relat ive Claus es
a trans itive RC is susce p5.1.1 Dede claim s that the subje ct of
neith er subje ct NP's
that
fact
tible to misco nstru al due to the
in Turki sh.
arked
casem
are
ts
objec
t
direc
nor indef inite
5.1
(18)
kopek koval iyor.
chase- PRES
child dog
'The child is chasi ng dogs. '
~ocuk
be casem arked accus ative .
Actu ally, indef inite direc t objec ts may
.
Consi der (19) below (and Dede' s (11))
ariyor um.
(19) Ben bir adami
a man-ACC look for-PR ES-ls
ls
'I am looki ng for a man.'
in (20) below , where the direc t
Here I have a parti cular man in mind;
not.
objec t is casel ess, I do
(20) Ben bir adam ariyor um.
be casem arked accus ative
It is not indef inite NP's which fail to
but NP's which are nonwhen they funct ion as direc t objec ts,
speci fic in refere nce.
restr icted to appea ring
Notic e that casel ess direc t objec ts are
ion.
posit
in imme diate pre-v erbal
(21) Ismai l { diin
} kitap *{dun
1
} okudu .
bah~ede
bah~ede
n.'
Ismai l read books yeste rday/ in the garde
Casem arked direc t objec ts are not.
{:~~~ede}
(22) Ismai l kitab i
okudu .
ACC
rday/ in the garde n.
yeste
1 Ismai l read the book
1
prope r nouns (Ahm et),
Certa in NP 1 s are inher ently speci fic:
i: Ahme t's cat),
kedis
et'in
(Ahm
NP's
ssed
posse
I),
prono uns (ben:
that cat).
kedi:
(o
ives
and NP's which occur with demo nstrat
ion as direc t
funct
they
when
ative
accus
arked
casem
Such NP's must be
objec ts. 7
(23) Ahmet Pinar i/*Pi nar koval iyor.
'Ahme t is chasi ng Pinar .'
BLS 5. Linguistic Society of America
189
DOI: http://dx.doi.org/10.3765/bls
5.1.2
When the direct object of the following sentence undergoes
relativiz ation,
(24) Cocuk kizi
seviyor.
child girl-ACC like-PRES
'The child likes the girl.'
one of the side-effe cts of deletion is that the subject NP appears
in immediate pre-verba l position.
(25)
[~ocuk
sev-] [kiz]
According to Dede, ~ocuk runs the risk of being interpret ed as the
non-spec ific direct object of sev-, not as its subject. Such misconstrual is not permitted in accordanc e with "the generally accepted
rule that transform ations should not decrease or change meaning."
(fn. 4) To block the object reading of ~ocuk, the genitive suffix
is attached to it. Why does the genitive succeed in unambiguo usly
indentify ing ~ocuk as the subject of the RC? The reason is that,
in general, only subject NP's in Turkish are genitiviz ed. 8
Suffixati on of the genitive is accompani ed by assignmen t of the
OP suffix and the possessiv e. The result is an OP RC.
(26) ~ocugun
sevdigi
kiz
child-GEN like-OP-PO SS girl
'the girl who the child likes'
Now, it is not at all clear why Dede claims that ~ocuk in (25)
would certainly be misconstr ued as a direct object were the genitive
not suffixed to it. It is more likely that the string would be ambiguous, with ~ocuk interpret ed either as the subject of the RC or
as the direct object. Dede's proposal could have been that such
ambiguity is not permitted to arise in Turkish. 9
In any case, compare (24) with the following sentence.
(27) Adamin
kizini
kopek kovaliyor .
man-GEN girl-POSS- ACC dog
chase-PRE S
'A dog/dogs is/are chasing the man's daughter .'
A subpart of the direct object, i.e., adamin, is relativiz ed.
(28) [kizini kopek kovala-] [adam]
As in the previous example, the subject of the RC is
in immediate pre-verba l position. But unlike the previous example, this
RC contains an accusativ e casemarke d NP (kizini). Kopek, then,
cannot be misconstr ued as the direct object of the clause and there
is no need for it to be genitiviz ed; (28) surfaces as a SP RC.
(29) kizini kopek kovalayan adam
SP
'the man whose daughter a dog/dogs is/are chasing'
BLS 5. Linguistic Society of America
190
DOI: http://dx.doi.org/10.3765/bls
Dede comments:
It follows that the process of relativiz ation and the assignment of the GEN case are based on whether or not it is possible
to identify the subject and the DO. In other words, the GSAR
operates whenever the relations of the subject and the DO are
not indicated by means such as other case suffixes, word
ordera semantic propertie s of the members of the sentence,
etc. 1 (p. 73)
In conclusio n, Dede's claim is that the function of the genitive
in transitiv e RC's is to prevent misconstr ual of the grammatic al relations of the subject NP. Crucially , the genitive is assigned as
a last resort: the GSAR operates if and only if everythin g else
has failed to indicate the relations of the subject.
5.1.3 In fact, the genitive is assigned when other casemarke rs, word
order, semantic propertie s, etc. make it possible to identify the
subject of the RC with ease.
Consider the following post-dele tion form:
(30)
kovala-] [kfz]
evde
girl
child house-LOC chase
[~ocuk
Although ~ocuk is not casemarke d, its position in the RC unambiguo usly identifie s it as a subject. (Recall that unmarked direct objects
must occupy immediate pre-verba l position. ) The genitive suffix thus
has no function to serve in the RC and the GSAR should not apply.
Neverthe less it must. 11
(31)
k~z
kovaladfg f
evde
child-GEN house-LOC chase-OP-P OSS girl
'the girl who the child chased in the house'
~ocugun
The underline d NP's in the following post-dele tion RC's also
admit of only one interpret ation: they must be subjects.
(32)a. [P~nar kovala-] [kfz]
b. [o kopek kovala-] [k~z]
If these inherentl y specific NP's were direct objects, they would
have to be marked accusativ e. Again, the genitive is not required
to prevent the subjects from being misconstr ued. But the GSAR must
in fact apply obligator ily.
(33)a. Pfnarfn kovaladfg~ kfz
b. o kopegin kovaladfg f kfz
'the girl who
P~nar/
that dog chased'
191
BLS 5. Linguistic Society of America
DOI: http://dx.doi.org/10.3765/bls
And consid er the follow ing interm ediate form; the subjec
t of
a senten tial object (under lined) has underg one relativ
e deletio n.
(34) [sen <;:it;:ek leri
yedigi ni
soyle- ] [kei;:i]
2s flower- PLU-AC C eat-PART-POSS-ACC say
goat
There are no less than three reason s why the subjec
t of the RC, the
pronou n sen, could not be interp reted as an object :
(i) pronom inal
object s cannot be unmark ed for case, and sen is; (ii)
sen does not
occupy immed iate pre-ve rbal positio n; and (iii) the
relativ e clause
alread y contai ns a direct object (i.e., the accusa tive
casema rked
senten tial object ). Thus (34) should surfac e as a
geniti veless ,
SP RC. This expect ation is not borne out.
(35) senin <;:ii;:ek leri yedigi ni soyled igini kei;:i
GEN
OP-POSS
'the goat which you said ate the flower s'
Finall y, if seman tic proper ties (and extra- lingui stic
knowle dge)
enter into the decisio n to assign the geniti ve, the
GSAR will not
apply in (36).
(36) [adam yaz-] [mektu p]
man write letter
Men write letter s; letter s cannot write men. A SP
RC is predic ted
becaus e the seman tic proper ties of yaz- pick out adam
as its subjec t.
(37) adam yazan mektup
SP
But if (37) is gramm atical at all, it only has the
anomol ous readin g:
'the letter which wrote men.' Speake rs cannot interp
ret adam as the
subjec t even though that is the only sensib le interp
retatio n of it.
5.2
Intran sitive Relati ve Clause s
The functio n of the genitiv e in intran sitive RC's is,
accord ing
to Dede, somewh at differ ent from its functio n in transi
tive RC's.
Since intran sitive verbs are not subcat egoriz ed for
direct object s,
there can be no questi on of mistak ing a subjec t for
an unmark ed object. Rather , Dede says that the problem in intran
sitive RC's is
mainta ining the defini teness of the subjec t NP.
5.2.1 In the unmarked case, the defini te subjec t of
an intran sitive verb is clause -initia l. An indefi nite subjec t
appear s in immediat e pre-ve rbal positio n. (Dede' s (30a) and,(3 la))
(38)
~ocuk
odada
uyuyor .
child room-LOC sleep-P RES
'The child is sleepin g in the room.'
192
BLS 5. Linguistic Society of America
DOI: http://dx.doi.org/10.3765/bls
(39) Odada 9ocuk uyuyor ,
'A child/ childr en is/are sleepin g in the room,'
12
positio ning
Relati ve deletio n of odada in (38) has the effect of
this "chang es
that
claims
Dede
verb,
the
before
iately
,:ocuk immed
ore,
Theref
NP,
nite
indefi
an
the subjec t from a defini te NP to
t." 13 (p. 75)
subjec
the
of
ties
proper
the
retain
to
needed
is
the GSAR
does requir e
In fact, relativ izatio n of the obliqu e object in (38)
in (39)
NP
same
the
of
n
izatio
relativ
while
,
uction
the OP constr
requir es the SP constr uction ,
oda
(40) 9ocugu n uyudug u
OP-POSS
GEN
'the room which the child is sleepi ng in'
(41) ,:ocuk uyuyan oda
SP
'the room which a child/ childr en is/are sleepin g in'
Consid er also the follow ing SP RC,
kapf
akan
su
(42) altfnd an
underne ath-PO SS-ABL water flow-S P door
'the door that water is flowin g out from under'
rbal positi on,
Althou gh the subjec t, su, appear s in immed iate pre-ve
te NP's
defini
only
nite:
indefi
is
Su
ve.
geniti
it is not marked
vized,
geniti
are
nite
indefi
as
in danger of being miscon strued
suffix ation
Dede clearl y intend s the follow ing: there will be
RC if and
sitive
intran
an
of
t
of the geniti ve to the defini te subjec
The
strued,
miscon
be
could
t
subjec
the
of
teness
defini
only if the
canverb
sitive
intran
an
of
t
defini teness of the senten tial subjec
the genitiv e is
not be miscon strued under any circum stance , Thus,
when a subpar t
t"
subjec
the
of
not needed "to retain the proper ties
ts for SP relaccoun
This
ized.
relativ
is
t
subjec
tial
of the senten
14
ativiz ation of yflan below,
~upheli olan yf lan
yedigi
(43) kabagf
doubtf ul be-SP snake
T-POSS
eat-PAR
ACC
squashsquash '
'the snake which it's doubtf ul that (it) ate the
te subjec ts of
5.2.2 The problem is that in many cases, the defini
defini teness
their
though
even
ized
genitiv
be
intran sitive RC's must
ized.
would have been readily appare nt were they not genitiv
the efConsid er an intran sitive RC where deletio n has not had
before the verb,
fect of positio ning a defini te subjec t immed iately
uyu-] [oda]
ak~am
(44) [,:ocuk kedi ile diin
sleep room
g
evenin
ay
yesterd
child cat with
~ocuk
fact that
must be suffix ed with the geniti ve in spite of the
193
BLS 5. Linguistic Society of America
DOI: http://dx.doi.org/10.3765/bls
it is imposs ible to miscon strue it as indefin ite,
(45)
~ocugun
kedi ile dun ak~am uyudugu
oda
GEN
OP-POSS
'the room in which the child slept last evening with the cat'
Additio nally, there are subject NP's which will be interpr
eted
as definit e regardl ess of their positio n relativ e to the verb.
Nevertheless , they must be genitiv ized,
( 46) a. Demet 1 in uyudugu
oda
GEN
OP-POSS
b, benim karde~imin
uyudugu
oda
ls-GEN sibling- POSS-G EN
OP-POSS
'the room where Demet/my sibling slept'
The only out in the face of these example s would be to say
that the
"defini teness recover y princip le" is grarnma ticized: the genitiv
e
is suffixe d to the definit e subject s of intrans itive verbs
regardl ess
of whethe r there could be miscon strual of their definit eness,
However, if the "defini teness recover y princip le" is gramma ticized
and
the genitiv e can be assigne d when other factors suffice to
prevent
miscon strual, then the rule will attach the genitiv e (incorr
ectly)
to the senten tial subject in (43),
6, Conclus ion
Recall that the GSAR is suppose d to apply obliga torily:
A. to the subject of a transit ive verb
B. to the definit e subject of an intrans itive verb and to
an
indefin ite subject when it does not occupy immedi ate preverbal positio n.
Also, the princip al functio n of the genitiv e in RC's is claimed
to
be prevent ion of miscon strual of the gramma tical relatio ns
or definitene ss of the subject of the RC. The genitiv e should be
suffixe d
if and only if there could be miscon strual. I shall refer
to this
as the 'functi onal princip le' of genitiv e assignm ent.
We have observe d that when the genitiv e is assigne d in accordance with conditi ons A and B, a number of false predict ions
are
made. Also, when the genitiv e is assigne d in accorda nce with
the
functio nal princip le, false predict ions are made. Interes
tingly,
the conditi ons and the functio nal princip le make differe nt
predictions about the distrib ution of the genitiv e. This is curious
in
the light of Dede's stateme nt that the GSAR applies when it
does
(i.e., in A and B) "in order to fulfill " the functio ns of
the genitive.
For instanc e, in accorda nce with conditi on A of the GSAR and
in violati on of the functio nal princip le, the subject s of
the follow-
194
BLS 5. Linguistic Society of America
DOI: http://dx.doi.org/10.3765/bls
ing RC's are genit ivize d:
(31) 9ocug un evde koval adfgf kfz
(33a) Pfnar fn koval adfgf kfz
tion of the funct ional
In accor dance with condi tion B and in viola
cts of the follow ing
subje
the
to
ed
suffix
is
ive
princ iple, the genit
RC's.
(45) 9ocug un kedi ile dun ak~am uyudu gu oda
(46a) Deme t'in uyudu gu oda
genit ive is in accor d
For the follow ing RC , the absen ce of the
GSAR incor rectly
the
of
A
ition
with the funct ional princ iple. Cond
here.
ive
assig ns the genit
(6b) kfzfn f kopek fsfra n adam
the SP RC's below .
The funct ional princ iple also accou nts for
OP RC's.
been
have
cond ition B of the GSAR, these shoul d
Given
(Sa) kabag f yedig i ~iipheli olan yflan
(9) bir ke9is i bende n ka9an adam
:
The predi ction s conve rge for these two cases
(26) 9ocug un sevdi gi kfz
(42) altfnd an su akan kapf
just the GSAR with
In sum, a gramm ar of Turki sh which conta ins
conta ins just the
which
ar
gramm
A
.
quate
condi tions A and B is inade
which conta ins
ar
gramm
A
funct ional princ iple is also inade quate .
r of cases .
numbe
large
a
in
s
ction
both makes contr adict ory predi
which of the proing
decid
for
basis
ipled
princ
no
be
to
There seems
even suppo sing that some
cedur es to follow for a given RC. Moreo ver,
Dede' s accou nt canno t avoid
such decis ion proce dure could be devis ed,
in terms of a disju nctio n
s
ciple
parti
RC
statin g the distri butio n of
of princ iples .
Footn otes
and Jonat han Press ler
am indeb ted to Jorge Hanka mer, Phil LeSou rd,
for many helpf ul sugge stion s.
*I
ablati ve
used in the gloss es are the follow ing: ABLFUTcase;
dative
DAT;
tense
aorist
case; ACC- accus ative case; AORPART -parti ciple
case;
ve
locati
LOCcase;
ve
geniti
GEN;
futur e tense
; POSS -posse ssive; PRESsuffi x; PASS -passi ve suffi x; PAST -past tense
1 Abbr eviati ons
BLS 5. Linguistic Society of America
195
DOI: http://dx.doi.org/10.3765/bls
present tense; l,2,3s/ l,2,3p -first, second, third person singula
r
or plural.
2 Upperca se letters
represe nt segmen ts which have a number of phoneti c
realiza tions due to the operati on of vowel harmony , final stop
devoicing , and consona nt assimil ation.
3 These terms are
Underh ill's (1972). Hankam er (1973) referre d to
(1) as a "posses sed partici ple" RC and to (2) as a "free partici
ple" RC. Hankam er's termino logy has the virtue of focusin g
attention on the crucial differe nce between the two RC constru ctions:
whethe r or not the genitiv e and the agreein g possess ive are
suffixed to the subject of the RC and the partici ple, respect
ively.
In the non-fu ture tense, the partici ple suffixe s have differe
nt
realiza tions; in the future tense, howeve r, the morpheme -EcEK
does double duty as the OP suffix and the SP suffix.
(a) kadfn-t n al-acag -f
halt
woman-GEN buy-PART-POSS rug
'the rug which the woman will buy'
(b) halfyf al-acak kadtn
rug-ACC buy-PART woman
'the woman who will buy the rug'
Suffixa tion of the genitiv e and the possess ive is the only
thing
that disting uishes (a) from (b).
Unfortu nately, the terms 'posses sed partici ple' and 'free participle' are not as well establi shed in the literat ure as OP
and SP.
For this reason, I have adopted Underh ill's termino logy.
4 In Dede's review
of HK 1 s proposa ls (which is discuss ed beginni ng
in section 2), the followi ng remark about the MNP appears :
"Although this princip le seems to work within the framewo rk of
transformati onal gramma r, it is not necessa ry to include such a
princip le
in the grammar of Turkish ••• " The major point of HK's paper
was
that the princip les governi ng partici ple choice in Turkish
make reference to notions like "subjec t" and "non-su bject" and thus
cannot
be stated in the framewo rk of a transfo rmation al gramma r.
5 Perlmu tter (1978)
and Perlmu tter and Postal (forthco ming) claim that
there is an NP in imperso nal passive s that functio ns as subjec
t, but
that it is an inaudib le dummy. This distinc tion is not relevan
t for
my purpose s here.
6 (15) does not exempl
ify the only circums tance in which an embedde d
subject NP fails to be casemar ked genitiv e. The subject s
of clauses
embedde d under verbs like san- ('think ') and bil- ('belie ve')
may be
uncasem arked:
(a) Cengiz Yakut erken kalktt
sandt.
early rise-PA ST think-PA ST
BLS 5. Linguistic Society of America
1
196
DOI: http://dx.doi.org/10.3765/bls
Cengiz though t that Yakut got up early .'
not suffix ed with
And the subje cts of certai n adver bial clause s are
the genit ive:
gider.
(b) Murat gelinc e Yasem in lokan taya
restau rant-D AT go-AOR
come-ADV
rant.'
'When Murat comes , Yasem in will go to the restau
cts be marked geniThus, there is no requir ement that embed ded subje
tive in Turki sh.
speci ficity , are
7 Addit ionall y, obliqu e objec ts, regard less of their
ve.
ablati
or
ve,
locati
,
dative
obliga torily casem arked
ve in, for examp le,
8 1 am of course ignori ng the assign ment of the geniti
posse ssive phras es.
nce of this type
Jorge Hankamer has remind ed me, ambig uity avoida
nce of relaexiste
the
out
points
(1977)
skey
is not unive rsal. Mcclo
be inter may
which
NP
an
tive clause s in Modern Irish which conta in
t.
objec
t
direc
a
as
or
t
preted eithe r as a subjec
9 As
role played
presen ts the follow ing senten ce to illust rate the
of the
dge
knowle
by seman tic prope rties and the extra -lingu istic
s.
applie
GSAR
the
not
or
er
speak er in determ ining wheth
10 Dede
kopek tstrd t.
bite-P AST
dog
'A dog/do gs bit the girl. '
(16) ktzt
girl~Acc
is relati vized , the
The claim is that when the direc t objec t (ktzt)
ve to preve nt its
geniti
d
marke
be
to
have
not
does
)
(kopek
subje ct
is that dogs usuall y
reason
The
t.
objec
t
being misco nstrue d as a direc
SP RC below is
the
then,
Dede
For
.
versa
vice
not
and
bite girls
gs bit.'
dog/do
a
who
girl
'the
gramm atical and has the readin g,
(18) kopek
fsfran ktz
SP
(18) is margi nal
For all the speak ers of Turkis h I have consu lted,
is, 'the girl who
it
of
on
retati
interp
le
possib
at best and the only
bit dogs. '
tempti ng to
respe ct to examp les (31) and (33), it might be
the
assign
to
fails
iple
princ
onal
functi
claim that althou gh the
be geniti vized
event
any
in
will
RC's
the
of
cts
subje
the
geniti ve,
5.
becau se they are embed ded. Not so; see footno te
1 ' 1 With
Const iin (39) is also interp retab le as a defin ite NP.
on for empha sis;
positi
erbal
pre-v
iate
immed
into
moved
are
tuents
12 9ocuk
BLS 5. Linguistic Society of America
197
DOI: http://dx.doi.org/10.3765/bls
on the defi nite read ing, 9-0Cuk is
ther efor e 'focu sed. '.
13 Dede does
not expl ain why the deri ved posi tion
of the subj ect shou ld
have the effe ct of "cha ngin g it" from
a defi nite NP to an inde fini te
NP.
1 "Ded e's actu
al disc ussi on of ( 43) is as follo ws:
" ••• the GSAR is
not need ed here beca use the dele tion
of the NP in the sent enti al subject does not caus e any chan ge in
the gram mati cal rela tion of the
sent enti al subj ect to the main verb
~upheli 'dou btfu l' and its
rela tion is indi cate d by the abse nce
of a case suff ix. That is,
a sent enti al NP whic h is in the nom
inati ve case can only func tion
as the subj ect. " (p. 74) It is not
true that unca sema rked sent ential NP's must be subj ects ; see foot
note 5. Furt herm ore, this
disc ussi on is conf usin g in ligh t of
Dede 's anal ysis of othe r intran sitiv e RC's , wher e misc onst rual
of defi nite ness , not gram mati cal rela tion s, was said to be
the prob lem the grammar face s.
Refe renc es
Brec kenr idge , Jane t (197 5) Rule s Whic
h Noth ing Unde rgoe s: An Inve stiga tion of Impe rson al Pass ives and
Obje ct Rais ing Con struc tions
in German, unpu blish ed unde rgra duat
e thes is, Harv ard Univ ersit y.
Dede , Mii~erref (197 8) "Why Shou ld
Turk ish Rela tiviz atio n Dist ingu ish
Betw een Subj ect and Non -Sub ject Head
Noun s?" BLS, Volume 4.
Hank amer , Jorg e (197 3) "Una ccep table
Amb iguit y," LI, 4.1.
Hank amer , J. and Laur a Knec ht (197
6) "The Role of the Subj ect/N onSubj ect Dist inct ion in Dete rmin ing
the Choi ce of Rela tive Clau se
Part icip le in Turk ish," in Aiss en
and Hank amer , eds. , Harv ard
Stud ies in Synt ax and Sem antic s, Volum
e II.
McC loske y, Jame s (197 7) "An Acce ptab
le Amb iguit y in Modern Iris h,"
LI, 8.3.
Perl mut ter, Davi d (197 8) "Imp erso nal
Pass ives and the Unac cusa tive
Hyp othe sis," BLS, Volume 4.
Perl mut ter, D. and Paul Post al (fort
hcom ing) "Imp erso nal Pass ives
and Some Rela tion al Law s."
Und erhi ll, Robe rt (197 2) "Tur kish
Part icip les, " LI, 3.1.