ABSTRACT

Title of Dissertation: MINIMALITY AND TURKISH RELATIVE
CLAUSES

Ilhan Merih Cagri, Doctor of Philosophy, 2005

Directed By: Professor Norbert Hornstein
Department of Linguistics


Turkish relative clauses display a subject/non-subject asymmetry. The subject
relative (SR) is licensed for relativization from [Spec, TP]. Whereas the non-subject
relative (NSR) is never acceptable for subject relativization, the SR is licensed in
clauses where there is no external argument, and when relativizing a non-subject in
clauses where the subject is non-specific. Within the framework of the Minimalist
Program, Turkish RCs are explained in terms of satisfaction of the EPP of T by a D
feature and Minimality effects. As long as no nominal expression intervenes between
the relative head and [Spec, TP], the SR is licensed. The SR, then, can be used as a
diagnostic for movement through TP. Minimality effects are incurred when there is
an intervening nominal between T° and the RC head, and the SR becomes
unacceptable. The proposal is that in Turkish, specific nominals, +human nominals,
and Experiencers of psych verbs all contain a DP projection. Non-specifics are NPs
which cannot satisfy the EPP. NP subjects cannot move to [Spec, TP], and thus

permit the SR form for relativization of non-subjects. NPs create intervention effects,
as does PRO, with the exception of subject control PRO which is perhaps a trace of
movement. Scrambling ameliorates intervention effects. Once scrambled,
expressions are frozen but remain porous for movement of a subconstituent.
Differences between inherent and structural Case are suggested with structural case
assignment limited to DPs and in a Spec-Head configuration. Structurally case-
marked DPs are barred from moving to case-assigning positions unless there is a
morphological match. Further proposals include structures for verb classes, including
Psych verbs, and structures for infinitivals and +human DPs. Contrastive focus is
briefly addressed. Though superficially complex, relativization in Turkish can be
accounted for with a minimum of technology. The suggestions here have
implications for the theory of the EPP, Case, its assignment and interface conditions,
feature satisfaction, and movement.






















MINIMALITY AND TURKISH RELATIVE CLAUSES



By


Ilhan Merih Cagri





Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park, in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2005










Advisory Committee:
Professor Norbert Hornstein, Chair
Professor Howard Lasnik
Professor Paul Pietroski
Professor Amy Weinberg
Professor Christopher Cherniak


UMI Number: 3202436
3202436
2006
Copyright 2005 by
Cagri, Ilhan Merih
UMI Microform
Copyright
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, MI 48106-1346
All rights reserved.
by ProQuest Information and Learning Company.





















© Copyright by
Ilhan Merih Cagri
2005















ii

Dedication





In memory of my parents.
Dedicated to Amina and Nadia.

iii

Acknowledgements
I have thought about what I would write in the acknowledgment section of my
dissertation long before I wrote the dissertation--sort of like planning where to hang
the pictures before you construct the wall. However, now that I am actually at this
juncture, I find myself at a loss for words. Before I begin acknowledging a few of the
people I feel grateful to, I must first thank God, the Creator, for blessing me with so
many good things in life and for allowing me to experience and succeed in this
wonderfully fulfilling endeavor.
This work is a testament to the strength and talent of my parents. My parents
were immigrants of the old school. They were industrious, independent, humble,
courageous and generous. I consider the noble character of my father and the
adventurous spirit of my mother gifts that I hope will be passed on to their
grandchildren and great-grandchildren. My mother did not live to see the completion
of my studies, but not a day went by when she wouldn’t ask, “So, my daughter, when
are we going to see your doctorate?” My mother was one to shamelessly brag about
her children. All aunts and uncles, all cousins, knew we were the special ones in the
family. Colleagues, friends, neighbors, taxi drivers, and store clerks were all told
how exceptional and how successful we were. I exaggerate not. To this day, I run
into people who knew about my mother’s children and grandchildren. People like
taxi drivers in Washington, DC, or the movie theater manager in Bethesda, MD, or
the hair dresser in Columbia, have asked me in detail about me, my siblings and our
kids. So, it would be remiss of me if I did not brag about my mother here. My
mother was born to Jewish parents in Ottoman Turkey. In two years of schooling,

iv

she was already in the fourth grade, at which point her father withdrew her from
school so that she could work in his laundry. My mother was engaged to a nice
Jewish boy by the time she was 17 but decided instead to run off with a very
handsome (“just like Errol Flynn!”) Muslim Turk 15 years her senior. Well, of
course, scandal ensued. The fiance was immediately married off to my mother’s
young aunt, my mother’s sister (believing the lament that the family was disgraced
and the other girls would never marry) ran off to Israel, and my mother was banished
from the family. No one in the family was permitted to speak to my mother. This
didn’t stop my grandfather from using her labor at his shop, however, and it was from
the shop doorway that my mother watched her brother’s Bar Mitzvah procession on
its way to the reception from which she was excluded. My mother and father moved
to Ankara shortly thereafter. Of my mother’s exploits there, I know only that she
enrolled in a school of French cuisine (she spoke French), worked as a nanny to an
American diplomatic family (she had learned English), worked at the Mexican
Embassy (she spoke Spanish), and entertained friends of different backgrounds
(smattering of Greek and Italian). My mother embroidered and knit beautifully, loved
poetry and art, and appreciated all kinds of music, Turkish, European, classic and
current. In the U.S., my mother was as active and as creative. Working in various
positions in many different embassies in Washington DC, my mother made our house
the scene of glittering soirees. On one occasion, I remember diplomats gushing over
the greasy jack my mother had found on the side of a road and placed on our
credenza. Needless to say, my father thought she was nuts. My mother lived every
day of her life. Not one moment was wasted. She hardly slept. And in her

v

characteristic way, she worked full-time until three days before her death at age
eighty. At her funeral, a Jewish cantor sang, Muslims, Christians, and Jews wearing
Yarmulkas stood side-by-side as they said the Muslim Jenazeh prayer, and Rodrigo’s
Concierto de Aranjuez blared loudly as she was interred by her family.
I thank my father for teaching me about strength and courage. My father was
a peaceful man of dignity and integrity. He hated to inconvenience anyone. Many
will understand when I say that my father loved his family so much that his eyes
teared as he gazed upon us. My father was a fixer. He taught me how to examine
things carefully and to reason things through. Syntax is about problem solving, and
this skill I learned from my father.
To my family, I owe my confidence. My aunts, Ida Dana, Meri Baruh, and
Mati Revah, and uncles, Nesim Revah and Leon Dana, gave me healthy doses of love
and encouragement. Their mantra was: “You can do anything!” I thank my sisters,
Çaya and Beyhan and my brother Kemal for their love and support. I thank my
brother-in-laws, Herb Conger and Bruce Trock for their encouragement. My children
watched in awe and suffered with tolerance as I made this long journey. It was by
loving them, and being loved by them, that I was able to keep my sanity. My
daughter, Homeyra, was my standard bearer and shield. I cannot express how proud I
am of her abilities and accomplishments. I am grateful to Zeki and Yahya for their
forbearance, and for requiring so little of me being during this period. Idris is owed a
special note of gratitude for maintaining his good-natured spirit in spite of quotidian
bouts of poverty, chaos, and stress. No matter how unpredictable the disaster, Idris
was there and willing to fix the problem. At different times, one or more of my

vi

children would engage in discussions related to an area of interest to me. On one
such occasion, I was inspired enough to produce a whole chapter of this thesis. I,
therefore, dedicate Chapter 7 on Infinitivals to Haroon. My two daughters-in-law,
Mariam and Fatimeh, spent many a day wondering just how mad their mother-in-law
was, but to their credit, they never waivered in their support of me nor stinted in their
kindness and generosity. (Conversation with Fatimeh: ME: I’m really sorry about the
appalling way I acted yesterday. You must think I’m crazy! FATIMEH: (weakly)
noooo.... not really....)
I have countless friends who carried me through this ordeal. My degree
belongs to the sisterhood of these friends who were happy for me, prayed for me,
encouraged me and uplifted me. This dissertation is in honor of the mothers and
wives among my friends, to their collective wisdom and their often selfless
aspirations.
I consider it an act of God’s good Grace that I ended up in the Linguistics
Department at UMCP. The professors here are truly dedicated to “teaching”. As
every student knows, it is not enough for an instructor to have made a name for
himself in the field. I was fortunate to have been able to study syntax with Norbert
Hornstein, Juan Uriagereka, David Lightfoot and Howard Lasnik. These scholars are
tireless in their efforts to help students develop an understanding of the field and
adopt sound research methods. Laura Benua taught phonology in such a way that
even today years after I took her classes, I have a firm grasp of the issues if not the
details. Stephen Crain and Rozz Thornton would give me a solid background in
language acquisition. Rozz, especially helped me develop sound research

vii

methodology. Other professors to whom I am indebted for their teaching talent and
erudition are Paul Pietroski, David Poeppel, Philip Resnik, and Amy Weinberg.
My studies in linguistics would have never gotten off the ground if it were not
for Norbert Hornstein. From my first syntax course, through years of painful highs
and lows, Norbert inspired, prodded, yelled, praised, critiqued and criticized, and
carried me through. I cannot relay how deeply grateful I am to Norbert for his
wisdom, and for always seeing the best in me. Amy Weinberg provided cheer and
counseling for the private troubles, and know-how and invaluable assistance in
professional matters. Paul Pietroski had faith in me when even I doubted myself.
Howard Lasnik taught me how to think and write like a linguist. The guidance of
these individuals during the final period of my dissertation writing was invaluable. It
is still amazing to me how accessible these people were. I am indebted to them for
their patience, open-mindedness, and professionalism, and for sincerely wanting
success for me. I must also express heartfelt gratitude to Jacek Witkos who was the
angel that pulled me out of a particularly barren trough.
The grad students during my tenure as a student were a gifted lot. They
provided the department an atmosphere of talent and intellectual stimulation. Early
on, I was encouraged by the special talents of people like Sachiko Aoshima, Cilene
Rodrigues, Max Guimarães, Ana Gouvea, Hirohisa Kiguchi, Mitsue Motomura,
Elixabete Murgia, Acrisio Pires, Itziar San Martin. I am obliged to these people for
their assistance and their example during the early years of my studies. I owe thanks
to Scott Fults, Ivan Ortega-Santos and Masaya Yoshida for providing inspiration in
the latter years. Especially helpful and generous with his time was Tomohiro Fujii.

viii

Thanks also to the generosity of Utako Minai. I formed special friendships during
this period. People like Cilene, Pritha Chandra, Mona Diab, Lydia Grebenyova,
Soomin Hong, Youngmi Jeong, Nina Kazanina, Leticia Pablos, Usama Soltan and
Heather Taylor gave of their friendship as well as their expertise. I feel honored to
know these linguists and am thankful for the privilege of having studied with them.
Every linguist needs good informants, people who can reliably provide
judgments and who remain patient for hours. I am indebted to Aylin Bener, Mehmet
Ergene, and Murat Aytekin for this invaluable assistance.
I believe God was instrumental in my success, and He provided Atakan Ince
for me just when I needed him. Without Atakan’s help, I would not have been able to
finish much of the work I accomplished in this thesis.
Finally, I wish to thank all those people who I cannot name for lack of space.
I owe a huge debt of gratitude to a few people in particular who made me believe that
a dream could become a reality and who sustained me through the early years. I will
name one here: Martha Price, I thank you.

ix

Table of Contents



Dedication .....................................................................................................................ii
Acknowledgements ......................................................................................................iii
Table of Contents ......................................................................................................... ix
Chapter 1: Introduction ............................................................................................. 1
1 Background ........................................................................................................... 4
Chapter 2: Explaining Turkish Relative Clauses................................................... 14
1 A little Turkish grammar..................................................................................... 14
1.1 Background ................................................................................................. 14
1.2 Overview..................................................................................................... 15
1.3 A look at Turkish nominals: specificity effects .......................................... 16
1.4 The EPP....................................................................................................... 19
1.5 NPs, DPs, Case and the EPP....................................................................... 22
2 Returning to relative clauses: Generalizations.................................................... 22
2.1 Clauses that lack external arguments: the SR form..................................... 26
2.2 More examples with the SR form: the possessor of a direct object ............ 27
2.3 Recap........................................................................................................... 29
2.4 Diagnostics for non-specific subjects.......................................................... 30
2.5 Optionality of RC forms.............................................................................. 33
2.6 The option of the SR form........................................................................... 33
2.7 The option of the NSR form........................................................................ 35
2.8 Genitive case ............................................................................................... 37
2.9 Clearing up the SR option........................................................................... 39
2.10 Relative clauses with complex arguments .............................................. 43
2.11 Semantic reflex of syntactic structure ......................................................... 48
3 A minimalist account: Pestesky and Torrego (2001) .......................................... 53
3.1 Pesetsky and Torrego (2001) analysis applied to Turkish relative clauses. 59
3.2 The NSR -DIK form.................................................................................... 61
3.3 The SR form................................................................................................ 63
3.4 Recap: our new story................................................................................... 69
3.5 Clauses that lack external arguments: the SR form..................................... 71
3.6 Relative clauses with complex arguments .................................................. 74
4 Conclusion........................................................................................................... 76
Chapter 3: Specificity................................................................................................ 78
1 Introduction......................................................................................................... 78
2 Toward an analysis of DP/NP structure and Case in Turkish............................. 82
3 The EPP and case assignment ............................................................................. 94
4 Looking at Turkish ‘Quirky’ relatives ................................................................ 96
5 The subject/non-subject asymmetry is a misnomer .......................................... 100
6 Repeat of Pesetsky and Torrego (2001) ............................................................ 101
7 Explaining the choice in RC forms ................................................................... 105
8 Summary ........................................................................................................... 109

x

Chapter 4: The EPP on T and Minimality............................................................ 109
1 Introduction and Background............................................................................ 110
1.1 Specificity, case, and displacement........................................................... 110
1.1.1 The EPP on T in Turkish................................................................... 112
1.2 Evidence for the EPP in Icelandic and German........................................ 119
2 Review of Turkish Relative Clauses ................................................................. 124
3 The SR and Movement through [Spec, TP] ...................................................... 126
3.1 Relative clauses with non-specific subjects and the SR form................... 127
4 Toward an explanation...................................................................................... 132
4.1 The base position of subjects and intervention by NPs............................. 136
5 Versions of the EPP........................................................................................... 139
6 Conclusion......................................................................................................... 140
Chapter 5: Human/Non-human Distinctions in Turkish ................................... 143
1 Background ....................................................................................................... 143
2 The Problem of Human Subjects ...................................................................... 147
3 Toward a Solution............................................................................................. 152
4 Explaining the Behavior of Human Nominals .................................................. 161
5 Summary ........................................................................................................... 179
6 Contrastive Focus and Human Subjects............................................................ 180
Chapter 6: Relativization in Psych Verb Constructions .................................... 187
1 Background ....................................................................................................... 187
1.1 Classes of Turkish psych verbs ................................................................. 190
2 Turkish Psych Verbs and Relativization........................................................... 193
2.1 Experiencer subjects.................................................................................. 194
2.1.1 Subject Experiencers with accusative Theme ................................... 204
2.2 Experiencer objects ................................................................................... 205
3 Conclusion......................................................................................................... 209
Chapter 7: Relativization from Infinitivals in Turkish ....................................... 213
1 Background ....................................................................................................... 213
2 Uninflected Infinitivals ..................................................................................... 220
2.1 Review of assumptions.............................................................................. 226
2.2 Non-subject infinitivals............................................................................. 228
3 Inflected infinitivals .......................................................................................... 241
4 Taking stock...................................................................................................... 245
Chapter 8: Conclusion............................................................................................ 252
Bibliography ............................................................................................................ 258











1

Chapter 1: Introduction



While this work centers on relative clause constructions in Turkish, it includes
research into several syntactic properties that have been attested cross-linguistically
and theories in generative grammar that have been proposed to explain them. By way
of introduction, I must clarify at the outset that the account presented here is solidly
within the Minimalist Program (mainly) of Chomsky (1995, 2000). Thus, it is
assumed outright that the reader agrees with the movement (or displacement) account
of syntactic derivations. An attempt has been made throughout to keep theory
internal assumptions to a minimum (modulo the initial assumptions regarding
movement (both A- and A-bar), X-bar structure and binary branching, theta-
assignment at first merge and structural case-assignment/checking by functional
heads and other basic hallmarks of generative linguistics), and, where applicable, to
point out areas that may be controversial or where an alternative account would work
just as well without distracting from the arguments being presented.
Much of this work, however, ends up being about subject-hood and the EPP,
specifically the EPP on T. Turkish has two relative clause (RC) strategies. The one
which is the topic of this dissertation is native to the language; the other is borrowed
from Persian. The native strategy is pre-nominal in the sense that the restricting
clause which contains a gap precedes the nominal that is the external head of the

2

clause. This is expected as Turkish is a consistently head-final language.
1
There is
no relative pronoun and an overt resumptive pronoun is not permitted in the gap site.
This contrasts with the form borrowed from Persian which is post-nominal (i.e. the
restricting clause follows the head) there is a relative pronoun, the Persian ki, and it
requires a resumptive pronoun when relativizing anything other than a subject or
direct object. This form also employs the verbal inflections of matrix sentences,
whereas the native Turkish RC employs non-finite or subordinate clause inflections.
Furthermore, what has not been noted in the literature, as far as I know, is that
whereas the native form is generally a restrictive relative
2
, the borrowed Persian form
functions as an appositive in Turkish.
3
This work has nothing further to say regarding
the borrowed RC form.
The native RC itself has two forms, commonly identified (using a variety of
similarly mnemonic labels) as the Subject Relative (SR) clause form and the Non-
Subject Relative (NSR) clause form based on the grammatical function of the
expression that would have appeared in the clause internal gap position. It has been
noted that this description is not quite accurate in the sense that the SR form is
licensed in some circumstances where its function is something other than the subject

1
For arguments against Kayne’s (1994) LCA entailment that specifier-head-complement is the
universal order see Kural (1997).
2
As evaluated against the language relevant diagnostics in Del Gobbo (2003: pp. 152-162).
3
A. Ince (p.c.) pointed out the non-restrictive interpretation of the ki RC form in Turkish. In sentence
(i) containing the native RC (i), it cannot be true that the girl lost her bag if she did not buy the book,
whereas in sentence (ii) with the ki RC, it can still be true that the girl lost her bag, even if she didn’t
buy the book (note the commas around the English equivalent).
(i) kitab-ı al-an kız çanta-sı-nı kaybetti
book-ACC buy-SR girl bag-AGR-ACC lost
‘The girl who bought the book lost her bag’
(ii) kız ki kitab-ı al-dı, çanta-sı-nı kaybetti
girl COMP book-ACC buy-PST bag-AGR-ACC lost
‘The girl, who bought the book, lost her bag’


3

of the RC. Interestingly, whereas the SR is sometimes licensed for non-subjects, the
NSR is never acceptable for relativization of the subject.
We will see in Chapter 2, that the SR is, in fact, licensed when the relativized
expression moves to (and through) [Spec, TP]. The SR examples will include
relativization of nominals bearing a variety of grammatical functions and theta roles.
It turns out that under appropriate circumstances, the SR form is acceptable for
relativization of all nominals except accusative direct objects. This means that given
the syntactically permissible conditions for the movement, any nominal may move to
[Spec, TP], except one that is marked with accusative case. This fact has theoretical
implications. As will be seen in later chapters, I conclude that Case is checked at PF,
and it is the morpho-phonetic mismatch of structural case at PF that disallows
movement of an accusative case-marked expression to the structural case-assigning
position of [Spec, TP].
Svenonius (2002) notes that the notion of subject is “no more than a
descriptive label for an epiphenomenal collection of properties” (p.3). Although the
Turkish Subject Relative can be used when relativizing expressions bearing a variety
of theta roles, it is licensed only when that expression has passed through the case-
assigning position for the canonical subject. Put simply, there are three components
to “subject-hood”: thematic (the most prominent argument of a predicate), syntactic
(identified by case or agreement), and discourse-informational (the topic of a
proposition) (Svenonius 2002). This gives us another way of viewing the Subject
Relative; that is, it is licensed when syntactic and discourse prominence converges.
To put it another way, if we think of the relativized expression as the topic within the

4

clause, and if we agree that [Spec, TP] is the most prominent position in the morpho-
syntactic arena of case and agreement, then it is when the clausal “topic” is also the
clausal “subject” (i.e. occupies the canonical subject case and agreement position of
[Spec, TP]) that the SR form is licensed. The only aspect of subject-hood missing in
the definition of Subject Relative is that the relativized expression does not have to be
the most prominent in the thematic hierarchy of the predicate. It is too early here to
discuss these issues in greater detail. In fact, we will readdress them at the conclusion
of this work when the reader has many more facts under his belt. I mention them here
to point out that although much of what follows is centered on Turkish, the reader is
urged to consider throughout the larger issues and to test their applicability to his own
research interests.



1 Background
4



A few facts about Turkish: Turkish is a head-final agglutinative SOV language.
There is no overt Wh-movement, nor any complementizers in the trivial sense. It
exhibits both subject and object drop, and indeed subjects and objects appear only for
contrast, emphasis, or other marked discourse purposes. To be more precise, overt
pronouns are unacceptable unless they signal a change of topic or contrast, as in (1).


4
Abbreviations used: ABL (ablative), ACC (accusative), AGR (agreement) AOR (aorist), DAT (dative),
GEN (genitive), INF (infinitive), INST (instrumental), LOC (locative), NOM (nominative), NPI (negative
polarity item), NSR (non-subject relative), NoEA (no external argument), PASS (passive), POSS
(possessive agreement), PST (past), RC (relative clause), SR (subject relative).

5

1) Ahmet oda-ya gir-di. (*O) sandalye-ye otur-du.
Ahmet room-DAT enter-PST. (*He) chair-DAT sit-PST
‘Ahmet entered the room. He sat down in a/the chair.’

Turkish exhibits rich morphology, and is regular in the stacking of its case and
inflectional morphemes which are generally suffixival. There is only subject
agreement, no object agreement.
5
What makes Turkish useful for research is that,
unlike synthetic languages where several inflectional elements can be fused into one
morpheme, there is generally a one-to-one mapping between function and morpheme
in Turkish which yields more transparency in the syntax.
As noted above, Turkish relative clauses demonstrate a subject/non-subject
asymmetry. There are two verbal suffixes which mark relative clauses in Turkish, -
An and -DIK
6
, the choice of which is generally determined by whether the clause
internal gap site is the subject of the relative, the SR -An form as in (2)a, or a non-
subject, the NSR -DIK form as in (3)a. The -An verbal form bears no agreement
morphology. The -DIK suffix, on the other hand, is followed by possessive
morphology which shows agreement with the subject, which, when overt, bears
genitive case morphology. Because the morpho-phonological processes on the verbal
morphemes are rather complex, I will denote the Subject Relative /-An/ form as SR
and the Non-Subject Relative /-DIK/ form as NSR.



5
Although it will be argued in Chapter 5 on human DPs that the assumption regarding subject
agreement is not straightforward.
6
The capital letters indicate positions that undergo vowel harmony and consonantal assimilation. I
will use Turkish spelling throughout and not convert the Turkish examples to the IPA. The –An form
can appear as ‘-an’, ‘-en’ and the -DIK form can appear as ‘-dik’, ‘-di’, ‘-tik’, ‘-ti’, ‘-dık’, ‘-dı’, ‘-
tık’, -tı, ‘-duk’, ‘-du’, ‘-tuk’, ‘-tu’, ‘-dük’, ‘-dü’, ‘-tük’ and ‘-tü’ as a result of phonological
processes.

6

2) a. [Ø
i
divan-da otur-an] bayan
i

Ø sofa-LOC sit-SR lady
‘the lady who is sitting on the sofa’

b. *[Ø
i
divan-da otur-du-u] bayan
i

Ø sofa-LOC sit-NSR-3s lady

3) a. [bayan-ın Ø
i
otur-du-u ] divan
i

lady-GEN Ø sit-NSR-3s sofa
‘the sofa that the lady is sitting on’

b. *[Ø
i
divan-da otur-an] bayan
i

Ø sofa-LOC sit-SR-3s lady

Kornfilt (1997a) (among others
7
) describes the asymmetry by postulating that the
NSR form is the elsewhere case. It is used for subordination structures in general.
On the other hand, the SR form is the marked option used in relative clauses when:
a. the gap site is a subject or part of a larger subject
b. the gap site is a non-subject in a construction where there is no surface subject
bearing a thematic role, as in impersonal passives and existentials.

The relative clauses in (4)
8
exemplify part (b) of Kornfilt’s generalization. These
phrases contain no external argument (NoEA), and the gap site is the (oblique) object
of an impersonal passive construction. Note that only the SR form is acceptable.

4) a. [Ø
i
Ankara otobüs-ün-e bin-il-en ] durak
i

Ø Ankara bus-CM-DAT board-PASS-SR stop
9

‘the stop where the Ankara bus is boarded’


7
The first modern analysis, based on Chomsky (1965), is attributed to Underhill (1972) by Hankamer
and Knecht (1976) whose own explanation of the asymmetry is based on grammatical relations such as
subject and object. Csató’s (1985) analysis was along the lines of Chomsky (1981).
8
Examples from Kornfilt 1997.
9
I use Kornfilt’s gloss. She refers to one of the nominal morphemes as a compound marker (CM).

7

b. *[Ø
i
Ankara otobüs-ün-e bin-il-di-i ] durak
i

Ø Ankara bus-CM-DAT board-PASS-NSR stop
‘the stop where the Ankara bus is boarded’

c. [Ø
i
bu durak-tan bin-il-en] otobüs
i

Ø this stop-ABL board-PASS-SR bus
‘the bus which is boarded from this stop’

d. *[Ø
i
bu durak-tan bin-il-di-i] otobüs
i

Ø this stop-ABL board-PASS-NSR bus
‘the bus which is boarded from this stop’


What is interesting, however, is that the SR form also appears in phrases such as
those in (5)a where the gap site is not the subject. This sentence contains an overt
clause-internal subject, ship. As shown in (5)b, the relativized expression, harbor,
bears dative case clause-internally and can in no way be identified as a subject. The
example in (5)c
10
demonstrates that, as expected, the NSR form is also possible.

5) a. [gemi yana-an] liman
ship sidle-SR harbor
‘the harbor that a ship is sidling up to’

b) Liman-a gemi yana-ıyor.
harbor-DAT ship sidle-pres.prog.-3s
‘A ship is sidling up to the harbor’

c) [gemi-nin yana-tı-ı] liman
ship-GEN sidle-NSR harbor
‘the harbor that the ship is sidling up to’

Furthermore, as pointed out by Barker, Hankamer, & Moore (1990), in one common
dialect there seems to be some “optionality” in the choice of verbal forms particularly

10
As the glosses indicate, there is a difference in interpretation based on whether the SR or the NSR
form is used: in 5)a), the subject, ship, is non-specific, whereas in 5)c), ship is specific, and the NSR
form is required. This difference will later be addressed.

8

when relativizing from within a complex argument. The pairs in (6) and (7) are
examples of extraction from within a subject and which, according to the
generalization, should not allow the NSR form. But these phrases do, in fact, permit
the NSR form as in (6)b and (7)b.
11


6) a. [[Ø
1
kız-ı] kitab-ı getir-en] adam
1

Ø girl-POSS-3s book-ACC bring-SR man
‘the man whose daughter brought the book’

b. [[Ø
1
kız-ı]-nın kitab-ı getir-di-i] adam
1

Ø girl-POSS-3s-GEN book-ACC bring-NSR-3s man
‘the man whose daughter brought the book’

7) a. [[Ø
1
biz-e güven-ece-i] üpheli ol-an] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS-3s doubtful be-SR man
‘the man who that (he) will trust us is doubtful’

b. [[Ø
1
biz-e güven-ece-i]-nin üpheli ol-du-u] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS-3s-GEN doubtful be-NSR-POSS-3s man
‘the man who that (he) will trust us is doubtful’


In sum, what we have in Turkish RC’s is two verbal suffixival elements, the choice of
which is based on whether the gap site is a subject or non-subject, and which can
sometimes be violated in simple relatives, but which permit a wider (although not
completely free) optionality of choice in complex relatives.
Turkish is a Wh-in-situ language. The only way to track A-bar movement is
with relative clause constructions.
12
There are no internally-headed relatives in
Turkish, and resumptive pronouns are also banned in simple relatives. Thus, we can
assume that Turkish RCs operate within the hallmarks of classic RCs: there is an

11
Examples from Barker, et.al. (1990).
12
I exclude Topic and Sluicing here.

9

obligatory gap in the (case) position of the RC head internal to the RC which is co-
referenced with the nominal expression (the external head) which the RC is the
complement of. Turkish being a head-final language, the structure of an RC, then, is
as in (8) where X
1
is the N° head and the CP its complement. The resulting structure,
[
NP
[
CP
. . .] N°], may optionally merge with D° giving us the DP shown in (8).

8) [
DP
[
NP
[
CP
. . . [gap]
1
. . . ] X
1
] D°]


I assume, and the evidence indicates, that the internal gap site (sometimes referred to
as the focus) of the RC is a +Wh-expression (or Operator) that undergoes A-bar
movement at least to the CP projection. I argue throughout this work that the SR
relative clause form is indicative of the Wh-expression, i.e. the relative head, having
moved to (and then out of) [Spec, TP]. If this is correct, it provides us with a
diagnostic for A-movement to [Spec, TP]. The diagnostic works like this: Given a
phrase with several nominals, one can formulate a RC targeting any one of those
nominals as the head of the RC. It turns out that of the two RC forms, only one will
be acceptable. The SR form will be required when the relative head A-bar moves
from [Spec, TP] to [Spec, CP], and the NSR form will be required otherwise.
13
For
example, in the phrase in (9), there are three DPs that can potentially be targetted for
relativization, the subject, the direct object and the locative. Assuming that the
subject-DP must first A-move to [Spec, TP] for case, after which it A-bar moves to

13
There is an apparent exception in extraction from sentential subjects as in (6) and (7), but it will be
shown that even here the generalization holds.

10

[Spec, CP], we predict that extraction of DP1-subject will trigger the SR form, as in
(10)a and disallow the NSR form, as in (10)b.

9) [DP1-subject DP2-direct object DP3-locative Verb]


10) a. [
CP
Ø
1
[
TP
Ø
1
(+case) [
vP
DP2-DO-(ACC) Ø
1
(+theta) DP3-locative Verb-SR]] ] DP1
1





b. *[CP Ø
1
[
TP
. . . DP2-direct object DP3-locative Verb-NSR]] DP1
1



Extraction of DP2, the direct object, on the other hand, should require the NSR form
and prohibit the SR form because the non-Wh-subject (DP1) must move to [Spec, TP]
to be assigned case (to avoid a Case Filter Violation). In (11), the NSR form is
obligatory.

11) [
CP
Ø
1
[
TP
DP1-Subj+case [
vP
Ø
1
+ACC [
VP
Ø
1
+ DP3-Loc Verb-NSR]]]] DP2
1





In (12), the focus is now a dative object, and again the NSR form is required because
the subject must move to [Spec, TP] for case. Any time [Spec, TP] is occupied by a
subject that is non-Wh, the NSR form is required and the SR form barred.

12) [
CP
Ø
1
[
TP
DP1-Subj+case [
vP
DP1-Subj [
VP
Ø
1
+DAT DP3-Loc Verb-NSR]]] ] DP2
1





11

But, we will see that there are occasions where the subject does not raise for case. In
this case, if we target either the dative DP2 (or the locative DP3) as the relativized
expression, we can assume that if the SR form is acceptable, then the dative must
have moved through [Spec, TP], as shown in (13). On the other hand, if the SR form
is bad, and the NSR form is required, as in (12), then the dative did not move through
[Spec, TP]; this can be either because [Spec, TP] was occupied (by the subject) or an
intervening element blocked the movement.

13) [
CP
Ø
1
[
TP
Ø
1
+DAT [
VP
Ø
1
+DAT DP3-Loc NP1-Subj Verb-SR]]]] DP2
1




This then, in a nutshell, is the diagnostic that will be used throughout this work to
tease apart movement. If one wants to see if an expression can move to [Spec, TP]
one simply targets that expression as the relative head, and sees if the SR form will be
acceptable or not. As will be seen, much of this unacceptability will be due to effects
similar to what we saw in (11) and (12), where the subject occupies [Spec, TP], or to
intervention effects created by the presence of a nominal between [Spec, TP] and the
relativized Wh-expression. The workhorse in this thesis is the latter, what is termed
Minimality effects, constraints on (A-)movement of a DP induced by the presence of
intervening expressions.
As the details in this work are laid out, we see that there is a correlation
between specificity and case, and specificity and movement. In addition, contra

12

Chomsky’s assumption even as recently as 2005,
14
we will see evidence that
structurally case-marked expressions are not “frozen” for further movement but are
simply barred from moving to another structural case-assigning position.
15
Because
this work is mainly about movement of DPs, the controversial issue of the EPP is
addressed early. Initially I assume outright a definition of the EPP as a feature on a
functional head which forces movement of a DP to its Spec. As we encounter more
and more Turkish data, we will revisit the EPP and will be led to conclude that, call it
what you like, something along the lines of the EPP as a feature that needs to be
checked seems to be working in Turkish. Although scrambling is not a topic per se in
this work, the effects of scrambling and constraints on scrambling are discussed when
they become germane.
The bulk of the findings in this project leads to one conclusion: that the SR
form is an instantiation of movement to [Spec, TP]. Chapter 2 is dedicated to
explaining why this conclusion is viable. Chapter 3 is a continuation of the argument
in the sense that assumptions made in Chapter 2 about specificity and the nature of
NPs and DPs are worked out in more detail, and hopefully presented in a manner that
is more compelling. Chapter 4 argues for the EPP as a formative feature. Here we
also examine the structural hierarchy of various verb classes finding support for
Perlmutter’s (1978) “Unaccusative Hypothesis” and Burzio’s (1986) similar findings
regarding predicate structure in Italian. In Chapter 5, we see evidence that Turkish is

14
In lectures delivered at LSA Summer Institute, MIT, 2005, in addition to works outlining the
Minimalist Program.
15
Actually, even this is not quite accurate. The evidence seems to suggest that a structurally case-
marked expression is barred from moving to a case position where it will be assigned a case with a
different morphological form than the one it already bears. This restriction does not hold for inherent
case-marked expressions.

13

sensitive to, not animacy, but human vs. nonhuman features. Rather than being
merely a semantic notion, the facts in Chapter 5 make the case that these features play
a role in the syntax. Also in Chapter 5, the effects of contrastive focus on movement
and case-marking are demonstrated. In Chapter 6, we look at similar effects with
psych verbs. In both these chapters, syntactic referentiality, i.e. a D feature, is
imposed on human subjects and on Experiencers of psych verbs with consequences in
terms of movement and intervention effects. In Chapter 7, we look at relativization
out of infinitival clauses, both inflected and uninflected. The controversy regarding
control PRO (movement or not) emerges because it seems that other than subject
control PRO, all other control PRO positions serve as interveners for movement.
Does this mean that subject control PRO is a trace of movement as proposed by
Hornstein (1999). The facts lead to this conclusion by the end of Chapter 7.
The final Chapter of this research project is a compendium of issues that have
been visited in this work and that are relevant cross-linguistically. Many of the
observations made in the course of this research are useful in presenting a different
perspective with which to view phenomena in other languages, and indeed are
remarkable because they seem to reemerge in language after language. In Chapter 8,
I review conclusions that I have reached and point to theoretical questions that this
work highlights.


14

Chapter 2: Explaining Turkish Relative Clauses


1 A little Turkish grammar


1.1 Background

The subject non-subject asymmetry in Turkish relative clauses has been of interest to
many linguists: Underhill (1972), Hankamer and Knecht (1976), and Knecht (1985),
as well as others, have attempted to provide an account. More recently, Kornfilt
(1984, 1988, 1991, 1997b) and Barker, Hankamer, and Moore (BHM) (1990) have
provided analyses under a Government and Binding framework.
Let’s briefly look at Kornfilt’s (1997b) proposal. Recall that the NSR -DIK
form bears agreement with the RC subject. Using this fact as an indication of a strong
AGR, Kornfilt assumes that the NSR -DIK form is not licensed in subject gap RC’s
because the strong AGR of this form would, and indeed, must license pro in subject
position. This pro would be (A-bar) bound by an Operator in [Spec,CP] violating the
A-disjointedness Requirement (Aoun and Li 1989
16
), (1)b. Conversely, the SR form
does not bear any agreement, and according to Kornfilt, the weak AGR of the SR -An
verbal form cannot license pro. The unavailability of pro in the subject position
permits a non-pronominal empty category at the subject gap, as in (1)a.

16
“The A’-disjointedness Requirement: A pronoun must be (A’-)free in the smallest Complete
Functional Complex (CFC)”, i.e. its Governing Category which in this case is the CP. The
A’-disjointedness Requirement is a sub-clause of a generalized version of Condition B and was argued
for in Aoun & Li (1989), Borer (1984), Kornfilt (1984, 1991), McCloskey (1990), and Ouhalla (1993).

15

1) a. [[ e
i
okul-a gid-en] Op
i
] adam
i

school-DAT go-SR man
‘the man who goes/went to school’

b. *[[pro
i
okul-a git-ti-i] Op
i
] adam
i

pro school-DAT go-NSR-poss3s man
Intended: ‘the man who goes/went to school’

1.2 Overview

Let us first be clear about the logical possibilities for Turkish relatives, and which
forms appear in the grammar. Table 1 demonstrates that of the four possible
combinations with external arguments, the NSR form in sentences which contain a
subject gap (item 3) is the only form not found in the grammar. Of the two possible
combinations in sentences with “No External Argument” (NoEA), only the SR form
(item 5) is licensed despite the fact that these sentences have no canonical subject to
extract.


GAP SITE RC STRATEGY EXAMPLE
1- Subject -An (SR) (2)a
2- non-subject -DIK (NSR) (3)a
3- * Subject -DIK (NSR) (2)b
4- non-subject -An (SR) (5)
5- NoEA -An (SR) (4)a
6- * NoEA -DIK (NSR) (4)b
Table 1: Acceptability of possible strategies

2) a. [Ø
i
divan-da otur-an] bayan
i

Ø sofa-LOC sit-SR lady
‘the lady who is sitting on the sofa’

b. *[Ø
i
divan-da otur-du-u] bayan
i

Ø sofa-LOC sit-NSR-3s lady


16

3) a. [bayan-ın Ø
i
otur-du-u ] divan
i

lady-GEN Ø sit-NSR-3s sofa
‘the sofa that the lady is sitting on’

b. *[Ø
i
divan-da otur-an] bayan
i

Ø sofa-LOC sit-SR-3s lady

4) a. [Ø
i
bu durak-tan bin-il-en] otobüs
i

Ø this stop-ABL board-PASS-SR bus
‘the bus which is boarded from this stop’

b. *[Ø
i
bu durak-tan bin-il-di-i] otobüs
i

Ø this stop-ABL board-PASS-NSR bus
‘the bus which is boarded from this stop’

5) [gemi yana-an] liman
ship sidle-SR harbor
‘the harbor that a ship is sidling (or, that ships sidle) up to’


1.3 A look at Turkish nominals: specificity effects

Arguments do not always bear overt case morphology in Turkish. In addition, case-
marked expressions are in a different structural position than their bare counterparts.
This is demonstrated for the direct object in (6).
17
Assuming that Turkish adverbs of
manner mark the edge of the VP,
18
sentence (6)c shows that a case-marked object
cannot remain inside the VP, while an object without overt case must remain inside
the VP as in (6)a-b.
19



17
Examples from Tosun (1999).
18
See Kural (1992).
19
The interaction between specificity and (especially accusative) Case has been noted by modern (e.g.
Dede (1986), Enç (1991), Erdal (1981), Erguvanlı-Taylan (1984), Kornfilt (1997), Nilsson (1986),
Tura (1986)) and traditional grammarians. Kornfilt (2004) shows that the correlation between
specificity and overt case holds for all structural cases, which she identifies as nominative, genitive (of
subordinate clause subjects), and accusative. Kornfilt claims that inherently case-marked nominals are
ambiguous in their specificity as these expressions enter a derivation already case-marked.

17

6) a. Ben hızlı kitap oku-r-um
I quickly book read-AOR-1sg
‘I read books quickly.’

b. *Ben kitap hızlı oku-r-um
I book quickly read- AOR -1sg

c. *Ben hızlı kitab-ı oku-ru-m
I quickly book-ACC read- AOR -1sg
‘I read the book quickly.’

d. Ben kitab-ı hızlı oku-ru-m
I book- ACC quickly read-aor-1sg
‘I’ll read the book quickly.’

Enç (1991) notes that in Turkish, indefinite nominals in object position always
unambigously receive a specific or non-specific interpretation depending on whether
or not they bear overt case morphology. The object in (7)a bearing accusative case
must be interpreted as a specific piano,
20
whereas the non-case-marked object in (7)b
must receive a non-specific reading.

7) a. Ali bir piyano-yu kirala-mak isti-yor.
Ali one piano-ACC rent-INF want-3s-PRES
‘Ali wants to rent a certain piano.’

b. Ali bir piyano kirala-mak isti-yor.
Ali one piano rent-INF want-3s-PRES
‘Ali wants to rent a (non-specific) piano.’

So, for objects the facts are as follows: a case-marked object must raise from its base
position and receive a specific interpretation and conversely, a bare object must
remain in-situ and be non-specific. This means we have a diagnostic for raising for
objects; the presence of overt case.

20
For example, ‘specific’ for piano in sentence (9a) means there is a certain piano such that Ali wants
to rent it.

18

The same correlation between specificity, overt case and raising can be seen
for subjects in embedded environments. Whereas nominative case is the Ø or null
morpheme in Turkish, embedded subjects receive genitive case, as in (8).

8) Ali-nin Ankara-ya git-ti-i-ni duy-du-lar
Ali-GEN Ankara-DAT go-NSR-3s-ACC hear-PST-3p
‘They heard that Ali went to Ankara’

We know that subjects of existential constructions must be non-specific, and as
expected, the subject of an embedded existential construction in Turkish does not
bear overt case (9)a. Note the position of the embedded subject in (9)b, as well as its
interpretation when the embedded subject is case-marked. This is consistent with
what we have observed so far—that specifics must bear case. The case-marked
specific subject must raise, as in (10). If it is correct that temporal adjuncts are
generated high in the structure (adjoining perhaps to vP or TP), then the subject is
required to raise above temporal adjuncts, as shown by the unacceptable (10)b.
21


9) a. Yan-ın-da bir kız ol-du-u-nu gör-dü-ler.
side-his-LOC one girl be-NSR-3s-ACC see-PAST-3p
‘They saw that there was a girl by his side.’

b. Bir kız-ın yan-ın-da ol-du-u-nu gör-dü-ler.
one girl-GEN side-his-LOC be-NSR-3s-ACC see-PAST-3p
‘They saw that (of the salient girls) one (of them) was by his side.’

10) a. Ali-nin bu sabah Ankara-ya git-ti-i-ni duy-du-lar
Ali-GEN this morning Ankara-DAT go-NSR-3s-ACC hear-PST-3p
‘They heard that Ali went to Ankara this morning’



21
What is relevant for us, of course, is the unmarked and non-scrambled cases.

19

b. *bu sabah Ali-nin Ankara-ya git-ti-i-ni duy-du-lar
this morning Ali-GEN Ankara-DAT go-NSR-3s-ACC hear-PST-3p

Because case morphology and displacement seem to go hand in hand, we can
conclude that (at least in embedded environments) case marking on an argument—in
the form of genitive case on the subject and accusative case on the object—is
evidence of raising. Furthermore, the facts show that only case-marked arguments
receive a specific interpretation; specifics must be case-marked, and non-specifics
cannot bear overt case. The examples in 11) demonstrate that the correlation between
raising, case and specificity holds for objects in embedded environments.

11) a. Aye-nin pasta-yı kaık-la /bahçe-de /hızlı ye-di-i-ni gör-dü-ler
Aye-GEN cake-ACC spoon-INST/garden-LOC/fast eat-NSR-3s-ACC saw-PST-3p
‘They saw that Aye ate the cake quickly/in thegarden/with a spoon’

b. *Aye-nin pasta kaık-la /bahçe-de /hızlı ye-di-i-ni gör-dü-ler
Aye-GEN cake spoon-INST/garden-LOC/fast eat-NSR-3s-ACC saw-PST-3p

c. Aye-nin kaık-la /bahçe-de /hızlı pasta ye-di-i-ni gör-dü-ler
Aye-GEN spoon-INST/garden-LOC/fast cake eat-NSR-3s-ACC saw-PST-3p
‘They saw that Aye ate (some) cake quickly/in thegarden/with a spoon’

d. *Aye-nin kaık-la /bahçe-de /hızlı pasta-yı ye-di-i-ni gör-dü-ler
A.-GEN spoon-INST/garden-LOC/fast cake-ACC eat-NSR-3s-ACC saw-PST-3p

1.4 The EPP

I assume that in Turkish, functional heads have an EPP feature
22
and that all phrasal
movement is driven by the EPP.
23
Thus T° (and vº) has an EPP feature that must be

22
To be more specific, I assume that the EPP is a feature of the functional heads v°, T°, and C°.
Although the outcome is the same, this description of the EPP is different from the notion of the EPP
as a structural requirement of an occupied specifier. At this point, I am not wedded to any particular
definition of the EPP; for me the “EPP” is merely a label for whatever it is that drives overt XP
movement. A more detailed discussion of the EPP will follow in Chapter 4.

20

satisfied. This is supported by the pair of sentences in 12).
24
The subject in (12)a, is
non-specific and cannot raise. It is generally assumed in the literature that Turkish
locatives are generated in the VP (Kural 1992). By parity of reasoning from the
examples we saw above with respect to objects and embedded subjects, we can
assume that the locative expression has raised from VP to [Spec, TP] to satisfy the
EPP on Tº. Compare (12)a with (12)b where the locative is lower than the specific
subject. Recall that nominative case is the Ø-morpheme, so the subject in (12)b is
presumably “overtly” case-marked and has raised to [Spec, TP]. We saw a similar
pair in (9). The embedded subject to the right of the locative in (9)a is non-specific
and bears no overt case. In (9)a, T’s EPP feature is satisfied by the locative “at/by his
side” whereas in (9)b, the specific subject raised to T, and receives overt case (albeit
genitive
25
).

12) a. Sokak-ta köpek havl-ıyor.
street-LOC dog bark-PRES
‘A dog/dogs are barking in the street.’

b. Köpek sokak-ta havl-ıyor.
dog street-LOC bark-PRES
‘The dog is barking in the street.’

Another potential argument for an EPP feature on Tº comes from the pair of
sentences below from Kural (1992). I assume that the unacceptability of the sentence
in (13)a is due to failure to satisfy the EPP of T. Compare with the minimally

23
By movement, I exclude all scrambling-type movement as it is not germane. In later chapters I will
distinguish between feature driven movement, i.e. to satisfy the EPP, and scrambling, which is optional
and apparently costless.
24
from Kelepir (2001).
25
It is beyond the scope of this thesis to address how embedded subjects receive genitive rather than
nominative case. One approach is Hiraiwa (2001) which suggests that genitive case is assigned by a
v-T-C amalgam.

21

different acceptable sentences in (13)b which contains a locative expression. A non-
specific subject cannot satisfy the EPP, but a locative can raise to T and save the
derivation. Note that the word for ‘here’ in Turkish is a nominal expression with
locative case, as demonstrated in (13)c.
26


13) a. *[Bir tavuk] pi-iyor
a chicken cook-PRES-AGR
‘A chicken is cooking’

b. Burada [bir tavuk] piiyor
here a chicken is-cooking

c. bu-ra-da
this-“place”-LOC
‘here’ [Literally: ‘at this place’]

I will return to a lengthier discussion on the EPP in Chapter 4. I include this much
here to justify my assumption that in Turkish sentences, [Spec, TP] must be occupied,
and in the absence of a specific subject, another nominal is required to move to that
position. I have encoded these facts by assuming that it is the EPP on T that must be
satisfied or the derivation will crash.



26
I take the expressions ‘here’ and ‘there’ in Turkish to be nominal because they can take a variety of
cases, as shown in (i) and (ii). Although bu and o can function as independent lexical items denoting
‘this’ and ‘that’, respectivelly, I do not know what -ra- is and assume it means something like ‘place’.
(i) a. bu-ra-dan b. o-ra-dan
this-??-ABl that-??-ABL
‘from here’ ‘from there’
(ii) a. bu-ra-ya b. o-ra-ya
this-??-DAT that-??-DAT
‘to here’ ‘to there’


22

1.5 NPs, DPs, Case and the EPP

In order to capture the complementarity between raising, overt case, and specificity
on the one hand, and the fact that bare arguments in-situ must be non-specific, on the
other, I will adopt an NP/DP distinction for Turkish. That is, I assume that non-
specific nominals are NPs and specific nominals are DPs. I will discuss this
assumption in more detail in the next chapter, but let’s take it as reasonable and adopt
it for now. The facts about case morphology and displacement fall out if I further
assume that only DP’s need satisfy the Case Filter and that the EPP can only be
satisfied by DP’s.
27
Neither of these assumptions seems far-fetched, and I will
address the arguments behind these assumptions later as well. In the next section, we
will see how much mileage these assumptions buy us in formulating an account of
Turkish relatives.


2 Returning to relative clauses: Generalizations


I have argued that, in Turkish, functional heads
28
, Tº and vº, have an EPP feature
which must be satisfied by a DP, or a specific nominal. I have also assumed that non-
specifics are NPs which do not need to satisfy the Case Filter nor can they satisfy the
EPP. In order to account for the facts in relative clauses, I must further assume that

27
In fact, Chomsky (1995) defines the EPP as a “strong D feature”.
28
I will not argue whether or not v° has an EPP feature. The issue throughout this thesis is the EPP on
T° (and on C°). I mention v° merely for uniformity although the facts do show that accusative objects
must raise.

23

Cº also has an EPP feature which attracts a +Wh DP to its specifier. With this much
technology in hand, let’s return to the facts in Table 1.
First, let’s look at the NSR form, item 2 in Table 1. In sentence (14)a, the
subject lady, being specific, is a DP. As the simplified derivation in (14)b shows, the
EPP of Tº attracts the subject DP, lady, to [Spec, TP] in movement . The +Wh DP
sofa moves to [Spec, CP] to satisfy the EPP on Cº in movement . We will not
concern ourselves with whether this element further moves to a position external to
the clause or whether some matching operation co-indexes it with an external head.
At this point, we are only concerned with operations internal to the relative clause. I
should mention that rather than Operator movement, I assume that the +Wh element
itself moves all the way to [Spec, CP].
29


14) a. [bayan-ın Ø
i
otur-du-u ] divan
i

lady-GEN Ø sit-NSR-3s sofa
‘the sofa that the lady is sitting on’

b.
CP
¸ ¸¸ ¸
sofa ¸ ¸ ¸ ¸
+Wh TP C°
¸ ¸ ¸ ¸
lady+GEN ¸ ¸¸ ¸
VP T°
¸ ¸ ¸ ¸
DP ¸ ¸¸ ¸
t-lady DP V°
t-sofa sit
+uWh




29
Throughout this paper, I will be assuming the raising analysis (Brame 1968; Schachter 1973;
Vergnaud 1974; Kayne 1994; Bianchi 1999) for relative clauses, although the matching analysis would
also work for the issues being presented here, with modifications. Nothing being argued rests on a
specific analysis for relative clauses.
.

24

Now, let’s shift gears and look at all the RCs that require the SR form to see if we can
find any commonality among them. Recall that when the subject is being relativized,
the SR form must always be used, No.1 in Table 1. The SR form is also licensed
when the subject is non-specific and a non-subject is being relativized, as noted in
No. 4 in Table 1.
Let us begin with subject gap clauses as in sentence (15)a, the derivation of
which appears in (15)b. (For simplicity, I have omitted details in the tree such as the
vP projection and accusative case on the object.) As demonstrated in (15)b, T has an
EPP feature which must be satisfied. The DP subject, bee, is attracted to [Spec, TP],
movement . The subject is also the +Wh relative head. It is attracted by the EPP of
Cº. So, in movement , the subject bee moves from [Spec, TP] to [Spec, CP].

15) a. [[Ø
1
] [ [kız-ı sok-an ] ] arı
1

Ø girl-ACC sting-SR bee
‘the bee that stung the girl’

b. CP
¸ ¸¸ ¸
bee ¸ ¸ ¸ ¸
+Wh TP C°
¸ ¸ ¸ ¸
t-bee ¸ ¸¸ ¸
+Wh VP T°
¸ ¸ ¸ ¸
DP ¸ ¸¸ ¸
t-bee DP V°
+Wh girl sting


Let’s turn now to clauses with non-specific subjects. Recall that the NSR form is
obligatory (when relativizing a non-subject) when the subject is specific. In contrast
to the example in (14) above, the SR form is licensed in the RC we saw in (5) (item 4
in Table 1), repeated as (16)a, precisely because the subject ship is non-specific. The

25

subject is an NP and cannot satisfy the EPP on a head. As demonstrated in the tree in
(16)b, ship cannot (and has not) raised from its base-generated position. The EPP of
Tº must be satisfied by another nominal. In this case it attracts the +Wh-DP, harbor,
movement . This +Wh-element in [Spec, TP] now moves to [Spec, CP], , to
satisfy the EPP features of Cº.

16) a. gemi yana-an liman
ship sidle-SR harbor
‘the harbor that a ship is sidling up to’

b.
CP
¸ ¸¸ ¸
harbor ¸ ¸ ¸ ¸
TP C°
¸ ¸ ¸ ¸
t-harbor ¸ ¸ ¸ ¸
+Wh VP T°
¸ ¸ ¸ ¸
NP ¸ ¸¸ ¸
ship DP V°
t-harbor sidle.up.to
+Wh




Notice that in the minimally different example (17)a, with derivation (17)c, the
subject ship is specific, and the NSR form is required. What is different between the
derivations of the RCs in (16)a and (17)a? Notice that in the tree in (17)c, [Spec, TP]
was occupied by a non-Wh element, the DP-subject, whereas in (16)b, the NP-subject
remained in-situ, leaving [Spec, TP] vacant for the +Wh-expression to move to.

17) a. [gemi-nin yana-tı-ı] liman
ship-GEN sidle-NSR harbor
‘the harbor that the ship is sidling up to’

b. *[gemi-nin yana-an] liman
ship-GEN sidle-SR harbor

26

c. CP
¸ ¸¸ ¸
harbor ¸ ¸ ¸ ¸
+Wh TP C°
¸ ¸ ¸ ¸
ship+GEN ¸ ¸¸ ¸
VP T°
¸ ¸ ¸ ¸
DP ¸ ¸¸ ¸
t-ship DP V°
t-harbor+DAT sidle.up.to
+Wh




In fact, every time the specifier of T is occupied by a non-Wh expression, as in (14)
and (17), the NSR form is required. This means that where the EPP of T is satisfied
by an expression that is also +Wh, the SR form is licensed. I do not mean this to be
an explanation; at this stage, I am merely making an observation.

2.1 Clauses that lack external arguments: the SR form

Having formulated a generalization, we can now predict that phrases with no external
arguments will also require the SR form. Because there is no “subject” to occupy it,
[Spec, TP] will be vacant for a +Wh non-subject to move to it to satisfy T’s EPP
feature. To demonstrate this idea, let’s look at the clause (18)a, No. 5 in Table 1. I
have included the sentence in (18)b to demonstrate that neither of the two nominal
expressions in the relative clause, “this stop” and “bus”, requires structural case: the
PP ‘from this stop’ is rendered in Turkish as the nominal this stop with ablative case,
and bus receives inherent dative case. As the derivation in (19) shows, the +Wh DP
bus moves to [Spec, TP] to satisfy the EPP on T in . This element then moves to

27

[Spec, CP] to check the EPP on C, . Again, we have a +Wh element in [Spec, TP]
and the SR form is required.

18) a. [Ø
i
bu durak-tan bin-il-en] otobüs
i

Ø this stop-ABL board-PASS-SR bus
‘the bus which is boarded from this stop’

b. Otobüs-e bu durak-tan bin-il-ir.
bus-DAT this stop-ABL board-PASS-AOR
‘The bus is boarded from this stop.’

19) CP
¸ ¸¸ ¸
+Wh-bus ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
t-(+Wh-bus) ¸ ¸¸ ¸
VP T°
¸ ¸¸ ¸
t-(+Wh-bus) VP
¸ ¸ ¸ ¸
DP V°
this stop-ABL board


2.2 More examples with the SR form: the possessor of a
direct object

Let us now look at the derivation of a sentence that is more complicated. In the SR
example in (20)a, the subject is non-specific, and the relativized element originates
within the accusative object. In this sentence, the subject, bee, is a non-specific NP
which cannot satisfy the EPP. As illustrated in the derivation in (20)b, bee remains in
its base-generated position. Because the object, girl’s leg, is a DP, it is attracted by
the EPP of v,
30
and receives accusative case, as in . In , the EPP of T attracts the

30
I had up to now ignored the vP projection but must include it here to have the object raise above the
subject so that we do not incur Minimality violations or intervention from the subject. We assume that
there is a vP projection that assigns accusative case (in a [Spec-head] configuration) from sentences

28

possessor of the object, +Wh-girl, from the specifier of the object in [Spec, vP]. Of
course, girl raises again from [Spec, TP] to [Spec, CP] to satisfy the EPP on C.
31

Again, the SR form is licensed when a +Wh element has moved to [Spec, TP].

20) a. [[Ø
1
baca-ın-ı] arı sok-an] kız
1

Ø leg-POSS-ACC bee sting-SR girl
‘the girl whose leg a bee/some bees stung’

b. CP
¸ ¸¸ ¸
girl ¸ ¸ ¸ ¸
+Wh TP C°
¸ ¸¸ ¸
+Wh-girl ¸ ¸¸ ¸
vP T°
¸ ¸¸ ¸
DP+ACC ¸ ¸¸ ¸
NP-bee ¸ ¸¸ ¸
VP v°
¸ ¸ ¸ ¸
DP V°
¸ ¸¸ ¸ sting
+Wh- DP ¸ ¸ ¸ ¸
girl-GEN leg+Agr D°





Notice in (20)b, that movement is of the +Wh element out of the now-accusative
object-DP in [Spec, vP]. The direct object has had its case checked/assigned, and
thus is frozen for movement into another case-checking (A-) position. However, it is

such as (i). Perhaps a more or more accurate interpretation for (i) is ‘The road is blocked by a car’
where what is important in the utterance is not that a car has blocked the road, but rather that the road
is blocked, by some car or other. Compare (i) with (ii), where car must now receive a specific
interpretation, and where it has presumably raised to [Spec, TP] and been assigned nominative case.
(i) Yol-u (bir) araba tıkamı
road-ACC one car blocked
‘A car has (or ‘Some cars have) blocked the road’
(ii) Araba yol-u tıkamı
car road-ACC blocked
‘The car has blocked the road’
31
The Wh-expression also deletes C’s +Wh feature. I have ignored the issue of features so far and will
address this in more detail later.

29

porous for movement from within it. Although I will not elaborate here, I do not
assume a left-branch condition exists in Turkish.
32

2.3 Recap

Let us review our assumptions thus far. T has an EPP feature which can only be
satisfied by a DP. If the subject is a DP, it must be attracted to the Spec of T.
33
The
SR form is licensed when the expression that satisfies T’s EPP feature is +Wh. In a
clause where the subject is non-specific, i.e. is an NP which cannot satisfy the EPP on

32
Interestingly, neither RC form permits relativization of the entire DP, girl’s leg. Both the SR
strategy in (i)a and the NSR strategy in (i)b are unacceptable. Whereas relativization of a complex
possessor-possessee DP object is not acceptable, extraction of the DP from the specifier of that DP
object is acceptable. (ii)a is the NSR form with a specific subject and (ii)b is the SR with a non-
specific subject. This makes sense if we are assuming that in relativization, only the head, or N°, of
the Wh-DP is promoted, and that nominals with specifiers are DPs.
i. a. *[ Ø
1
arı sok-an ] [kız-ın baca-ı]
1

Ø bee sting-SR girl-GEN leg-POSS
‘the girl’s leg that a bee/some bees stung’
b. *[ Ø
1
arı-nın sok-du-u ] [kız-ın baca-ı]
1

Ø bee- GEN sting-NSR- POSS.3S girl-GEN leg-POSS
‘the girl’s leg that the bee stung’
ii. a. [arı-nın [[Ø
1
baca-ın-ı] sok-tu-u] kız
1

bee-GEN Ø leg-POSS-ACC sting-NSR-POSS.3S girl
‘the girl whose leg the bee stung’
b. [[Ø
1
baca-ın-ı] arı sok-an] kız
1

Ø leg-POSS-ACC bee sting-SR girl
‘the girl whose leg a bee stung’
Even in English, some possessor-possessee relatives sound odd: ‘the woman that bought the fish’
sounds much more natural than ‘Anne’s mother that bought the fish’. Because ‘Anne’s mother’ picks
out a unique individual, possessor-possessee DP’s may be more like appositives. ‘John’s car which I
washed’ does not mean ‘(of John’s three cars) the car that I washed’. RC’s in Turkish fail a range of
del Gobbo’s (2003) diagnostics for appositives. In fact, to construct an appositive in Turkish one must
make use of the borrowed Persian complementizer ki which introduces a clause, though subordinate,
nevertheless marked by matrix verbal and case properties. The SR and NSR RCs can only be
restrictive (Meral 2004) and this may be one reason possessor-possessee DP’s cannot be relativized.
Another way of looking at this might be to remember that we had assumed the raising
analysis of RP’s in which it is only the N° of the RC head that is promoted beyond the CP to the matrix
clause. A possessor-possessee structure is a DP and too big to be promoted. This would also explain
why a RC such as (i) is bad. If proper names are referential and must therefore be a D, then (i) is bad
because a D(P) cannot be promoted.
(i) *Anne that bought the fish
33
Except for the accusative case-marked object also in [Spec, vP], attraction of any DP other than the
subject will violate Minimality. I assume that DP arguments receive structural case in a Spec-Head
configuration. When the subject is specific, it must obligatorily move to [Spec, TP] to avoid a Case
Filter crash. Thus, even though both arguments are equidistant from the point of view of T’s EPP, the
subject must raise for the derivation to converge.

30

T, relativization of any other element will license the SR morpheme. This is because
the specifier of T will be free to host the +Wh-non-subject. This is shown in sentence
(21), where the locative DP, fields is the relative head. Because the subject is non-
specific, +Wh-fields is attracted by T. At some point in the history of this phrase,
there was a +Wh element in [Spec, TP]; thus, the SR form is required.

21) a. [Ø
1
mısır yeti-en] tarla
1

Ø corn grow-SR field
‘the field where corn grows’

b. CP
¸ ¸¸ ¸
field ¸ ¸ ¸ ¸
+Wh TP C°
¸ ¸ ¸ ¸
t-field ¸ ¸¸ ¸
+Wh VP T°
¸ ¸ ¸ ¸
PP/DP ¸ ¸ ¸ ¸
t-field NP V°
+Wh corn grows



2.4 Diagnostics for non-specific subjects
34


Enç (1991) points out Turkish has quantificational determiners
35
and NPI expressions
that have selectional restrictions for specificity. There are two determiners in Turkish
which both mean ‘some’ but differ in terms of their specificity. Birkaç patterns like
the English ‘some’ in that it can receive either a specific on non-specific
interpretation (22)a. Nominal expressions with bazı are always specific (22)b.


34
I will return to specificity and the nature of DPs and NPs in Chapter 3, but I include a brief
discussion here as it is important to the account I am proposing.
35
Although the determiners in question translate as ‘some’ and could be called quantifiers, I am
interested in their determiner-like properties and so will refer to them as determiners.

31

22) a. Ali Zeyneb-e birkaç kitap/kitab-ı postala-dı.
Ali Zeyneb-DAT some book/book-ACC mail-PAST
‘Ali mailed some /some-of-the books to Zeyneb.’

b. Ali Zeyneb-e bazı *kitap-lar/kitab-lar-ı postala-dı.
Ali Zeyneb-DAT some books/books-ACC mail-PAST
‘Ali mailed some of the books to Zeyneb.’

The determiner bazı (but not birkaç) is ungrammatical in existential constructions
(23)a but not in the non-existential locative construction (23)c, precisely because bazı
requires a specific, i.e. presupposed, interpretation. Notice the word orders in (23)b
and (23)c: in the existential construction with non-specific subject (23)b, the locative
must raise above the subject whereas in (23)b the specific subject has raised to T.

23) a. *Bahçe-de bazı çocuklar var.
garden-LOC some children exist
‘There are some children in the garden.’

b. Bahçe-de birkaç çocuklar var.
garden-LOC some children exist
‘There are some children in the garden.’

c. Bazı çocuklar bahçe-de.
some children garden-LOC
‘Some of the children are in the garden.’

The same pattern can be seen in the Turkish negative polarity determiner hiçbir ‘any’
(literally ‘any one’.) This determiner always forms a specific nominal expression in
Turkish: it requires accusative case morphology (24)a and is banned from existential
constructions (24)b.

24) a. Ali hiçbir *kitap/kitab-ı al-ma-dı.
Ali any book/book-ACC buy-neg-PAST
‘Ali didn’t buy any of the books.’

32

b. *Bahçe-de hiçbir çocuk yok.
garden-LOC any child doesn’t-exist
‘There aren’t any of the children in the garden.’

Thus certain determiners and NPI items are incompatible with a non-specific
interpretations: nominal expressions with bazı ‘some’ and with the negative polarity
determiner hiçbir ‘any’ must always be interpreted as specific.
If we are on the right track, that is, if the SR form is licensed when non-
relativized subjects are non-specific NP’s, we would predict that the SR form would
be unacceptable when the subject contains the obligatorily specific determiner, bazi,
or the specific NPI, hiçbir. We see that this is indeed the case. Both bazı and the
determiner birkaç (which allows both the specific and the non-specific readings) are
acceptable with the NSR form in (25)a, while (25)b demonstrates that bazı is
unacceptable in the otherwise grammatical SR clause we saw in (20)a. Likewise, the
NPI item, hiçbir, yields ungrammaticality in (26)b.

25) a. [Bazı/birkaç arı-nın [[Ø
1
baca-ın-ı] sok-tu-u ] kız
1

some of the/some bee-GEN Ø leg-POSS-ACC sting-NSR-POSS.3S girl
‘the girl whose leg some of the/some bees stung’

b. *[[Ø
1
baca-ın-ı] bazı arı sok-an] kız
1

Ø leg-POSS-ACC some bee sting-SR girl
‘the girl whose leg some (of the) bees stung’

c. [[Ø
1
baca-ın-ı] birkaç arı sok-an] kız
1

Ø leg-POSS-ACC some bee sting-SR girl
‘the girl whose leg some bees stung’

26) a. [hiçbir arı-nın [[Ø
1
baca-ın-ı] sok-ma-dı-ı] kız
1

any bee-GEN Ø leg-POSS-ACC sting-NSR-POSS.3S girl
‘the girl whose leg no bee stung’

b. *[[Ø
1
baca-ın-ı] hiçbir arı sok-may-an] kız
1

Ø leg-POSS-ACC any bee sting-SR girl

33

This further supports our argument that the SR form (which must be used when
extracting a subject) may be used when a non-subject is being relativized only when
the clausal subject is non-specific, i.e. is an NP and cannot satisfy T’s EPP.

2.5 Optionality of RC forms

Turkish RCs seem to permit a certain degree of optionality in the choice of verbal
paradigm. In the sentences in (27) both the SR and NSR forms are acceptable
36
.

27) a. [[Ø
1
kız-ı ] kitab-ı getir-en] adam
1

Ø girl-POSS.3s book-ACC bring-SR man
‘the man whose daughter brought the book’

b. [[Ø
1
kız-ı]-nın kitab-ı getir-di-i] adam
1

Ø girl-POSS.3s-GEN book-ACC bring-NSR-3s man
‘the man whose daughter brought the book’

According to our assumptions, a +Wh DP must have moved to [Spec, TP] in the SR
clause in (27)a whereas a non-Wh DP must have been attracted to [Spec, TP] in the
NSR clause in (27)b. Let’s look at how this can be.

2.6 The option of the SR form

In the relative clauses in (27), the gap site is not the subject; it is the possessor of the
subject. We have determined that the SR form is licensed when a +Wh expression
has moved to [Spec, TP]. This can happen in RCs where the subject has not been
relativized only when the subject is an NP and not a candidate for the EPP or Case.

36
Although there seems to be some dialectical variation as to the acceptability of both these forms, my
aim is to provide an account for the dialect where both forms are acceptable.

34

In the SR RC in (27)a, a +Wh element must have moved to [Spec, TP]. If the
subject [man’s daughter] in (27)a were a DP, it would be attracted to [Spec, TP].
Since the subject itself is not +Wh—only the possessor man is—the SR form would
be barred. Thus, in (27)a, the subject cannot have moved to [Spec, TP]; only the
+Wh element from within the subject must have raised to T. The subject itself must
be a non-specific NP that has not raised from its base-generated position. In (27)a,
because the SR form is acceptable, under our assumptions, we must assume that it
was the +Wh possessor of the subject that moved to [Spec, TP].
Let’s look at the derivation of (27)a. First, note that in both examples in (27),
the direct object, book, must raise to [Spec, vP] because it is specific and therefore a
DP. In the tree in (28) for (27)a, the object was attracted by the EPP of v and has had
its case checked/assigned, in . The entire subject, man’s daughter, is a non-specific
NP. But, the possessor, man, in the Spec of the subject is a DP with +Wh features.
The +Wh DP-man, is attracted by the EPP of T°, after which it moves to [Spec, CP].

28) the SR form
CP
¸ ¸ ¸ ¸
¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
+Wh-man ¸ ¸ ¸ ¸
vP T°
¸ ¸¸ ¸
book-ACC vP
¸ ¸ ¸ ¸
NP ¸ ¸ ¸ ¸
¸ ¸¸ ¸ VP v°
+Wh-DP-man ¸ ¸¸ ¸ ¸ ¸ ¸ ¸
N° DP V°
¸ ¸ ¸ ¸ book bought
daughter+Agr





35

Although this account may explain the acceptability of the SR form, this derivation
will give us the wrong word order: [t
1
book-acc [[t
1
daughter]] man
1
. And it looks
like this movement violates Minimality (at least under some definitions). We will
return to this problem shortly.

2.7 The option of the NSR form

According to our generalization, in the NSR clause in (27)b, the element in [Spec,
TP] must be a non-Wh DP. Since the subject of the clause is not a +Wh expression,
this is what must have moved to T. The subject man’s daughter must therefore be a
DP in this example. This is shown in derivation (29) for (27)b. In move , the direct
object book, being a DP, raises to [Spec, vP]. (It also receives overt accusative case.)
The entire DP subject is attracted to satisfy the EPP of T in . The DP subject has its
case assigned/checked by T. Crucially, in (29), the subject is not a +Wh DP; the DP-
man in its specifier is. Thus the SR form is not licensed. This differs from the tree in
(28) where the +Wh DP was a specifier of an NP. The Wh-element, man, in the
specifier of the subject in [Spec, TP] then moves in to C.
Let’s look at the derivation in (29) for the NSR example 27)b. The EPP of T
targets the closest DP. Let us assume for now that the subject and the element in its
Spec are equidistant from T, and thus both candidates for Attract by T’s EPP. Thus,
T’s EPP can be satisfied by both the in situ subject as well as the expression, man, in
its Spec. However, if the DP-man raises out of the subject to T, the derivation won’t
converge. This is because if the entire DP subject does not raise to T, it will remain

36

without case, violating the Case Filter.
37
The EPP of C°, on the other hand, can only
attract a +Wh element. Whereas there are more options for the EPP of T, the EPP of
C must specifically target a +Wh DP for convergence, as in .
In sum, in (29), of the two DP’s (the subject and the object) in [Spec, vP], the
subject DP must raise to [Spec, TP] to avoid a Case Filter violation. Furthermore,
raising of the +Wh DP from the specifier of the subject to [Spec, TP] is barred for the
same reason (the subject would remain without case).
38


29) the NSR form
CP
¸ ¸ ¸ ¸
+Wh- man ¸ ¸ ¸ ¸
TP C°
_¯ _¯ _¯ _¯
DP ¸ ¸¸ ¸
¸ ¸¸ ¸ vP T°
man-GEN ¸ ¸¸ ¸ ¸ ¸¸ ¸
+Wh daughter D° book ¸ ¸¸ ¸
DP ¸ ¸¸ ¸
¸ ¸ ¸ ¸ VP v°
¸ ¸ ¸ ¸
book V°
bought






37
The evidence from the strict correlation between structural case assignment and displacement leads
to the conclusion the structural case in Turkish must be assigned in a Spec-head configuration. This
assumption and possible variations will be discussed in later chapters.
38
Perhaps a more problematic issue is that in , the +Wh element was a constituent of the subject. I
am assuming that, at least under certain conditions, subjects are not islands in Turkish. Even in
English, examples in the literature regarding the so-called Subject Condition effect are not definitively
unacceptable. For example, in his discussion of the SC in extraction from non-finite clauses, Stowell
(1991) includes sentences such as (i) with the following proviso: “Although there is some variability in
the judgements …”
(i.) a. ?*Who
i
do you consider [[the oldest sister of t
i
] to have left]?
b. ?*Which book
i
did you find [[the author of t
i
] very eloquent]?
c. ?*Who do you judge [[John’s having visited t
i
] very unwise]?
I find that the sentences in (ii), which should be worse because we are extracting out of the subject of a
finite verb, are not any more degraded than Stowell’s examples.
(ii.) a. ?*Who
i
do you believe [[the oldest sister of t
i
] left]?
b. ?*Which book
i
did you say [[the author of t
i
] was very eloquent]?
c. ?*Who do you feel [[John’s having visited t
i
] was very unwise]?

37

To review what we have determined thus far: whenever the subject is specific,
no matter what non-subject nominal is being relativized, the NSR form must be used
because [Spec, TP] will not be available for that element to move into. To put it
another way, when the subject is a DP, the element that checks T’s EPP feature must
be the subject which also receives case from T.
Thus, the choice of the relative clause form depends on whether [Spec, TP] is
available for the +Wh element or not. When the subject is non-specific, it cannot be a
target of T’s EPP and [Spec, TP] will be available for another DP. If the relativized
element cannot move into [Spec, TP], the SR form will be barred.
Before we can resolve the word order problem noted in the derivation of (27)a
in (28), we need to look at case assignment inside NPs and DPs.

2.8 Genitive case

Recall that we assumed that DPs require case and that NPs do not. Let’s turn our
attention now to the specifier position inside DPs and NPs. Except for nominative
case, all other cases in Turkish are overtly case-marked. We will use this overt case
marking as a diagnostic for DP’s: a nominal expression without case morphology is
an NP, while a structurally case-marked expression must be a DP—the structural
cases being nominative (with the phonetically null Ø morpheme), accusative, and
genitive. Referring back to the example in (29), note that the EPP of T can be
satisfied by the DP subject, [man’s daughter] as well as the expression, man, in its
Spec. However, if the DP-man raises out of the subject to [Spec, TP], its in situ
remnant, the DP [trace-man daughter] will remain without case, and the derivation

38

won’t converge. As we will see throughout this thesis, Turkish does not permit case
assignment via Agree. Thus, in RCs, when a subject bears overt genitive case, we
must assume it is a DP, and that it has raised to [Spec, TP] where it satisfied T’s EPP
feature and was assigned case by T in a Spec-Head configuration.
Looking now at the examples in (30), city is in the specifier position and can
be either case-marked or not. When city has genitive case, it must also receive a
specific interpretation. This is consistent with what we have determined thus far
about Turkish nominals. Note in (30), as well as in (31), that there is agreement
between the possessor and the possessee regardless of whether either gets case.
39
In
(30)a, we have a nominal, walls, with an NP in its specifier; in (30)b, a case-marked
element, therefore a DP, is in the specifier position.

30) a. ehir duvar-lar-ı [[
NP
city] walls]
city wall-pl-AGR
‘city walls’

b. ehir-in duvar-lar-ı [[
DP
city-GEN] walls]
city-GEN wall-pl-AGR
‘walls of the city’

In (31) I have listed all possible NP/DP combinations. In (31)a, we have an NP with
an NP in its specifier. In (31)b, we have a DP with a DP in its specifier. In (31)c, we
have a DP with an NP in its specifier. And finally, in (31)d, note that a DP in the
specifier of an NP is bad.

39
The morpheme I call ‘AGR’ is frequently referred to as a compound marker in the literature. In
Chapter 5, I argue that this is a (possessive) agreement morpheme. On another note, these facts are
contra Chomsky who assumes that case and agreement go hand-in-hand. In Turkish, we may be able
to say that case-marked elements must agree but that the converse need not hold: agreeing elements
need not be case-marked. This is similar to the asymmetry I assume with nominative case on T. A
subject DP must be assigned case by T, but T doesn’t have to assign case.

39

31) a. ehir duvar-lar-ı gördüm. [
NP
[
NP
city] walls]
city wall-pl-AGR see-PST-1s
‘I saw city walls.’

b. ehir-in duvar-lar-ı-nı gördüm. [
DP
[
DP
city-GEN] walls]-ACC
city-GEN wall-pl-AGR-ACC see-PST-1s
‘I saw the walls of the city.’

c. ehir duvar-lar-ı-nı gördüm. [
DP
[
NP
city] walls]-ACC
city wall-pl-AGR-ACC see-PST-1s
‘I saw the [city walls].’

d. *ehir-in duvar-lar-ı gördüm. *[
NP
[
DP
city-GEN] walls]
city-GEN wall-pl-AGR see-PST-1s
‘I saw walls [the city].’

These facts lead us to conclude that whereas D° assigns genitive case to a DP in its
specifier, N° does not assign case. Hence, when the subject of a relative clause is an
NP, a DP in its specifier cannot get case and must raise to a case-assigning head in
order to receive case. Failure to raise will result in a Case Filter violation.

2.9 Clearing up the SR option

Looking back at the derivation in (28), we can now see why the DP in the specifier of
the NP subject had to raise to [Spec, TP] as in (32). If the DP does not raise, it will
remain without case and the derivation will crash.
Notice that derivation (32) still gives us the wrong word order for (27)a
repeated below as (33)a. Assuming that man is further promoted to the RC external
head position, the surface word order of (32) will be as in (33)b.




40

32) the SR form
CP
¸ ¸¸ ¸
¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
DP-man ¸ ¸¸ ¸
+uWh,+Case vP T°
¸ ¸¸ ¸
book-ACC vP
¸ ¸ ¸ ¸
NP ¸ ¸ ¸ ¸
¸ ¸ ¸ ¸ VP v°
DP-man ¸ ¸ ¸ ¸ ¸ ¸ ¸ ¸
+uWh,-Case N° DP V°
¸ ¸ ¸ ¸ book bought
daughter+Agr






33) a. [[Ø
i
kız-ı] kitab-ı getir-en] adam
i

Ø girl-POSS.3S book-ACC bring-SR man
‘the man whose daughter brought the book’

b. *[kitab-ı [[Ø
i
kız-ı] getir-en] adam
i

book-ACC Ø girl-POSS.3S bring-SR man

In (32), the DP-specifier of the NP-subject, was attracted by the EPP of T, leaving the
NP-subject in its base-generated position inside the vP. This move seems to violate
Minimality because although the entire subject and the object are now specifiers of
the vP, and thus equidistant from T, it is not so clear that the element that is a
constituent of the subject is in the same minimal domain as the object.
Hornstein and Witkos’ (2001) analysis of transitive expletive constructions
(TECs) offers a possible solution. They argue that existential constructions are
formed by the merge of the expletive and the associate, and the overt movement of
the expletive to [Spec, TP]. Furthermore, what happens in TECs is that the object
and the [expletive-associate] pair are both specifiers of vP at the point when T merges

41

with the vP. TECs do not exist in English because movement of the expletive from
the [expletive-associate] pair will violate Minimality because the expletive (which is
a constituent of the [expletive-associate] pair) is not in the same minimal domain as
the object. On the other hand, in languages where another projection is available
above the vP, the [expletive-associate] pair can move to that position, from which the
expletive is now free to move without the issue of minimality.
For (32), one way around violating Shortest Move would be as follows: we
could say that although the DP within the NP is attracted by the EPP of T, the whole
NP, [
NP
[
DP
man-GEN D°] daughter], moves to T°. It is pied-piped by the DP-man.
This movement is allowed because the NP is equidistant to [Spec, T] being in the
same minimal domain as the fronted object now in [Spec, vP]. This movement is
similar to that of whose book in sentences such as “Whose book did you borrow?” In
this sentence, it is the element with the +Wh feature, whose, that is being attracted to
[Spec, CP], but pied-piping of the remnant of the category allows for the convergence
of a derivation that would otherwise crash with the movement of whose alone.
I repeat the relevant examples in (34). In (34)a, the specifier of the subject
receives case from T, whereas in (34)b, the entire DP subject receives case.

34) a. [[Ø
1
kız-ı ] kitab-ı getir-en] adam
1

Ø girl-POSS.3S book-ACC bring-SR man
‘the man whose daughter brought the book’

b. [[Ø
1
kız-ı]-nın kitab-ı getir-di-i] adam
1

Ø girl-POSS.3S-GEN book-ACC bring-NSR-3s man
‘the man whose daughter brought the book’


42

Adopting the Hornstein and Witkos proposal, it is the DP in the Spec of the NP-
subject in (34) (35)a that is attracted to T, and the NP-subject is pied-piped with it.
This strategy of avoiding the minimality violation noted for (32) is shown in (35):
When the DP-man in [Spec, NP] is attracted by the EPP of T, the entire NP, man’s
daughter, pied-pipes to [Spec, TP], as in move .

35) The SR form with pied-piping of the subject


CP
¸ ¸¸ ¸
+Wh-man ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
¸ ¸ ¸ ¸
vP T°
¸ ¸¸ ¸
book-ACC vP
¸ ¸ ¸ ¸
NP ¸ ¸ ¸ ¸
¸ ¸ ¸ ¸ VP v°
DP-man-GEN ¸ ¸ ¸ ¸ ¸ ¸ ¸ ¸
+uWh N° DP V°
¸ ¸ ¸ ¸ book bought
daughter+Agr






Another reason for proposing this sort of analysis, i.e. that it was only a constituent,
the +Wh-DP in the Spec of the subject, rather than the entire subject that was
attracted to T, is because had the subject been attracted by T’s EPP, it would have
received overt genitive case.
40
In the analysis being proposed here, the SR form is

40
I have not addressed the possibility of scrambling. One might argue that the entire subject has
scrambled to a position higher than the object but lower than [Spec, TP]. I reject this idea because
Kural (1992) and Kornfilt (2003), among others, shows that non-specifics in Turkish cannot scramble.
Persian shares many of the same phenomena regarding specificity, case and displacement as Turkish.
In her study of scrambling in Persian, Karimi (2005) also shows that scrambling is not possible for
non-specifics.

43

licensed when the expression that is Case-marked by T is also +Wh. Although the
entire NP-subject is presumably sitting in [Spec, TP], it is the +Wh-DP in its Spec
that is receiving case from T.

2.10 Relative clauses with complex arguments

We can now extend our analysis to RCs with even more complex arguments.
Relative clauses with sentential subjects permit both RC forms, as shown in (36).
41



36) a. [[Ø
1
biz-e güven-ece-i] üpheli ol-an] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS.3S doubtful be-SR man
‘the man who/such that (he) will trust us is doubtful’

b. [[Ø
1
biz-e güven-ece-i]-nin üpheli ol-du-u] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS.3S-GEN doubtfu be-NSR-POSS.3S man
‘the man who/such that (he) will trust us is doubtful’


Note that the subject in the RCs in (36) is something akin to “the fact that [the man]
trusts us”. I propose that the structure of these clauses is as in (37), where there is a
null element “fact” in Nº whose complement is the CP “that [the man] will trust us”.
42

It is thus the “factive” NP that receives the theta-role from the predicate “is doubtful”.

37) VP
¸ ¸¸ ¸
NP V°
¸ ¸ ¸ ¸ is.doubtful
¸ ¸ ¸ ¸
CP N°
_ _ _ _ (fact)
[the man] will trust us



41
A discussion of these examples first appeared in Csató (1985) and are again discussed in Barker,
et.al.(1990), and elsewhere.
42
This is akin to transforming a predicate of individuals, [CP], to a predicate of states, [CP – Nº].

44

This structure is similar to the null nominal head selected by factive verbs as analyzed
by Kiparsky and Kiparsky (1971). I assume that the +Wh head of the RC, man, is not
base-generated inside the “factive”-CP, but rather is first-merged as the specifier of
the NP. There is a null resumptive pronoun (RP) bound by man in the subject
position of the “factive”-CP. I will justify this assumption a little later; for now, note
that (38) better represents the structure of the RC subject for the RCs in (36).

38) VP
¸ ¸¸ ¸
NP V°
¸ ¸ ¸ ¸ is.doubtful
mani ¸ ¸ ¸ ¸
+uWh CP N°
¸ ¸ ¸ ¸ (fact)
¸ ¸¸ ¸
TP C°
¸ ¸¸ ¸
RPi ¸ ¸¸ ¸
VP T°
¸ ¸¸ ¸
us-DAT VP
¸ ¸¸ ¸
t-RP trust



Furthermore, the subject can be an NP as in (38) or the “fact”-clause can be
embedded in a DP
43
in which case the +Wh DP man will raise from [Spec, NP] to the
specifier of the DP (and get genitive case). When the subject is a DP, the entire DP-
subject is attracted by the EPP of T and assigned genitive case. Once this subject, [
DP

man [
NP
t-man [
CP
…] Nº ] D°] is in [Spec, TP] of the relative clause, the +Wh-
element, man, raises to [Spec, CP]. The derivation in (39), for the NSR phrase in

43
We would expect there to be a semantic difference based on whether the “factive” subject is an NP
or a DP. Some speakers do feel a slight difference but I have not been able to pin down a definitive
diagnostic that will yield consistent results. It is possible that the verbs that allow sentential subjects,
is doubtful, is certain, themselves cloud the semantic effects of the old information-new information
distinction that the NP-DP subject should entail, but see Section 2.11 where time adverbial scopal
differences are noted.

45

(36)b, demonstrates a RC with a DP subject. To be clear, as has been the case
throughout, we are looking at only the internal structure of RC’s, prior to the
promotion of the relativized element to the external head position.

39) the NSR form (36)b

CP
¸ ¸¸ ¸
+Wh-man ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
¸ ¸¸ ¸
VP T°
¸ ¸ ¸ ¸
DP V°
¸ ¸ ¸ ¸ is.doubtful
DP-mani ¸ ¸ ¸ ¸
+GEN,+Wh NP D°
¸ ¸ ¸ ¸
t-mani ¸ ¸ ¸ ¸
CP N°
¸ ¸ ¸ ¸ (fact)
¸ ¸¸ ¸
TP C°
¸ ¸¸ ¸
RPi ¸ ¸ ¸ ¸
VP T°
¸ ¸¸ ¸
us-DAT VP
¸ ¸¸ ¸
t-RP trust




The difference between clauses (36)a and (36)b can be attributed to whether the
subject is a DP or an NP. We saw that in (36)b the subject was a DP. In (36)a, on the
other hand, the subject is an NP which cannot satisfy the EPP and consequently does
not raise to [Spec, TP]. Again, the +Wh head of the RC is base-generated in the
specifier position of the NP-subject. As we saw in (31), a DP in the specifier of an
NP must raise to the spec of a case assigning head. This is demonstrated in the tree in
(40), for example (36)a: to avoid violating the Case Filter, the +Wh DP-man must

46

raise to [Spec, TP] out of the NP-subject. Note that in (40), pied-piping of the NP
subject is not required because there is no competing DP.

40) the SR form (36)a
CP
¸ ¸ ¸ ¸

¸ ¸¸ ¸
TP C°
¸ ¸¸ ¸
DP-man ¸ ¸¸ ¸
+Wh, +case VP T°
¸ ¸ ¸ ¸
NP V°
¸ ¸ ¸ ¸ is.doubtful
DP-man ¸ ¸ ¸ ¸
+Wh, -case CP N°
_ _ _ _ (fact)


One more assumption must be clarified: I assume that all relativized elements are DPs
because they must be specific, or have topic-like properties, and must be visible for
attraction by the EPP.
44
As DPs, relativized elements must be case-marked prior to
A-bar movement.
Let’s take stock and look at the possible options for subjects. A subject can
either be a DP or an NP. A possessor in its specifier can be either a DP or an NP.
When the subject is a DP, it must raise to a case-assigning head. It won’t matter
whether the possessor of a DP subject is an NP or a DP because an NP doesn’t
require case and a DP possessor will be assigned genitive case by the subject D°. On
the other hand, when the subject is an NP, the subject itself does not need case. If the
possessor in its Spec is also an NP, it too does not require case. However, a DP

44
Kayne (1994) proposes that in wh-relatives, the element that moves to [Spec, CP] is a DP headed by
a relative Dº, as in [
DP
which NP]. I adapt this analysis for Turkish which has neither overt
determiners nor overt complementizers, and assume that the relativized element in Turkish is a null
+Wh-Dº and its NP complement. See Bianchi (1999, 2000). Borsley (1997), although disagreeing
with Kayne’s raising analysis, demonstrates that the RC gap acts as a DP-trace with respect to
binding, licensing of parasitic gaps, and weak islands.

47

possessor in the Spec of an NP subject must raise for case or violate the Case Filter.
With this in mind, let’s look at derivations (43) and (44) for the RCs in (36).
Continuing derivation (40) in (41), we see why the SR is required in (36)a.
The +Wh DP-man in Spec of the NP-subject must raise to [Spec, TP] in to satisfy
T’s EPP and to receive case. The relative DP-man then moves to [Spec, CP] in .

41) the SR form (36)a

[
CP
[
NP
Ø
1
biz-e güven-ece-i ] üpheli ol-an ] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS.3S doubtful be-SR man
‘the man who [such that] (he) will trust us is doubtful’

CP
¸ ¸ ¸ ¸

+Wh-man ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
t-man ¸ ¸¸ ¸
+case VP T°
¸ ¸ ¸ ¸
NP V°
¸ ¸ ¸ ¸ is.doubtful
DP-man ¸ ¸ ¸ ¸
+Wh CP N°
_ __ _


By way of contrast, the subject of the relative clause in (36)b is a DP. Derivation (39)
repeated as (42) demonstrates that the entire DP-subject must raise to receive case.
The element in the Spec of the DP subject is assigned case by D°, and does not need
to move to an A-position for case. The expression in [Spec, TP], the entire subject, is
not a +Wh element—a constituent is—and as expected the NSR form is required.

42) the NSR form (36)b

[
CP
[
DP
Ø
1
biz-e güven-ece-i-nin] üpheli ol-du-u ] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS.3S doubtful be-NSR-POSS.3S man
‘the man who [such that] (he) will trust us is doubtful’

48


CP
¸ ¸ ¸ ¸
+Wh-man ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
¸ ¸¸ ¸
VP T°
¸ ¸ ¸ ¸
DP V°
¸ ¸ ¸ ¸ is.doubtful
DP-man-GEN ¸ ¸ ¸ ¸
+Wh CP D°
_ __ _




2.11 Semantic reflex of syntactic structure

In (41) and (42), we are making a claim about the category of the subject in the
matrix RC which has consequences in terms of the syntactic position the subject will
occupy. We would expect there to be some semantic reflex of the differences in
structure. This does seem to be the case as demonstrated in the examples 43) which
contain the time adverbial, Monday. The word order of the RCs in (43) is identical,
but in (43)a, the adverb can only modify the verb of the embedded RC subject,
whereas in (43)b Monday modifies the matrix RC verb.

43) a. [[Pazartesi [Ø
i
kız-ı-nın] Ankara-ya git-ti-i]-nin
Monday daughter-poss3s-GEN A.-DAT go-NSR-3s-GEN
anla-ıl-dı-ı] adam
i

understand-PASS-NSR-3s man
‘the man (such that) it was discovered that his daughter [went to Ankara on Monday]’

b. [Pazartesi [[Ø
i
kız-ı-nın] Ankara-ya git-ti-i] anla-ıl-an] adam
i

Monday daughter-poss3s-GEN A.-DAT go-NSR-3s understand-PASS-SR man
‘the man (such that) on Monday it was discovered that his daughter went to Ankara’


49

The derivations in (44) and (45) demonstrate where the difference in interpretation
comes from. Assuming the adverbial expression merges in T, there are two positions
in each derivation in which Monday can merge: in the embedded sentential subject or
in the RC (which I will call matrix) T. In (44), for the NSR in (43)a, if Monday had
merged in the matrix T, we would get the wrong word order because the sentential
subject will raise above Monday, yielding [[his daughter [ Ankara go]] Monday]. On
the other hand, Monday merging in T of the sentential subject will give us the right
word order but will yield an interpretation where Monday can only be interpreted as
modifying the embedded verb, the event of going to Ankara occurred on Monday.

44) the NSR form (43)a
CP
¸ ¸ ¸ ¸
+Wh-man ¸ ¸ ¸ ¸
TP C°
¸ ¸ ¸ ¸
k ¸ ¸ ¸ ¸
VoiceP T°
¸ ¸ ¸ ¸
¸ ¸ ¸ ¸
VP Voice° (Passive)
_ _ _ _
DPk V°
¸ ¸ ¸ ¸ discover
DP-mani ¸ ¸ ¸ ¸
+GEN,+uWh NP D°
¸ ¸ ¸ ¸
t-mani ¸ ¸ ¸ ¸
CP N°
¸ ¸ ¸ ¸ (fact)
¸ ¸¸ ¸
TP C°
¸ ¸¸ ¸
Monday TP
¸ ¸¸ ¸
DPj ¸ ¸ ¸ ¸
VP T°
¸ ¸ ¸ ¸
t-DPj ¸ ¸¸ ¸
¸ ¸ ¸ ¸ Ankara V°
RPi ¸ ¸ ¸ ¸ go
NP D°
daughter


50

As shown in the tree in (44), the time adverbial Monday raises with the entire subject
to a position higher than the RC verb discover, and crucially fails to c-command the
matrix (RC) verb. Compare this with the position of the time adverbial in the tree in
(45) for the SR example in (43)b where Monday remains below the matrix (RC) T.
In the SR form in (43)b, again the adverb can possibly merge in two positions, the
T of the embedded subject and the matrix (RC) T. As shown in (45), Monday
merging in matrix T will give us the correct word order and the interpretation that
Monday modifies the RC verb discover. As can be seen in the tree in (45), although
Monday c-commands the embedded verb (in the in situ sentential subject), it is too far
away to modify it; there is a VP, NP and CP between the time adverbial and the
Tense and verbal projections of the sentential subject.
45
The reading where the event
of going happened on Monday is not available in this structure.
Monday merging in the subject should conceivably also be possible, but my
informants were not able to get a reading where Monday would modify the embedded
subject verb “going to Ankara” with the word order of (43)b. The interpretation that
the going to Ankara will occur on Monday was only possible with the SR form with a
different word order within the clause of the sentential subject, one where the subject
of the sentential subject raises above Monday, yielding the phrase [[his daughter]
1

Monday [
vP
t
1
Ankara go]].




45
This inability for the time adverbial to modify the embedded verb is straightforward in a Phase-
based story. The complement of C is Spelled-Out when D merges in the structure of this complex
subject. This story supports Chomsky’s (1999/2001, and 2001) suggestion that that in addition to vP
and CP, DP is also a Strong Phase. Another way of looking at the inability of the adverbial to modify
the verb, would be to say that adverbial modification obeys Subjacency.

51

45) the SR form (43)b

CP
¸ ¸¸ ¸
+Wh-man ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
DP-man TP
+GEN,+Wh ¸ ¸¸ ¸
Monday TP
¸ ¸¸ ¸
VoiceP T°
¸ ¸¸ ¸
VP Voice° (Passive)
¸ ¸ ¸ ¸
NP V°
¸ ¸ ¸ ¸ discover
DP-mani ¸ ¸ ¸ ¸
+uWh CP N°
¸ ¸ ¸ ¸ (fact)
¸ ¸¸ ¸
TP C°
¸ ¸¸ ¸
DPj ¸ ¸¸ ¸
VP T°
¸ ¸ ¸ ¸
t-DPj ¸ ¸¸ ¸
¸ ¸ ¸ ¸ Ankara V°
RPi ¸ ¸ ¸ ¸ go
NP D°
daughter



For those speakers whose dialects permit both RC forms, there is no ambiguity in the
interpretations of the RCs in (43) when spoken with normal intonation.
46
That is, in
(43)a, the adverb can only modify the embedded verb of the sentential subject, and in
(43)b, the adverb can only be interpreted as modifying the matrix (RC) verb.
Let’s look at another example. The relative clauses in (46) are passivised
expressions. The phrase that was the complement of the RC verb, “that he will sell
his house”, has become the sentential subject of the RC. It is the subject of this
sentential subject that is the relative head. I have highlighted the adverb galiba

46
That is, without pauses. Pauses would denote a scrambled position similar to “We found out, [next
week (it is)]
1
, that the girl will go to Boston t
1
.”

52

‘apparently’ and the verb it modifies. Notice the interpretations: the adverb modifies
the RC verb in the NSR form in (46)a whereas the adverb modifies the sentential
subject verb in (46)b.
47


46) a. [ galiba [Ø
i
[pro
i
ev-i]-ni sat-aca-ı]-nın söyle-n-di-i ]
adam
i

apparently Ø pro house-AGR-ACC sell-FUTNSR-3s-GEN tell-PASS-NSR-3s man
‘the man
i
who that (he
i
) will sell his
i
house was apparently announced’

b. [Ø
i
[galiba [pro
i
ev-i]-ni sat-aca-ı] söyle-n-en] adam
i

Ø apparently pro house-AGR-ACC sell-FUTNSR-3s tell-PASS-SR man
‘the man
i
who that (he
i
) will apparently sell his
i
house was announced’

The example in (46)a is particularly interesting because even with changed intonation
and pauses, the adverb cannot be interpreted in the sentential subject. It seems galiba
cannot merge in TP. In (46)a, galiba has scrambled and adjoined to TP, but it must
be interpreted in its base position. Let’s take a closer look at the positions of galiba
in (47). We are still assuming that the sentential subject is a DP or NP Factive, but
for ease of exposition, I am abstracting away from much of the structure of this
phrase, and simply labeling it FDP or FNP. I am using the Ø symbol to denote the
+Wh-expression man with the understanding that this expression is generated in the
specifier of the Factive phrase, and is coindexed with the null resumptive pronoun
subject of the sentential subject. In the illustrations in (47), I have highlighted the
sentential subject in bold-type. In (47)a, galiba was generated in vP/PassiveP, and has
scrambled to TP adjoining above the sentential subject, FDP. In (47)b, the sentential
subject, FNP, is in the verbal domain, while the +Wh-expression has raised to [Spec,

47
Again, we are assuming normal intonation without pauses which would reflect scrambling or
discourse driven movement, such as Topic or Focus.

53

TP] and then to [Spec, CP]. As expected, (47)b is ambiguous and allows the
alternative reading where galiba modifies the RC verb ‘was announced’, but this
interpretation requires a pause after galiba.

47) a. [
CP
Ø
i
[
TP
galiba
j
[
FDP
Ø
i
[pro
i
evi]-ACC satacaı]-GEN [
vP/PassP
t
j
FDP söylen-NSR]]]






b. [
CP
Ø
i
[
TP
Ø
i
[
vP/PassP
[
FNP
Ø
i
[
vP
galiba [pro
i
evi]-ACC satacaı]] söylen-SR]]] adam
i



These examples seem to provide further support that the structure being suggested for
the two RC forms is on the right track.


3 A minimalist account: Pestesky and Torrego (2001)


In this section I would like to present one theoretical account for the facts in Table 1
within the framework of Minimalism (Chomsky (1995, 2000)). More concretely, I
will adopt a version of Pesetsky and Torrego (2001) (P&T).
P & T (2001) use the principle of Economy to explain the that-trace effect.
They begin by taking do-support in English as evidence that C must contain an
uninterpretable T feature, uT, which must be deleted by T. Using an adaptation of

54

Travis’s (1984) Head Movement Constraint, P&T assume that uT on C must be
deleted by movement of Tº to Cº.
P&T propose a tighter version of Chomsky’s (1995) Attract Closest F, Attract
Closest X (ACX), which prevents a head with multiple uninterpretable features from
targeting via Attract across any element that could potentially delete one of its
features. In effect, this constraint imposes ordering on feature checking of C. In
P&T’s system, C° always has an uT feature which can be deleted by head movement
of Tº to Cº. Interrogative Cº, and embedded Cº that hosts successive cyclic Wh-
movement, have an additional feature, uninterpretable Wh (uWh). Both these
features, uT and uWh, host an EPP feature. The features of C target the closest
element with matching features. The consequence of ATX is that C’s uT feature
must always be satisfied before its uWh feature; the local movement from the TP to
delete C’s uT feature, will always precede movement of another element to check C’s
uWh feature, unless this other element is just as close (i.e. within the same minimal
domain). When the +Wh expression of a sentence is not also the subject, ATX will
force T-to-C movement.

When two elements in TP can check different features of C°
(i.e. a +Wh subject that can check uWh and T° that can check uT), ATX applies
vacuously, and either element is a candidate for movement.
In sum, the uninterpretable features of C° can be satisfied by:
1- movement of T° to C° to check uT, and
2- movement of a Wh-element to [Spec, CP] to check uWh.
Because the TP is the closest projection to C, the movement in (2-) can precede the
movement in (1-) only when the +Wh-element is in [Spec, TP].

55

P&T explain lack of do-support, or T-to-C movement, in the sentence Who
bought the book? as follows. Motivated by a desire to unify nominative case on DP
and agreement on T, P&T argue that nominative case is, in fact, uT on D. Both T and
D have uninterpretable features that once checked by the other result in D properties
being borne on T, called “agreement”, and T properties borne on D, called
“nominative”. The outcome of this analysis is that a nominative DP is able to delete
the uT on C by moving to [Spec, CP] in the same way that T° to C° movement can.
P&T further assume that once a feature has been checked, it is “marked for
deletion” but remains “alive” for further operations until the end of the (strong) phase.
This means that although the uT features of the subject DP have already been checked
by T°, they are not deleted until the end of the CP phase. So, we need to revise our
summary above adding the additional way that C’s uT can be checked.
The uninterpretable features of C° can be satisfied in the following ways:
1- movement of T° to C° or movement of a nominative DP to [Spec, CP] to
check uT on C°, and
2- movement of a Wh-element to [Spec, CP] to check uWh on C°.
But, according to P & T, even though there are two ways to delete C’s uT, T-to-C
movement is required unless the element in [Spec, TP] is +Wh.
48

Beginning with the derivation for the sentence What did Mary buy? in (48),
let’s review P&T’s explanation of the subject/non-subject asymmetry of do-support in
English interrogatives.

48
The required head movement of T° to C° seems rather stipulative to me. P&T do not offer an
explanation as to why, for example, in English yes-no questions, the nominative subject cannot delete
uT on C° by moving to [Spec, CP]. I assume this is because a non-Wh expression is barred from the
CP projection, but see fn. 49 and 52.

56

Although there are two ways in which uT on C can be deleted, P&T’s analysis
(of sentences in which the Wh-element is not the subject) rests on the assumption that
in matrix interrogatives, uT is obligatorily deleted by head movement of T.
49
After T
moves to C, movement , the Wh-element moves to [Spec, CP] to check the uWh
feature of C, movement .

48) What did Mary buy?

CP
¸ ¸¸ ¸
¸ ¸ ¸ ¸
C° TP
did ¸ ¸¸ ¸
Mary ¸ ¸¸ ¸
T° VP
¸ ¸¸ ¸
¸ ¸¸ ¸
V° what
bought





Central to Pesetsky and Torrego is their Economy Condition (49) based on the
generalization that heads enter into Agree and Move relations only to the extent
necessary.


49
In discussing the ungrammaticality of “*What Mary bought” where the nominative subject has
deleted uT on C, P&T state: “The obligatoriness of T-to-C movement … might lead us to search for a
factor that favors T-to-C over subject movement... We suspect that this is not the right approach.”
P&T never explain why T-to-C movement is obligatory, and conclude the discussion with the
following statement: “We ... leave it as an observation for further research…. that movement of the
nominative subject to C is available as an alternative to T-to-C movement — even in matrix clauses
headed by a C that contains uWh. One factor that may explain this is unique specifier positions, that
is, if a nominative subject occupied [Spec, CP], movement of the +Wh-expression to [Spec, CP] would
not be possible with the result that neither C’s uWh, nor the +Wh-expression’s Wh features could be
checked. The evidence in Turkish certainly points to unique specifier positions in the functional
projections.

57

49) Economy Condition

A head H triggers the minimum number of operations necessary to satisfy
the properties (including EPP) of its uninterpretable features
50
.

This condition plays a crucial role in sentences where the Wh-element is the subject.
Recall that uT on C° can be deleted in one of two ways: i) by movement of the
nominative DP (carrying its still “alive” uT feature) from [Spec, TP], or ii) by head
movement of T° to C°. Even though both T° and the DP in [Spec, TP] are in the
same minimal domain and thus candidates for Attract Closest, the movement of T to
C is obligatory (see fns. 48 and 49). However, movement of the Wh-subject who
from [Spec, TP] to [Spec, CP] can check uT as well as uWh on C, as in derivation
(50)a. The Economy Condition disallows T-to-C in (50)b because movement of the
+Wh-subject results in a more economical derivation: all features of C are checked
with one move. The derivation which converges with less moves, (50)a, wins out
over an alternative derivation, (50)b, which requires more moves to check features.

50) a. Who bought the book?

CP
¸ ¸¸ ¸
¸ ¸ ¸ ¸
C° TP
¸ ¸¸ ¸
who ¸ ¸¸ ¸
+uT T° VP
+uWh ¸ ¸¸ ¸
¸ ¸¸ ¸
V° the book
bought



50
Note that, on this view, the EPP is an uninterpretable feature.

58

b. *Who did buy the book? (with normal intonation)

CP
¸ ¸¸ ¸
¸ ¸ ¸ ¸
C° TP
¸ ¸¸ ¸
who ¸ ¸¸ ¸
+uT T° VP
+uWh ¸ ¸¸ ¸
¸ ¸¸ ¸
V° the book
bought


P&T also use the Economy Condition to explain the that-trace effect. Rejecting the
traditional view that that is a complementizer merged as a sister to TP, they propose
that that is an instantiation of T-in-C. For P&T, the declarative C of embedded
clauses that hosts successive cyclic-wh-movement bears uT and uWh features, each
of which also bear EPP features. Again, when the Wh-phrase is not the subject, the
uT feature of C must be checked by T-to-C movement. In their system, C is null in
English, but T-in-C in this embedded environment is pronounced as that. Non-
subject wh-movement is demonstrated in derivation (51).
51


51) What
i
did John say [
CP
t-what
i
[
T
that]
j
+[C] [
IP
Mary will
j
buy t-what
i
]]?

In a sentence such as Who
i
did John say [
CP
t-who
i
[
TP
t-who
i
bought the book ] ]
where the Wh-word is the embedded nominative subject, movement of the subject to
the embedded CP can simultaneously check the uT and uWh features of C. Economy
dictates that this derivation be chosen over the less economical one where the
separate features of the embedded C are checked by two separate moves. More

51
Although not relevant for Turkish relatives, P&T account for ‘that deletion’ by allowing the
nominative embedded subject to delete uT on C, rather than via T-to-C, an option not available in
matrix clauses.

59

specifically, we can view Economy here as a local valuation: at a given point in the
derivation, choose the move that maximizes the number of features checked. T to C
movement, i.e. that to C, is precluded in this instance because movement of who to
[Spec, CP] checks more (in fact, all) features on C.
52


3.1 Pesetsky and Torrego (2001) analysis applied to Turkish
relative clauses

Although Pesetsky and Torrego did not address relative clauses, their analysis can be
extended to Turkish relative clauses. I assume that the Turkish NSR verbal
morpheme, –DIK, is a compound of a Tense morpheme, –DI, and the –K morpheme,
also found in uninflected infinitival verbs, as in (52). This view is supported by the
fact that the NSR form also allows the future tense as in (53)a. Note that –cE is the
future tense morpheme in matrix sentences (53)b.

52) a. gel -me -k b. ye -me-k
come-INF-K eat-INF-K

53) a. adam-ın gele-ce-i gün
man-GEN come-FUTNSR-3s day
‘the day the man will arrive

b. adam yarın gele-ce-k
man tomorrow come-FUT-3s
‘The man will come tomorrow.’


52
P&T suggest that the lack of T-to-C movement in sentences such as (i) is due to the absence of an
EPP feature on the embedded interrogative C in Standard English. They point to the dialectal
difference in the Belfast English example in (ii), and propose that, in (ii), uT on the embedded C is
deleted via movement, whereas in (i), it is accomplished via Agree.
(i) I wonder what Mary bought. (Standard English)
(ii.) I wonder which dish that they picked. (Belfast English)

60

I propose that -DI- is Tense specified for Past and that the -K morpheme is a reflex of
T to C movement, i.e. it signals T-in-C.
53
In the Pesetsky and Torrego story, a subject
that is being extracted contains both uT and uWh features. Movement of the subject
to [Spec, CP] deletes uT on C and renders the movement of T to C superfluous. The
derivation in (54)b of the clause in (54)a demonstrates that the nominative subject girl
can check both uT and uWh features of C.
If we assume that the -DIK (NSR) form is indicative of uT features of C being
checked by T to C movement, we would predict that whenever the subject is the
relative head, we would never see the -DIK (NSR) form because T to C movement
would be an additional unnecessary move. In the illicit (54)c, T to C movement has
deleted uT on C, which would have been deleted in any event by the obligatory
movement of girl to [Spec, CP] to check the uWh features of C (54)d.

54) a. [ hediye-yi ver-en ] kız
gift-ACC give-SR girl
‘the girl who gave the gift’


b. [ [C
+uT, +uWh
] [
TP
[girl
+uT, +uWh
] T [
VP
gave the gift]]]



c. *[ hediye-yi ver-di-ı ] kız
gift-ACC give-NSR girl
‘the girl who gave the gift’


d. [
CP
[Cº
+uT, +uWh
] [
TP
[
DP
girl
+T, +uWh
] T [
VP
bought the gift ]]]





53
For a detailed discussion, see Kural (1993) who also argues that -DI is past tense and -K is Cº. Note
that for me the -K morpheme is specifically an instantiation of Tº in Cº.

61

P&T’s analysis can be straightforwardly applied to account for the NSR -DIK
morpheme in non-subject RCs. The C head simply targets the closest head,
specifically T, to check its uT feature. There is no +Wh element in [Spec, TP] to
outcompete T-to-C.

3.2 The NSR -DIK form

Let us now go back to the NSR form, No. 2 and No. 3 in Table 1. Using P&T’s
intuition, we can see how Economy accounts for the unacceptability of the NSR form
in sentence (2)b repeated with its derivation in (55).

55) *[Ø
i
divan-da otur-du-u] bayan
i

Ø sofa-LOC sit-NSR-3s lady
Intended: ‘the lady that is sitting on a/the sofa’


CP
¸ ¸ ¸ ¸
lady ¸ ¸ ¸ ¸
TP C°
¸ ¸ ¸ ¸
t-lady ¸ ¸ ¸ ¸
+uT VP T°
+uWh ¸ ¸ ¸ ¸
DP ¸ ¸¸ ¸
t-lady PP/DP V°
+uT _ _ _ _ sit
+uWh on the sofa



In (55), the NSR -DIK morpheme, an instantiation of T having moved to C, is
unacceptable. The subject lady is attracted to [Spec, TP] by T’s EPP feature,
movement , and is assigned nominative case. The subject has both uT and uWh

62

features; in movement of the subject to [Spec, CP] deletes these two features.
Movement of T° to C° is redundant, and so disallowed.
On the other hand, the NSR -DIK morpheme is acceptable in (3)a, repeated as
in (56), because when the extracted element is not the subject, there is no alternative,
more economical, move.

56) [bayan-ın Ø
i
otur-du-u ] divan
i

lady-GEN Ø sit-NSR-3s sofa
‘the sofa that the lady is sitting on’

CP
¸ ¸¸ ¸
sofa ¸ ¸ ¸ ¸
+uWh TP C°
¸ ¸¸ ¸
lady ¸ ¸¸ ¸
+uT VP T°
¸ ¸ ¸ ¸
DP ¸ ¸¸ ¸
t-lady PP/DP V°
+uT sofa-LOC sit
+uWh




In the tree in (56), again the subject lady is assigned nominative case by T and is
attracted to [Spec, TP] to satisfy T’s EPP feature, movement . Because the subject
does not have a +Wh feature it cannot check uWh on C. Movement of the +Wh DP
sofa to [Spec, CP] is required. This derivation requires two moves to delete both uT
and uWh on C. The movement of T to C is obligatory for convergence as it must
check C’s uT feature. When a non-subject is being extracted, there is no alternative
derivation that would converge with fewer moves.


63

3.3 The SR form

We saw that in a subject relative, Economy dictates use of the -An SR form and bars
the -DIK NSR form. As demonstrated in (55), movement of the +Wh-subject deletes
both features of C simultaneously, making T-to-C movement unnecessary. Thus (55)
is aptly the derivation for the SR sentence (2)a repeated in (57).

57) [Ø
i
divan-da otur-an] bayan
i

Ø sofa-LOC sit-SR lady
‘the lady who is sitting on the sofa’

To be clear, P&T rely on Shortest Move to explain the obligatory movement of T to
C (rather than the alternative nominative-DP movement to [Spec, CP]) when the
subject is not a Wh-element, and they rely on Economy to disallow T to C movement
when the subject is a Wh-element, with all moves driven by the EPP. For P&T, the
crucial feature is uT on C.
Whereas the P&T analysis explains the asymmetry in simple relative clauses
in Turkish, it cannot account for the exceptions (numbers 4 and 5 in Table 1) nor can
it explain the optional cases. We will need to make adjustments to the P&T analysis.
Let’s begin with the first complication. How do we explain the acceptability of the
SR form when a non-subject is relativized, as in the SR example (5) repeated as (58)
(item 4 in Table 1)? Rather than T moving to C, the +Wh-element in the verbal
domain has moved to C; this violates Shortest Move in P&T and should be banned.

58) a. [gemi yana-an] liman
ship sidle-SR harbor
‘the harbor that a ship is sidling up to’

64

b.
CP
¸ ¸¸ ¸
harbor ¸ ¸ ¸ ¸
+Wh TP C°
¸ ¸ ¸ ¸
¸ ¸¸ ¸
VP T°
¸ ¸ ¸ ¸
NP ¸ ¸¸ ¸
ship DP V°
t-harbor+DAT sidle.up.to
+Wh


In section 1.3, we saw that Enç’s (1991) work was of interest because she
demonstrated the correlation between specificity, raising and overt case-marking in
Turkish. In sum, non-specific nominal elements neither get overt case nor do they
raise from their base-generated position whereas specific nominal expressions must
raise and must bear overt case.
54
Based on this complementarity, I assumed an
NP/DP division based on specificity: in Turkish, non-specific nominals are NP’s that
lack a D projection.
55
Conversely, specific nominals are DP’s.
Furthermore, I consider the reciprocity between obligatory case assignment by
functional heads (the Inverse Case Filter) and the Case Filter an unnecessary
redundancy. I therefore assume that all DP’s must have their case checked/assigned
but that all case-assigning heads need not necessarily discharge their case
56
. In a
sentence that has no DPs as arguments, T° (and v°, for transitives) will have no DP on

54
We saw in (6) that case-marked objects must raise; (9) and (10) demonstrate that the same is true for
subjects.
55
Turkish lacks determiners, so the D head in DP’s must be null. Longobardi (1994) provides evidence
for DP’s with a null D in Italian. I assume a similar structure for Turkish.
56
I realize this complicates the issue of Full Interpretability. An uninterpretable Case feature on T that
remains unchecked or undeleted cannot lead to a derivational crash under this view. There is the
possibility that T does assign Case to whatever element sits in its Spec, and because the EPP of T must
be satisfied for convergence, T will always be assigning Case. I will discuss this option in more detail
later.

65

which to discharge its case feature. An example is the Turkish sentence in (59)a in
which the sole argument, the subject, being non-specific, must be an NP as in (59)b.

59) a. (Bir) köpek havla-dı.
one dog bark-PST
‘A dog barked.’

b. [
TP
[
VP
[
NP
dog] bark ] -PAST]


The assumption that convergence requires that DP’s receive case but not that relevant
heads assign case enables us to account for the acceptability of sentences with NP
arguments. The question remains though, how is the EPP of T checked in (59)a? As
demonstrated in (59)b, the subject, being an NP, has not raised to T. Note that this
sentence would be quite odd without a context, for example in response to the
question, “What happened?” Crucially, this question entails a contextually relevant
time and place. Therefore, I assume the sentence in (59)a contains a pro-form
locative, as in the response in (60)a. It is the pro-locative that raises to T and checks
its EPP feature, as in (60)b.
57
At this point, I am not assuming that T assigns
Nominative Case to the Locative, but I will return to this in greater detail later.

60) a. Question: (Sokak-ta) ne ol-du?
street-LOC what happen-PST
‘What happened (in the street)?’

Answer: [ora-da] (bir) köpek havla-dı.
there-LOC one dog bark-PST
‘A dog barked [there].’

b. [
TP
pro-there-LOC [
VP
[
NP
dog] bark ] PAST]


57
This is analogous to locative inversion in Spanish, where there is evidence to suggest that because a
bare NP subject cannot raise to a preverbal position, a locative must raise to satisfy the EPP of T. See
Ortega (200).

66

In Section 1.3 of Chapter 2, we saw the correlation between displacement, specificity
and case. Recall that a specific object must obligatorily raise and receive overt case,
whereas a non-specific one does neither. Nominative case in Turkish is the Ø-
morpheme, but similar facts were demonstrated for subjects in embedded
environments where the subject is marked with overt genitive case. In the sentences
in (61) and (62), note both the position of the subject and its case: no raising nor case
on the subject when it is non-specific, obligatory raising and case when it is specific.
These facts provide evidence that all DPs not only receive case, but must also raise to
receive/check case, presumably because case is assigned in a Spec-Head
configuration.
58


61) a. Sokak-ta köpek havla-dı-ın-ı duydum.
street-LOC dog bark-NSR(COMP)-AGR-ACC hear-PST-1s
‘I heard a dog barked in the street.’

b. [pro [
vP
[
CP
[
TP
street [
VP
[
NP
dog] bark ] ] ]-ACC heard]


62) a. Köpe-in sokak-ta havla-dı-ın-ı duydum.
dog-GEN street-LOC bark-NSR(COMP)-AGR-ACC hear-PST-1s
‘I heard that the dog barked in the street.’

b. [pro [
vP
[
CP
[
TP
[
DP
dog]-GEN [
VP
street bark ] ] ]-ACC heard]


58
Referring to Chomsky (2001), Boeckx (2001) points out that not every DP can satisfy the EPP; the
DP must be “featurally related” to the EPP bearing head. In Chomsky (1995), EPP driven movement
is comprised of Attract F (head adjunction of formal features FF) followed by pied-piping of the
category for PF convergence (so that the category will be “close enough” to its FF so that the features
of the category will not be scattered). In Chomsky (2001), the Spec-Head relation is considered an
outcome of Move which is defined as Agree + Pied-piping + (internal) Merge. An important
generalization is that in Turkish a DP receiving structural case must raise to the specifier of the case-
assigning head. However, I concur with Boeckx (2001) that Agree need not be a prerequisite for
Move, thus allowing for the possibility of Move to take place in some cases under Match, a looser
requirement of “feature-relatedness” (see fn. 59).

67

Deviating from Pesetsky and Torrego, I propose that in Turkish relative clauses the
EPP feature of T is simply a feature of some uninterpretable feature of C. That is, if
C is in the derivation, it has an uninterpretable T feature that makes it select T.
59
By
the same token, T also has a comparable uninterpretable feature that must be checked.
Let us say that this feature is some sort of Wh- (i.e. A-bar) feature
60
because it was
selected by C. For simplicity, let us call the matching features uT when on C, and uC
when on T. When a non-Wh-element is in [Spec, TP], T must necessarily move to C
because the uC feature it bears is still unchecked. On the other hand, a Wh-element
in [Spec, TP] is able to delete/check the uC feature on T
61
, after which it will raise to
[Spec, CP] and delete/check uWh on C. The proposal for Turkish then is as in (63).

63) Theoretical Assumptions

i. The EPP feature on T is a feature of an uC feature; T has an EPP feature
only when it has been selected by C.
62

ii. The uC feature of T may be checked/deleted by movement of a +Wh-
element to [Spec, TP] or by movement of T° to C°.
iii. T can assign case but need not.
63


59
In Turkish matrix sentences, the specific subject has topic-like properties. Thus these sentences may
have a Topic projection which selects T with an uninterpretable “Topic” feature. I assume that
sentences such as (59) have neither a Topic projection nor an EPP feature on T. This contrasts with the
embedded sentence in (61), where absent a DP requiring structural case, a featurally-related DP, the
locative, raised to T to satisfy its EPP feature.
I avoid a discussion as to whether v may or may not have an EPP feature. Following the reasoning
for Tº, however, there may be an optional A-bar “topic-like” projection above vP that selects for v with
an EPP feature. Although not a topic of this dissertation, the data we incur in later chapters suggests
there may indeed be an EPP on vº as well as an A-bar projection above vP. I will point these out as we
encounter them. Certainly, this issue requires further inquiry.
60
The EPP of this feature attracts a D; Wh-features are irrelevent for the operation Attract by T.
61
It does not concern us here as to exactly how uC on T is deleted by a Wh-element in [Spec, TP]. One
can imagine several plausible ways, none of which will detract from the proposal here.
62
This also holds for sentences with TopicP, that is, Topic° will select for T° with an uninterpretable
A-bar-like feature which itself will have an EPP feature, but see fn. 59.

68

iv. Nominals that are specific are DP’s; non-specifics are NP’s.
v. Only DP’s need case.
vi. NP’s cannot be attracted by the EPP and so are invisible for movement.

Let’s look closer at the consequences of assumption (ii) above. We know that T’s
EPP attracts the closest DP. If this DP has a +Wh feature, it satisfies two features: uC
on T and the EPP on uC. If the closest DP is non-Wh, the only way for T to
discharge its uC feature is by movement from T to C. This is what is required for T°.
The C° of a RC has an interpretable Wh feature to be checked. T-to-C movement
will delete uC on T, but not uWh on C° which still must be satisfied by movement of
a +Wh-expression to [Spec, CP] checking uWh and satisfying C’s EPP feature.
With this new approach, let’s look again at the SR in (57), repeated as (64).

64) [Ø
i
divan-da otur-an] bayan
i

Ø sofa-LOC sit-SR lady
‘the lady that is sitting on the sofa’

CP
¸ ¸¸ ¸
lady ¸ ¸ ¸ ¸
+Wh TP C°
¸ ¸ ¸ ¸
lady+GEN ¸ ¸ ¸ ¸
+Wh VP T°+uC
¸ ¸ ¸ ¸
DP ¸ ¸¸ ¸
t-lady DP V°
+Wh sofa-LOC sit



63
Chomsky (1999) refers to non-finite T which has an EPP feature but cannot check case, as in (i), as
T
def
.
(i) We expect there to be awarded several prizes.
For Turkish, I assume that finite T, when selected by Cº (or Topicº), has an EPP feature but need not
necessarily assign case. As shown in (61), a (inherently) case-marked element may be attracted by the
EPP of T. The exact nature of T’s case-assigning properties is not crucial here, thus I will remain
agnostic as to whether T actually assigns null nominative case to the already case-marked DP in its
spec, or the element being case-marked and “featurally-related” is able to delete T’s uninterpretable
case feature. (I take up the question of whether Tº must assign case in Chapter 4.)

69

In the derivation in (64), T has an uC feature and an EPP feature that must be
satisfied. In step , the subject lady is attracted to [Spec, TP] satisfying T’s EPP
feature. Because the subject has a +Wh feature, the uC feature on T has been deleted.
C merges with the TP, and its EPP and uWh features are checked by movement of the
subject lady to [Spec, CP], step . Movement of T° to C°, step , is unnecessary,
and so disallowed.
This contrasts with the NSR sentence in (56), repeated as (65), where the
subject is not a Wh-element and so cannot delete uC on T. T to C movement as in ,
is required for this derivation to converge.

65) [bayan-ın Ø
i
otur-du-u ] divan
i

lady-GEN Ø sit-NSR-3s sofa
‘the sofa that the lady is sitting on’

CP
¸ ¸¸ ¸
sofa ¸ ¸ ¸ ¸
+uWh TP C°
¸ ¸ ¸ ¸
lady ¸ ¸¸ ¸
+uT VP T°+uC
¸ ¸ ¸ ¸
DP ¸ ¸¸ ¸
t-lady PP/DP V°
+uT sofa-LOC sit
+uWh




3.4 Recap: our new story

Before we proceed to relative clauses with non-specific subjects, let’s review how we
have deviated from Pesetsky and Torrego (2001). In the Pesetsky and Torrego
analysis, the driving feature was on the C head. Notice that we have shifted the
perspective from uT features on C to uC features on T. After all, it only makes sense

70

that if there is a C projection, C will select a T that will ensure that features required
by C will be checked
64
. Thus relative clauses in Turkish have the following
functional categories: a C° with uWh and EPP features, and a T° with uC and an EPP
feature. Note that only a +Wh-expression can move to [Spec, CP]; [Spec, TP] has no
such restriction. In this story, T can be thought of as a hybrid, an A projection
bearing Wh-like features (which perhaps is the reason, when necessary, T can
undergo head-movement to C). uC on T can be checked/deleted either by movement
of a +Wh DP to [Spec, TP], or by head movement of T to C. The cycle ensures that
the former be the unmarked case, and that T to C is Last Resort-like.
65
The reason we
do not get the NSR -DIK morpheme (indicating T-in-C) when a subject is being
relativized, is that there is no motivation for T to move to C; all of T’s features have
been checked.
66
In this sense, T is Greedy: it only moves to check features on itself;
it cannot check any features on C.
These divergences from P&T (2001) enable us to explain the use of the SR
form in RC’s with no external argument as well as those with a non-subject gap in
clauses a non-specific subject.

64
This is different from S-selection; C° cannot check its features against T° merely by virtue of this
selection. This kind of selection is analogous to the Force projection (Rizzi 1997) of C selecting a
finite or non-finite T.
65
It does not really matter if T moves to C as soon as C merges with TP or after the +Wh element
moves to [Spec, CP]. Arguments could be made for both alternatives. The point is that uC on T must
be checked preferably within the TP projection or at the latest, by the next projection.
66
The implications for matrix sentences are that all sentences with specific subjects must have a CP
layer if we are to assume that DP subjects raise from their base-generated positions. We saw in (26)
and (27) that T° in sentences with non-specific subjects does not have an EPP feature. It is not far
afield to assume that matrix sentences with specific subjects in Turkish have a Topic projection that
selects for a T° with an uC feature that has an EPP feature. Thus sentences with non-specific subjects
are TP’s, whereas sentences with specific subjects have an A-bar projection, TopicP. It follows that
the same assumption must be made for the vP layer: there is some kind of “Topic-like” projection
above the vP which selects for v° with an EPP feature which attracts the DP (specific) object.
Otherwise, there is no extra projection; v° does not have an EPP feature and the NP (non-specific)
object remains in situ.

71

3.5 Clauses that lack external arguments: the SR form

We can now see why clauses that lack external arguments, item 5 in Table 1, as in
(4)a repeated as (66)a, require the SR form. I have included the comparable matrix
sentence in (66)b to demonstrate that neither of the two nominals in the sentence,
“this stop” and “bus”, requires structural case: the PP ‘from this stop’ is rendered in
Turkish as the nominal this stop with ablative case, and bus receives inherent dative
case. As shown in derivation (67), there are two (equidistant and non-structurally
case-marked) DP’s which can satisfy the EPP of T. Because one of them is +Wh,
Economy will choose movement of this element to [Spec, TP]. This move will delete
uC on T, bleeding T to C movement.

66) a. [Ø
i
bu durak-tan bin-il-en] otobüs
i

Ø this stop-ABL board-PASS-SR bus
‘the bus which is boarded from this stop’

b. Otobüs-e bu durak-tan bin-il-ir.
bus-DAT this stop-ABL board-PASS-AOR
‘The bus is boarded from this stop.’

67) 1- the +Wh-element bus merges with the verb board forming the VP:
[
VP
bus board]

2- the PP/DP this stop merges with (adjoins to) the VP:
[
VP
this stop [
VP
bus board]]

3- T° merges with this VP: [ T° [
VP
this stop [
VP
bus board ]]]

4- T° has uC with an EPP feature; there are only non-argument DP’s available.
Economy dictates that the +Wh DP move to [Spec, TP]:
[
TP
bus
i
T° [
VP
this stop [
VP
t
i
board ]]]

5- C° merges with the TP:
[ C° [
TP
bus
i
T° [
VP
this stop [
VP
t
i
board ]]]]

6- C’s uWh and EPP features are checked by movement of +Wh bus to its Spec:
[
CP
bus
i
C° [
TP
t
i
T° [
VP
this stop [
VP
t
i
board ]]]]

72

Let’s take a closer look at step 4 in (67). How might Economy dictate movement of
one DP over another? The Pesetsky and Torrego analysis provides us an answer. In
the tree in (68), there are two DP’s inside the VP. Being in the same minimal
domain, they are equidistant from the point of view of T’s EPP. Movement , of the
+Wh-DP to T, will check uC on T, and in , the element moves to [Spec, CP] to
check uWh on C. Only two moves are necessary to check all the features on T and C.

68)
CP
¸ ¸¸ ¸
+Wh-DP ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
+Wh-DP ¸ ¸¸ ¸
VP T°
¸ ¸¸ ¸ +uC
+Wh-DP VP
¸ ¸¸ ¸
DP V°


Compare with the alternative scenario: the equidistant, non-Wh DP is attracted to T,
as in (69). Although this element satisfies T’s EPP feature, , T to C movement is
required to delete uC on T, . The +Wh DP must also move to [Spec, CP] to check
the uWh feature on C, . This derivation requires three moves to check all features
on both T and C.

69)
CP
¸ ¸¸ ¸
+Wh-DP ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
DP ¸ ¸¸ ¸
VP T°
¸ ¸ ¸ ¸ +uC
+Wh-DP VP
¸ ¸¸ ¸
DP V°




73

Derivation (68) is preferable to the derivation in (69) also because of Shortest Move.
Both DP’s are equidistant from T. In (68), the +Wh DP took two short moves to
[Spec, CP]. In (69), the movement of the +Wh DP to [Spec, CP] in , is longer
move, i.e. crosses more nodes. All things being equal, the system prefers a derivation
where features—in this case, uWh on C—can be satisfied with shorter moves than
with longer ones.
Looked at another way, derivation (68) will trump (69) when evaluated
locally, i.e. by the amount of work accomplished at each point in the derivation.
When T is merged, it has two features that must be checked/deleted: uC and its EPP.
From the point of view of T, it is more economical to attract an element that will
satisfy both these features as soon as possible and at once, by attracting a +Wh
element to its specifier. The alternative is for T to attract a non-Wh-DP to its Spec
and then to undergo head movement to C to delete its uC feature. When the DP’s
providing both options are equidistant, the system will choose the more efficient
option.
67


67
This was noted in Hankamer & Knecht (1976) with the following examples:
(i) [Kapı-nın alt-ın]-dan [yer-in üzer-i]-ne su akıyor
door-GEN bottom-AGR-ABL floor-GEN top-AGR-DAT water is.flowing
‘Water is flowing from under the door onto the floor.’
(Literally: ‘From the door’s bottom, onto the floor’s top, water is flowing’)
(ii) [[Ø alt-ın]-dan yer-in üzer-ine su ak-an] kapı
bottom-AGR-ABL floor-GEN top-AGR-DAT water flow-SR door
‘the door that water is flowing under from onto the floor’
(iii) [kapı-nın alt-ın-dan [Ø üzeri]-ne su ak-an] yer
door-GEN bottom-AGR-ABL top-AGR-DAT water flow-SR door
‘the floor that water is flowing from under the door onto’
They conclude “...when there are two oblique phrases in an indefinite-subject construction, the [SR] is
used no matter which contains the target of relativization. ... In fact, no matter what is relativized out of
a clause with an indefinite subject, the RC is constructed with the [SR].” The principle of Economy
explains why this is so. Any or all of the Economy measures sited will give preference to the SR form
over the NSR.



74

3.6 Relative clauses with complex arguments

We are now ready to extend our analysis to RCs with even more complex arguments.
We saw in (36), now repeated as (70), that relative clauses with sentential subjects
permit both RC forms.

70) a. [[Ø
1
biz-e güven-ece-i] üpheli ol-an] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS.3S doubtful be-SR man
‘the man (such) that (he) will trust us is doubtful’

b. [[Ø
1
biz-e güven-ece-i]-nin üpheli ol-du-u] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS.3S-GEN doubtful be-NSR-POSS.3S man
‘the man (such) that (he) will trust us is doubtful’

I had argued in section 2.10 that the subject in these RCs is something akin to “the
fact that [the man] trusts us” or “such that [the man] trusts us” and that the structure
of these subjects is comprised of a a null “fact” in Nº whose complement is the CP
“that [the man] will trust us”. Recall that it was this “factive” NP that received the
theta-role from the predicate “is doubtful”.
Recall further that I had assumed that the subject could be an NP or that the
“fact”-clause could be embedded in a DP. The tree in (71) for the RC in (70)b
demonstrates the derivation of a relative clause with a “factive” DP subject. The EPP
of T attracts the entire DP-subject, [
DP
man [
NP
t-man [
CP
…] Nº ] D°]. Once this
subject is in the [Spec, TP] of the relative clause, the +Wh-element, man, from within
the subject raises to [Spec, CP]. Since the subject is not a +Wh element, uC on T
remains unchecked and T to C movement is required for convergence.



75

71) the NSR form for RC (70)b

CP
¸ ¸ ¸ ¸
+Wh-man ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
¸ ¸¸ ¸
VP T°+uC
¸ ¸ ¸ ¸
DP V°
¸ ¸ ¸ ¸ is doubtful
DP-mani ¸ ¸ ¸ ¸
+GEN,+uWh NP D°
¸ ¸ ¸ ¸
t-mani ¸ ¸ ¸ ¸
CP N°
¸ ¸ ¸ ¸ (fact)
¸ ¸¸ ¸
TP C°
¸ ¸¸ ¸
RPi ¸ ¸¸ ¸
VP T°
¸ ¸¸ ¸
us-DAT VP
¸ ¸¸ ¸
t-RP trust







This derivation contrasts with the tree in (72) for the RC in (70)a where the subject is
an NP which cannot satisfy the EPP of T. Again, the +Wh head of the RC is base-
generated in the specifier position of the NP-subject. As we saw in (31)d, a DP in the
specifier of an NP must raise to the spec of a case assigning head or it will violate the
Case Filter. Thus the +Wh DP-man must raise to [Spec, TP], , out of the NP-
subject for case or the derivation will crash. At the same time, this +Wh expression
deleted uC on T, rendering T to C movement unnecessary. The DP-man then moves
to [Spec, CP] to delete uWh on C, in .



76

72) the SR form for RC (70)a

CP
¸ ¸¸ ¸

+Wh-man ¸ ¸¸ ¸
TP C°
¸ ¸¸ ¸
DP-man ¸ ¸¸ ¸
+uWh, +case VP T°+uC
¸ ¸ ¸ ¸
NP V°
¸ ¸ ¸ ¸ is doubtful
DP-man ¸ ¸ ¸ ¸
+uWh, -case CP N°
_ _ _ _ (fact)


4 Conclusion


In this chapter, we saw an explanation for the two different relative clause forms in
Turkish that accounts for their distribution. The SR form can be explained by the
generalization that it is licensed whenever the +Wh relativized expression moves into
[Spec, TP] and then to [Spec, CP]. I demonstrated a correlation between specificity
and displacement, and proposed a DP/NP dichotomy that would capture the facts. I
argued for an EPP feature on T, and suggested a restricted definition of the EPP and
the Case Filter, such that these apply only to DPs. Furthermore, we saw evidence that
D°, but not N°, assigns case.
Although valuable, the account of Wh-movement proposed by Pesetsky and
Torrego (2001) failed to adequately explain the Turkish facts. I suggested a
modification of the features implicated by (P&T) that while still capable of handling
the Pesetsky and Torrego facts, is better able to account for the Turkish data. Rather
than nominative case on C, it is an uninterpretable C feature on T that needs to be

77

checked. With this revised account, plus the assumptions about DPs, Case and the
EPP, we were able to predict the distribution of RC forms. In addition, we had a
vehicle for explaining the seeming optionality in RCs with complex subjects.
This chapter provides us with two things. First, we have the beginnings of a
diagnostic for movement to [Spec, TP]. If we are right that the SR from is licensed
only when a +Wh expression moves through [Spec, TP], then, every time we have an
SR clause, we know the relative head must have A-moved to [Spec, TP]. This gives
us a vehicle for testing conditions of A-movement in Turkish sentences. Second, as
we proceed to apply our diagnostic, we will encounter issues that have theoretical
import. We have mentioned some of these topics above. There are others that we
have not yet addressed; for example, it seems that structurally case-marked elements
are “frozen” because EPP effects in Turkish seem to be sensitive to structural vs.
inherent case on DPs. In Chapter 4, we will also look into intervention effects. As
we apply the SR diagnostic, we will examine the A-movement to T and the effects of
Minimality in constructions with different verb classes, with psych verbs, and in
infinitival structures. The hope is that as we formulate explanations, the ideas will
simultaneously be specific enough to provide an accurate account of Turkish clauses,
and general enough to provide insight into the grammar of Natural Language.


78

Chapter 3: Specificity

1 Introduction


In Chapter 2, I made several proposals about specificity, case marking and the
structure of nominals. In this chapter we review some of these assumptions and the
arguments in their favor.
We saw that Turkish does not have overt determiners. Information about
specificity is encoded by case morphology and by displacement. For example, the
sentences in (1) demonstrate a direct object with and without overt case morphology.
Only when the object ‘book’ is marked with accusative case, as in (1)b, can it have
(indeed, must have) a specific interpretation.

1) a. Ali kitap oku-du.
Ali book read-PST
‘Ali read a book.’

b. Ali kitab-ı oku-du.
Ali book-ACC read-PST
‘Ali read the book.’

Overt case morphology is not the only phenomenon correlated with a specific
interpretation. The sentences in (2) establish that case-marked objects are in a
different structural position than their bare counterparts. Assuming that Turkish
adverbs of manner mark the left edge of the VP
68
, sentence (2)c demonstrates that a

68
See Kural (1992).

79

case-marked direct object cannot remain inside the VP, while an object without overt
case must remain inside the VP (2)a-b.
69


2) a. Ben hızlı kitap oku-r-um
I quickly book read-AOR-1sg
‘I read books quickly.’

b. *Ben kitap hızlı oku-r-um
I book quickly read-AOR-1sg

c. *Ben hızlı kitab-ı oku-ru-m
I quickly book-ACC read-AOR-1sg
‘I read the book quickly.’

d. Ben kitab-ı hızlı oku-ru-m
I book-ACC quickly read-AOR-1sg
‘I’ll read the book quickly

This is not actually the whole story. To understand the behavior of Turkish nominals,
one must distinguish between those expressions that merge into theta positions
without case, and those that are inherently or lexically case-marked. Thus, nominals
that are inherently case-marked, trivially always bear overt case and are ambiguous in
terms of their specificity.
70
For clarity, throughout this chapter, I will refer to
nominals that enter a derivation without case as ‘arguments’, admitting that
technically a Dative or Ablative expression is also an argument of the verb. As we
will see, there are behavioral differences both in the syntax (and at LF and PF)
between nominals that require structural case and those that are inherently case-
marked. For expository purposes, then, I will use the term ‘argument’ to refer to

69
Examples from Aygen-Tosun (1999).
70
See Kornfilt (2003).

80

those expressions that are typically referred to as the internal argument and the
external argument, and which would be assigned structural case by v or T.
Returning to our examples in (1) and (2), we saw that specific direct objects
must both raise and bear overt case, and non-specifics cannot raise and are bare. This
regularity provides us with a useful tool: we can safely use the presence or absence of
overt case as an indication of whether or not the object has raised from its base-
generated position.
This correlation is difficult to demonstrate for subjects in matrix clauses
(nominative case being the null morpheme), but we do observe the same pattern in
subjects in embedded clauses where a specific subject must bear overt genitive case.
For example, the subject of so-called “factive”
71
clauses in Turkish has genitive case,
as in (3). However, the non-specific subject of the embedded existential construction
in (4)a does not have case morphology. Compare with (4)b in which the embedded
subject has raised above the locative and is case-marked, and must receive a partitive
(i.e. specific) interpretation. Thus, (4)b is analogous to (3) where the subject has
raised to [Spec, TP] and bears case, whereas in (4)a we get the existential reading
because the subject has remained in situ while the locative has raised to [Spec, TP] to
check T’s EPP feature.

3) pro [Ali-nin hasta ol-du-u]-nu söyle-di-ler.
pro Ali-GEN sick be-NSR-3s-ACC say-PAST-3pl
‘They said that Ali was sick’



71
This is the term used in the literature (e.g. Kornfilt 1997) to identify complement clauses with the
NSR –DIK verbal morpheme, and not to be confused with factive clauses a la Kiparsky (1971).

81

4) a. pro [yan-ın-da bir yılan ol-du-u]-nu söyle-di-ler.
pro side-his-LOC one snake be-NSR-3s-ACC say-PAST-3pl
‘They said that there was a snake by his side’

b. pro [bir yılan-ın yan-ın-da ol-du-u]-nu söyle-di-ler.
pro one snake-GEN side-his-LOC be-NSR-3s-ACC say-PAST-3pl
‘They said that (of the salient snakes) one (of them) was by his side’

Because overt case only appears on nominals that have raised, let us adopt the idea
that structural case in Turkish is assigned in a Spec-Head configuration.
72
From the
examples so far, we have seen that an argument with overt structural case has raised
from its base position where it has received a -role. In fact, the so-called
definiteness effect that specific arguments raise and get case and non-specific ones do
not, has been observed in other languages. However, Enç (1991) among others
demonstrates convincingly that this is actually a specificity effect
73
. Mahajan (1992)
also refers to specificity in providing an account for object movement in Hindi, noting
that non-specific nominals cannot undergo overt object shift. I will adopt this insight
and assume that it is specificity (or rather, non-specificity) that constrains subject and
object raising.





72
This assumption has several implications: 1) I will assume that all case in Turkish is overt (including
the null nominative Ø-morpheme); and 2) case assignment via Long-distance Agree is not possible in
Turkish.
73
Enç reminds us that the term had already been used by Fiengo and Higginbotham (1981) and
Hudson (1989) to describe constraints on NP movement. For a detailed explanation of the distinction
between definiteness and specificity, with a particular emphasis for Turkish, see Enç (1991) and
Kennelly (2003).

82

2 Toward an analysis of DP/NP structure and Case in
Turkish

In Chapter 1, I proposed that the difference in syntactic behavior between specific and
non-specific arguments could be attributed to a structural difference. I suggested that
specifics were DPs and non-specifics were NPs. I used the insights from Longobardi
(1994) to conclude that Turkish specifics have a null DP layer. Longobardi argues
that, in Italian, nouns as arguments, in contrast to predicative nouns, must have a DP
layer. He shows that a singular count noun must have a lexical D unless it has a mass
interpretation that allows for quantification, in which case there is a null D. Having
argued for the existence of a null determiner in Italian, Longobardi presents evidence
for N-to-D movement and posits the structures in (5) as the possible structures of
arguments with a null D. In his story, proper names must always raise to D (i),
pronouns are base-generated in D (ii), and common nouns do not normally overtly
raise to D (iii).

5) Arguments with null D in Italian

(i) DP (ii) DP (iii) DP
¸ ¸ ¸
¸ ¸ ¸
D° NP D° D° NP
¸ pronoun ¸
N° N°
proper name common noun



83

Let’s see the Italian facts presented by Longobardi.
74
Italian sentences seem to
require a D in the preverbal subject position, as shown in examples (6) and (7).

6) a. *Acqua viene giù dalle colline.
water comes down from the hills

b. Viene giù acqua dalle colline.
comes down water from the hills

7) *(Un/Il) grande amico di Maria mi ha telefonato.
(a/the) great friend of Maria called me up

Italian adjectives and possessives may occur between D and N or postnominally, but
as the examples in 8) show, neither an adjective nor a possessive may precede a
determiner with either common or proper nouns.

8) a. *mio il Gianni
my the Gianni

b. *vecchio il tavolo
old the table

We see in (9) the expected paradigm: the possible surface orders, [D – Poss – N] in
(9)a and [D – N – Poss] in (9)b. So, what is going on with (9)c and (9)d that makes
one bad and the other good?

9) a. Il mio Gianni ha finalmente telefonato.
D Poss N
the my Gianni finally called up



74
We will see these facts repeated in Chapter 5 with regard to human nominals, as they are useful in
helping us formulate the structure of Turkish DPs.

84

b. Il Gianni mio ha finalmente telefonato.
D N Poss
the Gianni my finally called up

c. *Mio Gianni ha finalmente telefonato.
Poss N
my Gianni finally called up

d. Gianni mio ha finalmente telefonato.
N Poss
Gianni my finally called up

Longobardi proposes that in sentences such as (9)d above, with a determinerless
proper name as subject, there is in fact a null D head, and that there has been N to D
movement. Thus, (9)d is not an exception to the requirement of a DP as subject. The
subjects in (9)b and (9)d are almost identical except that in (9)d, there has been head
movement of N to a null D. The structural difference in the subjects is shown in (10).

10) a. Gianni ... (in (9)d) b. Il Gianni ... (in (9)b)
DP DP
¸ ¸¸ ¸ ¸ ¸¸ ¸
¸ ¸¸ ¸ ¸ ¸¸ ¸
D° NP D° NP
Ø ¸ ¸¸ ¸ Il ¸ ¸¸ ¸
N° N°
Gianni Gianni


If we assume that a DP projection is obligatory for proper names as subjects, we cn
understand why (9)c is unacceptable: mio precedes the D°-N° complex. The Poss can
follow D, or can follow D and the name, but it can never precede D. Assuming that
Nº has raised to Dº, the Poss should follow the name. Compare this with (9)d which
is acceptable because, beginning with the base order of [D – Poss – N], the

85

subsequent N-to-D movement yields a DP with a [D–N–Poss] surface word order, as
shown in (11).

11) [
DP
D°-null [
NP
A°-mio N°-Gianni]][
DP
Gianni-D° [
NP
mio trace-G Gi ia an nn ni i]]



Longobardi cites Benincà’s (1980) observation that Italian bare nouns as arguments
must be interpreted roughly as indefinite, existentially quantified NPs, as in (12).

12) a. Bevo sempre vino.
I always drink wine.

b. Mangio patate.
I eat/am eating potatoes.

c. Non c’era studente in giro.
There wasn’t a student around

Longobardi assumes an empty D even for these common nouns, with the proviso that
N does not raise to D. Departing from Longobardi, I propose that these nominals do
not have a D projection. I suggest that the structure of ‘pretty girls’ in the existential
sentence in (13)a is not as in (13)b, but rather as in (13)c.
75




75
To be clear, I make this assumption for Turkish. It need not hold for Italian. One line of reasoning
is as follows: Italian has lexical determiners. There may be parametric variation in how a grammar
expresses specificity and definiteness which is determined by whether a language contains lexical D’s.
Unlike Italian, Turkish does not have lexical determiners. The grammar must find an alternative way
to capture specificity/non-specificity. If we assume that Dº is the locus of specificity, then the overt
expression of that feature will differ based on the availability of functional items from the lexicon.
Thus, in Italian (and generally in English, as well) it is the phonological content of the Dº that provides
the specific/definite interpretation, whereas in Turkish, it is the absence of Dº altogether that marks
non-specificity.

86

13) a. Ci sono belle ragazze.
there are pretty girls

b. *[
DP
[
NP
A°-belle N°-ragazze]]

c. [
NP
A°-belle N°-ragazze]

Taking this line of reasoning a bit further, it has been argued that only DPs can be
arguments, NPs can only be nominal predicates
76
. In fact, Mendelbaum (1994) shows
that predicate NPs are basically adjectival. Translating this idea into an “event-ish”
semantic interpretation, a sentence with an NP subject like cat in (14)a, would be an
event of ‘cat-scratching’ which would have his arm as the Theme, as shown in
(14)b
77
. This differs from the sentence in (15)a where the subject is a DP in that we
now have an external argument cat. Thus, the semantic interpretation for (15)b
would be there is an event of ‘scratching’ which has cat as the Agent and his arm as
the Theme.

14) a. [pro kol-u]-nu kedi tırmala-dı
arm-POSS-ACC cat scratch-PST
‘A cat (i.e. some cat or other) scratched his arm’

b. ∃e [Cat-scratching (e) & Theme (e, his arm)]

15) a. kedi [ pro kol-u]-nu tırmala-dı
cat arm-POSS-ACC scratch-PST
‘The cat scratched his arm’

b. ∃e [Agent (e, the cat) & Scratching (e) & Theme (e, his arm)]

76
Higginbotham (1987) proposed that an argument is “saturated” and can thus be assigned a theta role.
By extension Szabolcsi (1987), Abney (1987) and Longobardi (1994) have argued that NPs are
nominal predicates (unsaturated) and do not bear a theta-role and DPs are arguments that do bear a
theta role. Stowell (1989b) has shown that NPs are non-referential, whereas DPs are referential.
77
I reject the idea that the non-specific subject incorporates into the verb for two reasons: it can be a
large expression such as “hundreds of cats” and it serves as an intervener for A-movement from its c-
commanding domain.

87

Longobardi’s reasoning is useful for Turkish because it provides evidence for a null
D. The question is, are all arguments in Turkish indeed DPs as Longobardi proposes
for Italian, or does Turkish allow both DPs and NPs as arguments? What must we
say to account for certain arguments not moving and not bearing overt case?
I assumed that non-specifics do not have a DP projection. There are two
alternatives worth considering. One we have seen, Longobardi’s suggestion that all
arguments have a DP layer.
78
The second is an attempt to account for the
displacement and case facts in Turkish by suggesting that non-specific arguments
(either subject or object) incorporate into the verb, in the style of Baker (1988).
Kornfilt (1984, 2003
79
), and Kural (1992) argue against this approach.
80
The first
problem with the incorporation account for non-specifics in Turkish has to do with
the fact that a variety of focus-question and adverbial particles can appear between
the supposedly incorporated noun and the verb. In (16)a, the Yes/No Focus Q-
morpheme separates the noun from the verb. Compare with (16)b, where the Q-
morpheme is in its canonical unmarked position post-verbally. In (17), we see that
the particle –DA, ‘also/too’ can appear between noun and verb, and (18),
demonstrates that the Turkish free morpheme bile ‘even’ can occupy this slot.


78
Longobardi’s wording does provide some flexibility: “...a ‘nominal expression’ is an argument only
if it introduced by a category D” (Longobardi 1994: 620). By disambiguating predicative nominals
from arguments, it may be possible for an NP (or perhaps just an N°) to be a nominal predicate, rather
than an argument, as in the structure [
VP
V° N°]. See fn. 76.
79
Although Kornfilt (2003) adopts the incorporation account, her arguments against incorporation are
stronger than those that support it (“In spite of these inconclusive points, I would like to claim that,
through the interaction of scrambling and incorporation, Turkish does make an interesting contribution
with respect to incorporation”, p.144). Kornfilt admits that certain properties of incorporation are not
found in Turkish. Her motivation for adopting the incorporation account which she had argued against
in Kornfilt (1984) is to explain puzzling scrambling facts.
80
The anti-incorporation arguments and examples presented in this section are from Kornfilt (2003).

88

16) a. Hasan pasta-mı ye-di
Hasan cake-FOC eat-PST
‘Hasan ate CAKE??’

b. Hasan pasta ye-di-mi
Hasan cake eat-PST-Q
‘Did Hasan eat cake?’

17) Hasan pasta-da ye-me-di
Hasan cake-too eat-NEG-PST
‘Hasan didn’t eat cake either’

18) Hasan pasta bile ye-me-di
Hasan cake even eat-NEG-PST
‘Hasan didn’t even eat cake’

The second argument against incorporation is that we would expect to see a reflex in
thematic assignment, which we don’t. Causatives in Turkish introduce an extra theta
role to the verb, that of causee. Reminiscent of ergative systems which have a case
assigning hierarchy based on the number of arguments introduced into a structure, the
causee of an intransitive verb in Turkish is assigned accusative case whereas the
causee of a transitive verb is assigned dative case. This is shown in (19).

19) a. Hasan Ali-yi ko-tur-du
Hasan Ali-ACC run-CAUS-PST
‘Hassan made Ali run’

b. Hasan Ali-ye kutu-yu aç-tır-dı
Hasan Ali-DAT box-ACC open-CAUS-PST
‘Hasan made Ali open the box’

The logic is that when a direct object incorporates into the verb, the resulting
syntactic structure should be analogous to an intransitive, and by way of extension,

89

the causee should surface with accusative case. As we see in (20), this is not borne
out; in a transitive sentence with a non-specific direct object, the causee must bear
dative case. The theta roles of the transitive verb, and accompanying case
requirements on the causee, remain constant, regardless of the specificity of the direct
object. This is unexpected if incorporation creates a new word resulting in the direct
object losing its independent status.

20) Hasan Ali-ye / *-yi kutu aç-tır-dı
Hasan Ali-DAT/*-ACC box open-CAUS-PST
‘Hasan made Ali open boxes’

These facts and others presented by Kural (1992, 1997) are sufficient for us to
abandon the idea that non-specifics objects in Turkish incorporate into the verb. In
addition, I reject the notion that the non-specific subject incorporates into the verb for
two reasons. First, taking the sentence in (14)a as an example, the subject can be a
large quantified expression such as “hundreds of cats”. Second, the non-specific
subject serves as an intervener for A-movement from its c-commanding domain.
Returning to Longobardi’s idea, what would we have to say to account for the
Turkish facts if we adopt Longobardi’s proposal that all arguments are DPs? To
account for the obligatory raising of specifics only, we could say that D may or may
not have a Specificity feature which is strong.
81
When D is [+Specific], the DP must
raise to check that feature. When the feature is missing on D, the DP does not raise.

81
Or, we could say that the [+Specific] feature is uninterpretable, and can only be deleted in a Spec-
Head configuration with a functional head.

90

The structure of a specific DP
82
would be as in (21)a, and that of a non-specific DP
would be as in (21)b where D does not have the Specificity feature.

21) a. (specific) DP b. (non-specific) DP
¸ ¸
¸ ¸
NP D+[Specific] NP D
¸ ¸
¸ ¸
N N

How do these DPs get case? Based on Chomsky (1994, 1995), there are presumably
two ways that case features can be checked: either by Match and Agree, in which case
no movement occurs and the nominal does not receive overt case-marking, or by
Move, which results in overt case morphology. In our story so far, the +Specific DP
would be a target of Attract, followed by Move, which would result in overt case
morphology, while the non-specific DP would get its case checked via Agree.
There are several shortcomings to approach. First, it posits two different case
checking operations: case-checking via Match and Agree and case-checking via
Attract and Move. To resolve this, one could converge the operations by assuming
that one operation is an extension of the other. In this system, the case-checking head
looks for D, Matches and Agrees with it, and if the D had the additional [+Specific]
feature, as in (21)a, the case-checking head would have to Attract the DP to its
specifier to check the [+Specific] feature.
83

This account would have consequences for the definition of the EPP.
Certainly, the EPP would have to be redefined so that it does not target D’s without

82
As the focus of here is to explain Turkish facts, I will use the head-final Turkish tree structure
83
It may be Agr that is responsible for case checking and agreement. Because the presence of an Agr
projection is tangential to this schema, I avoid making a commitment.

91

the [+Specific] feature. Furthermore, for non-specific DP’s, where a case-checking
head checks case without Attract, the EPP of that head would not be satisfied.
Separating case-checking from the EPP may not necessarily be a bad outcome, but we
would in effect be discarding the requirement of a Spec-Head configuration for case
checking and supplanting it with a rule of Spec-Head checking of a Specificity
feature. Plus, we would now have a language specific definition of the EPP whereby
it could only be satisfied by a DP with a [+Specific] feature.
It being preferable to reduce assumptions rather than to create them, let’s
simplify matters and return to the Chomsky (1995) definition of the EPP as a strong
D-feature. Let’s also assume for Turkish that all phrasal movement is driven by the
EPP
84
and that all functional heads have an EPP feature which must be satisfied by a
DP merging into the specifier position of that head. Let’s take these as given for now,
with the understanding that we will re-examine these assumptions more thoroughly in
the next chapter.
My proposal for Turkish is, what you see is what you have. Thus, when there
is no overt case morphology, there has been no case assigned. For the null
nominative case as well, I assume an overt Ø morpheme, but crucially, this is the only
Ø case morpheme. I do not adopt a null accusative, absolutive, partitive, or default
case. Contrary to Longobardi, I suggest that (at least in Turkish) non-specific
nominals lack a DP projection entirely and are only NPs. Specifics are DPs that
contain a null D.
85
In sum, in Turkish nonspecifics are NPs which do not need case,

84
This assumption does not extend to scrambling which is outside the scope of this paper. I have
nothing to say as to what drives scrambling.
85
In fact, Chomsky (1995) considers Dº to be the locus of specificity.

92

and specifics are DPs which must get case.
86
Following Longobardi, I assume that
names being referential must (almost) always undergo raising from N to D
87
and that
pronouns are base-generated in D
88
.
To be clear, I am proposing narrow definitions of the EPP and the Case Filter,
as follows:
The EPP: a feature
89
of a functional head, v, T, and C, which must be satisfied by a D
feature, i.e. a DP in the specifier of that functional head, for a derivation to
converge.
The Case Filter: a convergence requirement that all DP’s in a derivation must have
their case overtly checked.
The assumption for DP arguments is that they are attracted by the EPP of a
case-assigning head, satisfying the EPP, and are assigned structural case by that head,
satisfying the Case Filter. Again, for the sake of simplicity, I do not assume the
Inverse Case Filter, that is, all heads with an EPP feature must have that feature
checked/deleted, but a case assigning head need not necessarily assign case. This

86
Actually, this suggestion may, in fact, be along the lines of Longobardi in the sense that non-specific
direct objects, and non-specific subjects of unaccusative verbs, being complements of V, may be
predicative, and thus not require case. The reason, I have not adopted this idea entirely is because of
transitive and unergative subjects which presumably merge in [Spec, vP]. These cannot be considered
predicative, whereas the NP/DP distinction in case requirement would account for the different
behavior of these subjects, particularly in embedded structures where specific subjects bear overt case.
I have adopted the NP/DP difference for case assignment and down-played the predicative nature of
verbal complements for the sake of uniformity.
87
Exceptions to this generalization are cases where the name is non-referential as in the relative clause
in (i) where the direct object, Ali, is not case-marked, and refers to non-specific men all named Ali.
Compare with (ii) where Ali is case-marked and must refer to a specific individual.
i. [Ø Ali öp-en] kız-lar
Ali kiss-SR girl-pl
‘girls that (only) kiss Ali’s’, i.e. men named Ali
ii. [Ø Ali-yi öp-en] kız-lar
Ali-ACC kiss-SR girl-pl
‘girls that kissed Ali’
88
Postal (1966) among others has suggested that pronouns are D with features.
89
The EPP may, in fact, be a feature of a feature, for example a feature of the Case feature on v and T.

93

assumption will have consequences for movement to T of non-subjects, such as
locatives or datives.
90

The generalization that a specific DP must raise, whereas a non-specific NP
cannot, seems to be too strong, cross-linguistically. A more accurate generalization
for many languages is that non-specifics cannot raise. For example, in Icelandic
Object Shift
91
constructions, only a definite (specific, for our purposes) DP may raise
out of a VP. In (22)b), where the negative ekki marks the edge of the VP, raising of
‘the-book’ is felicitous whereas raising of ‘a-book’ is not.

22) a. Jón keypti [
VP
ekki bókina/bók. ]
Jon bought not the-book/a-book

b. Jón keypti bókina/*bók [
VP
ekki t ].

Although in Germanic and Icelandic, the shifting of pronouns is obligatory, a fact I
will attribute to their specificity,
92
full NPs need not shift, but crucially, if they do
raise, they must be interpreted as definite or specific. Non-specific objects, on the
other hand, must remain within the VP.
I will assume that there is parametric variation in obligatory raising for
specifics. The correlation between overt case and raising/non-raising cannot be a
maintained in Germanic or Icelandic because all NPs/DPs bear overt case in these

90
We will re-examine this assumption also in Chapter 4.
91
Called so by Holmberg (1986). Because the data in (22) appears in so many papers on this subject, I
am unable to give credit to the source of the sentences.
92
See fn. 88, pronouns may, in fact, be D.

94

languages. What is important here is that we are seeing evidence in different
languages of a parameter that constrains the movement of non-specific NPs.
93



3 The EPP and case assignment

I have proposed that NPs are non-specific, do not need case, and remain in situ,
whereas DPs are specific, require case, and must raise at least to a projection of a
case-checking head. Let’s look at the Turkish facts to see if this analysis can be
maintained. In (2) we saw that adverbs of manner cannot appear between a non-
specific nominal expression and the verb. As demonstrated in (23)b, a verb does not
tolerate any element between itself and its non-specific complement. This contrasts
with (23)c, where one or more adverbials can appear between the accusative specific
object and the verb. Clearly, a direct object does not raise from its base-generated
position unless it is specific, and case-marked.
94


23) a. Hasan dün /kaık-la pasta yedi.
Hasan yesterday/spoon-COMM cake ate
‘Hasan ate (some) cake with a spoon’

b. Hasan pasta *dün/*kaık-la yedi.



93
It may be that the ban on movement may not be due to non-specificity but rather a function of an NP
being unvalued. In Chapter 4, we will see that whereas an NP cannot satisfy the EPP, it does however,
give rise to intervention effects when it is the closest nominal to the Attracting head. One way of
explaining this would be to say that whereas an NP contains the kinds of features that can satisfy the
EPP, it is unspecified for a value and therefore causes a crash.
94
Contrastive Focus in Turkish is the immediate preverbal position. Although (i) is a grammatical
sentence, it has a marked and contrastive interpretation, as in “Hasan ate cake yesterday (rather than
today.)”.
(i) Hasan dün bu pastayı yedi.
Hasan yesterday this cake-acc ate.
The unmarked word order in sentence (23)c, does not entail this interpretation.

95

c. Hasan bu pastay-ı dün /kaık-la yedi.
Hasan this cake-ACC yesterday/spoon-COMM ate
‘Yesterday, Hasan ate this cake with a spoon’

For Turkish, classifying specific nominals as DPs that require case and non-specific
nominals as NPs that do not, enables us to capture the fact that overtly case-marked
arguments must receive a specific interpretation. Let’s be clear, though, on what I am
committed to and the implications that follow. First, note that the correlation between
the EPP and case assignment
95
is tangential. DPs do not raise for case, they raise to
satisfy the EPP. All movement is driven by the EPP, and must obey Attract Closest
or Minimality. Second, although I am not totally committed to this, I am suggesting
here that in Turkish case is assigned in a Spec-head configuration after the DP has
been attracted by the EPP, while in other languages case may be assigned via Agree.
Similar to what I am proposing for Turkish can be found in Icelandic seem-
type raising constructions (from Jonas and Bobaljik (1993)). In (24)a the dative
experiencer can remain in the VP if it is indefinite, but must move to subject position
if it is definite or a pronoun (24)b. Evidence of a Minimality effect is in (24)c, where
the embedded subject cannot be raised over the dative experiencer.

24) a. Það virðist einhverjum manni [hertarnir vera seinir]
there seems-sg some man-DAT the-horses-NOM be slow
‘It seems to some man that the horses are slow.’

b. Mér virðast t [hertarnir vera seinir]
me-DAT seem-pl the-horses-NOM be slow
‘It seems to me that the horses are slow.’

c. *Hertarnir virðast mér [ t vera seinir]
the-horses-NOM seem-pl me-DAT be slow

95
I remain agnostic as to whether the operation is case assignment or case checking.

96

In (24)b, the embedded verb is nonfinite and cannot assign case to the embedded
subject, the horses. We must assume that it was the matrix finite T that assigned
nominative case to the embedded subject (triggering agreement with the verb), while
the dative subject was attracted to [Spec, TP] by the EPP. I include this example to
demonstrate that the EPP and case assignment are separate operations.
A difference that may exist between Turkish and Icelandic is how case is
assigned. In (23)b, we see evidence of case checking via Agree. In Turkish,
however, case may need to be checked in a Spec-head configuration after the DP has
moved to the Spec of the case-assigning head.
What is crucial, however, is that specificity is a feature of D, and lacking overt
determiners Turkish manifests specificity with both case morphology and
displacement. I am thus suggesting parametric variation in how a language encodes
specificity. In many languages, in situ subjects and objects receive overt case
whether they are specific or non-specific. Whereas these languages use overt
determiners to mark specificity, Turkish makes use of both displacement and case
morphology to mark specificity. As to whether there is something universal about
(non)specificity and NP/DP movement, I leave for further research.


4 Looking at Turkish ‘Quirky’ relatives


By adopting the DP/NP split above, we can explain the rather puzzling phenomenon
of optionality in what are otherwise well-behaved relative clauses in Turkish.

97

We are now familiar with the two types of relative clauses in Turkish: the
Subject Relative (SR) (25)a, and the Non-subject Relative (NSR) (25)b.

25) a. [ Ø mektub-u gönder-en ] kız
letter-ACC send-SR girl
‘the girl who sent the letter’

b. [ kız-ın Ø gönder-di-i ] mektup
girl-GEN send-NSR-3s letter
‘the letter that the girl sent’

We saw that these forms are fairly predictable with the caveat that when there is no
subject in the clause, as in impersonal passive constructions, the SR form must be
used. The sentences in (26) are examples we saw in previous chapters where the gap
is the (oblique) object of an impersonal passive.

26) a. [ Ø
i
Ankara otobüs-ü-ne bin-il-en ] durak
i

Ø Ankara bus-CM-DAT board-PASS-SR stop
‘the stop where the Ankara bus is boarded’

b. [ Ø
i
bu durak-tan bin-il-en] otobüs
i

Ø this stop-ABL board-PASS-SR bus
‘the bus which is boarded from this stop’ (from Kornfilt (1997))

This phenomenon does not seem too troubling at this point, because one can
immediately hypothesize that the SR morpheme is probably licensed by movement of
an element (the relativized expression) from [Spec, TP] to [Spec, CP]. In cases where
there is no surface subject, [Spec, TP] remains vacant. The EPP of T attracts a lower

98

DP, after which the DP can move to [Spec, CP] thus satisfying the licensing of the SR
verbal form
96
.
We also saw in Chapter 2, SR examples as in (27) where the external head is
not the subject of the verb. These examples differ from the ones in (26) because the
relative clauses here have overt subjects, mouse and student, within the clause. Why
is the SR form felicitous when there is no notion of subjecthood of the extracted
elements hole and statue?

27) a. [ Ø
i
fare çık-an ] delik
i

Ø mouse come.out-SR hole
‘the hole which a mouse/mice comes out of’

b. [[ Ø
i
üst-ü-ne] örenci yaslan-an ] heykel
i

Ø top-3s-DAT student lean-SR statue
‘the statue that students are leaning on/lean on’

Let’s make our generalization about the SR form more specific: The SR form is
triggered when the EPP of T has attracted the relativized element, after which, the
element further moves to [Spec, CP]. Although this is not an explanation of the SR
form, but rather for ease of remembering, we are saying the conditions that license the
SR form are met when the EPP of T has been satisfied by a +Wh-DP.
Note now that in (27), the subjects of the clauses are non-specific. From the
examples in (28)
97
, we can deduce the generalization that when the clausal subject has
a non-specific interpretation, the SR morpheme is licensed, otherwise, the NSR form

96
This is reminiscent of Stylistic Fronting in Icelandic which is licensed only in impersonal
constructions and in constructions with a subject gap (Maling 1980/1990). In both Icelandic and
Turkish, an operation that would be barred for elements other than the subject, is licensed when the
structural position of the subject has not been otherwise occupied.
97
Example (13a) is from zsoy (1994)

99

must be used. The nominal ‘three goats in (28)a is non-specific whereas in (28)b, it
receives a specific interpretation.

28) a. [üç keçi otla-yan] bahçe
three goat graze-SR garden
‘the garden where three goats graze’, as in “at all times”

b. [üç keçi-nin otla-dı-ı] bahçe
three goat-GEN graze-NSR garden
‘the garden where the three goats grazed’

Let’s look at another example with similar properties. In the RC in (29)a, the subject
bee is specific and has raised to [Spec, TP] to satisfy T’s EPP and has received case.
The relativized element cannot move to [Spec, TP]. It follows that when the
relativized expression must move to [Spec, CP] from any projection other than [Spec,
TP], the NSR form is required and the SR form is barred. In contrast, the SR form is
acceptable in (29)c precisely because the subject bee is non-specific and has not
raised from its base-generated position. This frees up [Spec, TP]; here, the EPP of T
is satisfied by the relativized element, a +Wh-expression. I include (29)b to show that
a specific subject cannot remain without case, and neither can a non-specific subject
raise above an accusative object.

29) a. [arı-nın [Ø
1
baca-ın]-ı sok-tu-u ] kız
1

bee-GEN Ø leg-POSS-ACC sting-NSR-POSS-3s girl
‘the girl whose leg the bee stung’

b. *[arı [Ø
1
baca-ın]-ı sok-an] kız
1

bee Ø leg-POSS-ACC sting-SR girl
‘the girl whose leg a/the bee stung’



100

c. [[Ø
1
baca-ın]-ı arı sok-an] kız
1

Ø leg-POSS-ACC bee sting-SR girl
‘the girl whose leg a bee/some bees stung’


5 The subject/non-subject asymmetry is a misnomer

We have seen evidence that the subject/non-subject account for Turkish relative
clauses is actually illusory. The SR form is licensed when a +Wh-DP moves from
[Spec, TP] to the CP projection. For non-subject relatives, [Spec, TP] is usually
occupied by the non-Wh-subject. This is exactly the scenario where the SR form is
barred and the NSR form must be used.
In normal discourse, the subject of the relative clause will often be “old
information” and will therefore be specific.
98
Thus, in simple relative clauses, it will
usually be the case that the subject, a DP, will be in [Spec, TP], having been attracted
there by the EPP. If the subject is the relativized expression, i.e. +Wh, then the SR
form will be licensed. If a non-subject is relativized, the NSR form must be used
because [Spec, TP] will be occupied by the non-Wh subject, and movement to [Spec,
CP] will have to originate from a position lower that [Spec, TP]. In a construction
with no subject, as in impersonal passives, or when the subject is an NP and cannot
raise to [Spec, TP], the SR form will be triggered, even though the relativized element
is a non-subject DP.



98
It is a general property of languages that subjects tend to be about the topic. This carries over into
RCs where in normal discourse, the subject often refers to the topic.

101

6 Repeat of Pesetsky and Torrego (2001)


We saw in Chapter 2, that Pesetsky & Torrego’s (2001) analysis straightforwardly
accounts for the –DIK morpheme in non-subject RCs. Their Head Movement
Generalization requires the C to T to check C’s uT feature. There is no +Wh element
in [Spec, TP] to outcompete T-to-C. Thus, what we see in derivation (30), is simply
raising of T to C as in , resulting in a PF output of –DIK (the NSR form), followed
by movement of the +Wh-object letter to [Spec, CP] to check C’s uWh feature. As I
have done throughout this dissertation, I show movement only within the clause and
ignore promotion of the relativized element beyond CP.

30) [ kız-ın Ø gönder-di-i ] mektup
girl-GEN send-NSR-3s letter
‘the letter that the girl sent’

CP
¸ ¸¸ ¸
¸ ¸¸ ¸
TP C°
¸ ¸¸ ¸
girl ¸ ¸¸ ¸
VP T°
¸ ¸¸ ¸
t-girl ¸ ¸¸ ¸
letter V°
+Wh send



This contrasts with the SR form where the subject bears a +Wh-feature. The
derivation of (31)a shown in (31)b demonstrates that the nominative subject girl can
check both uT as well as uWh features of C, making the movement of T to C
gratuitous.


102

31) a. hediye-yi ver-en kız
gift-ACC give-SR girl
‘the girl who gave the gift’

b. [ C° +uT, +uWh] [
TP
[girl, uT, uWh] T° [
VP
bought the gift] ]



The P&T story can be used to account for Turkish relatives. One need only say that
T° always assigns nominative case, and that movement of a nominative expression to
[Spec, CP] triggers the SR form. However, this story requires the stipulation that
once nominative case is checked (note that deletion is not an option here) on the
+Wh-subject, those features remain “alive” to check uT on C. Recall the impersonal
passive constructions in (26) repeated as (32) which require the SR form. I have
included sentence (33) to demonstrate that both relativized elements had inherent
case, stop+ablative and bus+dative, prior to extraction. Because the phrases in (32)
lack a “subject”, [Spec, TP] was left vacant for another element, the relativized DP.
Under the P&T story we would have to assume that nominative case is assigned to
the inherently case-marked DP, so that it can check both uT and uWh on C and bleed
T to C movement. It is not clear to me that the T in impersonal passive constructions
assigns nominative case. This would be even more obvious in embedded structures
where T° presumably assigns genitive case. Thus, preposing (i.e. raising to T) the
subject and adding genitive case (either with or without the inherent case) on either
DP results in acceptability, (33)b-c.

32) a. [ Ø
i
Ankara otobüs-ün-e bin-il-en ] durak
i

Ø Ankara bus-CM-DAT board-PASS-SR stop
‘the stop where the Ankara bus is boarded’

103

b. [ Ø
i
bu durak-tan bin-il-en] otobüs
i

Ø this stop-ABL board-PASS-SR bus
‘the bus which is boarded from this stop’

33) a. pro [onlar-ın otobüs-e bu durak-tan bin-di-i]-ni duy-dum.
pro they-GEN bus-DAT this stop-ABL board-NSR-3-ACC heard-1s
‘I heard that they boarded the bus from this stop.’

b. *pro [otobüs(-e)-nin bu durak-tan bin-il-di-i]-ni duy-dum.
pro bus(-DAT)-GEN this stop-ABL board- PASS-NSR-3-ACC heard-1s
Intended: ‘I heard that the bus is boarded from this stop.’

c. *pro [bu durak(-tan)-ın otobüs-e bin-il-di-i]-ni duy-dum.
pro this stop(-ABL)-GEN bus-DAT board- PASS-NSR-3-ACC heard-1s
Intended: ‘I heard that [from this stop] is boarded the bus.’

This is why I proposed in Chapter 2, that rather than C having an uT feature, it is T
that has an uninterpretable Wh feature, which I called uC. The idea was that in order
to ensure that the Wh-features of C get checked, C selects for a T with an uC feature.
I did not go into detail as to what this feature on T might be. Perhaps it is some kind
of Topic-like referential requirement. It has been noted that except for the aorist
tense, all other tenses require that at least one nominal be referential (using this term
rather loosely) or “specific”, i.e. a DP. Thus, even sentences such as “Some dogs
barked” requires the subject “some dogs” to have a partitive construction that displays
the syntactic properties of a DP. A Turkish sentence with only non-specific
arguments is not acceptable.
99
It may be, then, that to guarantee convergence, every
non-existential, non-irrealis sentence has a Topic projection that selects a T with an

99
Isever (2003) argues against this idea presenting a host of sentences where both arguments are non-
specific. He fails to note two things, however. First, many of his examples include the +human
nominal biri ‘someone’ which I show in Chapter 5 is a partitive DP. Second, each of his examples
includes the evidential verbal morpheme –mi either as the sole TAM (Tense, Aspect, Mood) marker
or in addition to a future or aorist morpheme. This is evidence that only in restricted existential or
“irrealis” TAM environments can a sentence converge with no DPs.

104

EPP feature, as well as an uC feature. In matrix sentences, T° is prohibited from
moving to C°; once the EPP of T is checked by the closest DP (presumably the
subject), the +Wh-Topic-expression (whichever that is) A-bar moves to [Spec, CP].
In RCs, where T-to-C is not prohibitted, if the DP that the EPP of T is +Wh
then, uC on T is deleted, and the +Wh-DP moves to [Spec, CP] to check uWh on C.
If the closest DP to T is not +Wh, then head movement of T to C is required to check
T’s uC feature. A +Wh-DP must still move to C to check C’s uWh feature. We saw
how this played out for the impersonal passive constructions in Chapter 2. An
example is repeated in (34).

34) [ Ø
i
bu durak-tan bin-il-en] otobüs
i

Ø this stop-ABL board-PASS-SR bus
‘the bus which is boarded from this stop’

CP
¸ ¸¸ ¸
+Wh-bus ¸ ¸ ¸ ¸
TP C°
¸ ¸¸ ¸
t-(+Wh-bus) ¸ ¸¸ ¸
VP T°
¸ ¸¸ ¸
t-(+Wh-bus) VP
¸ ¸¸ ¸
DP V°
this stop-ABL board


In the derivation in (34), the EPP of T targets the DP bus which moves to [Spec, TP]
in . This DP is +Wh; consequently, uC on T is deleted and T to C movement is no
longer motivated. The relativized element then moves to [Spec, CP] in to check
C’s uWh.


105

7 Explaining the choice in RC forms

Let’s review. We have two principles that determine the choice of RC form: 1)
Specific nominals in Turkish have a DP projection; and 2) The EPP attracts a D°
feature. In the relative clauses in (28) repeated as (35), it seemed as if both the SR
and the NSR forms were acceptable. Notice, however, that the subject in (35)a can
only receive a non-specific interpretation, whereas the subject in (35)b is specific.

35) a. [üç keçi otla-yan] bahçe
three goat graze-SR garden
‘the garden where three goats graze’

b. [üç keçi-nin otla-dı-ı] bahçe
three goat-GEN graze-NSR garden
‘the garden where the three goats grazed’

Recall that in Section 2, we had argued that non-specifics are NPs that cannot satisfy
the EPP. In a sentence where the subject is an NP, the EPP of T must be satisfied by
a non-subject-DP. This is precisely what occurs in the RC in (35)a. Let’s look at the
derivation in (36). I should point out that I am assuming that the subject three goats
and the locative garden are in the same minimal domain and thus equidistant from
T.
100
The subject is an NP, and cannot be attracted by the EPP. T’s EPP targets the
only other nominal, the DP-garden. Because garden has +Wh features, movement of
garden to [Spec, TP] in deletes T’s uC feature. T to C movement is not motivated.
Garden then moves to [Spec, CP] and deletes C’s uWh feature.


100
If I did not assume equidistance of the two nominals, we would expect intervention effects from the
higher nominal. See fn. 93.

106

36) [üç keçi otla-yan] bahçe
three goat graze-SR garden
‘the garden where three (non-specific) goats graze’

CP
¸ ¸¸ ¸
+Wh-garden ¸ ¸ ¸ ¸
TP C° [+uWh]
¸ ¸¸ ¸
t-(+Wh-garden) ¸ ¸¸ ¸
VP T° [+uC]
¸ ¸¸ ¸
t-(+Wh-garden) VP
¸ ¸¸ ¸
NP V°
three goats graze


Now let’s turn to the RC in (35)b with the derivation in (37). In this derivation, either
garden or three goats is a target for Attract by T’s EPP; however, if garden moves,
the DP-subject will be left without case and the derivation will crash. We have
already seen that structural case in Turkish is assigned in a Spec-head configuration.
Thus, garden must raise to [Spec, TP] satisfying T’s EPP feature, and receiving
genitive case from T
101
. Because the subject is non-Wh, uC on T has not been
checked and T to C movement is required, as in . The +Wh-garden then moves to
[Spec, CP] to check C’s uWh feature, .

37) [üç keçi-nin otla-dı-ı] bahçe
three goat-GEN graze-NSR garden
‘the garden where the three goats grazed’



101
It is beyond the scope of this paper to discuss how genitive rather than nominal case is assigned in
this position. I will mention though that I believe it to be a reflex of the T-C amalgam formed by head
movement of T to C, along the lines of Hiraiwa (2001) who suggests the same for Japanese ‘NO’ on
subjects in relative clauses.

107


CP
¸ ¸¸ ¸
+Wh-garden ¸ ¸ ¸ ¸
TP C° [+uWh]
¸ ¸¸ ¸
three goats-GEN ¸ ¸¸ ¸
VP T° [+uC]
¸ ¸¸ ¸
t-(+Wh-garden) VP
¸ ¸¸ ¸
DP V°
t-(three goats) graze




In the derivation in (37) we are assuming that the NSR morpheme –DIK, is the
instantiation of T in C. One other point worth mentioning is that, unlike the SR form,
the NSR verb shows agreement, not a bad consequence when we are assuming that in
this sentence case was assigned to the DP-subject. The outcome is that structural
case-assignment and agreement with the verb seem to go hand-in-hand.
Let’s now look at another pair of relative clauses which seem to allow either
RC form. The pairs in (38) have a sentential subject. To account for the acceptability
of both forms, I assume that one has an NP subject which prevents it from raising to
[Spec, TP], leaving that position open for the +Wh expression, while the in the other,
the subject is a DP which raises to [Spec, TP]. This is better shown in (39).

38) a. [[Ø
1
biz-e güven-ece-i] üpheli ol-an] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS-3s doubtful be-SR man
‘the man that it is doubtful will trust us’

b. [[Ø
1
biz-e güven-ece-i]-nin üpheli ol-du-u] adam
1

Ø 1p-DAT trust-FUT-COMP-POSS-3s-GEN doubtful be-NSR-POSS-3s man
‘the man that it is doubtful will trust us’

108

First, recall that in Chapter 2, I had posited a null resumptive pronoun (RP) inside the
sentential subject co-indexed with the +Wh-expression in the specifier position of
the nominal phrase, either NP or DP. Note that in example (38)b where we assumed
the subject was a DP, the entire sentential subject is marked with genitive case. Thus,
the phrase cannot be an element in-situ and must have raised to [Spec, TP]. The
phrase being non-Wh, the NSR form is required, that is, T-to-C movement is
necessary to delete T’s uC feature.

39) a. [
CP
[
TP
Ø
1
T° [
VP
[
NP/CP
Ø-RP
1
biz-e güven-ece-i] üpheli ol-an] adam
1

us-DAT will.trust doubtful be-SR man

b. [
CP
Ø
1
[
TP
[
DP/CP
Ø-RP
1
biz-e güven-ece-i]
2
-nin [
VP
t
2
üpheli ol-du-u] adam
1

us-DAT will.trust-GEN doubtful be-NSR-3s man


It is evident that there is really no “optionality” in the choice of RC. DP subjects
must move to [Spec, TP] or they will violate the Case Filter. The movement is
motivated by the EPP on T. If the DP subject does not have +Wh-features, T to C
movement will be required to save the derivation. This is instantiated as the NSR
form. On the other hand, NP subjects cannot satisfy the EPP and so T’s EPP feature
will attract another DP in the structure, the +Wh DP that is to be relativized. This
move by the relativized expression to [Spec, TP] checks uC on T, thereby bleeding T
to C movement.






109

8 Summary


Determining that specific arguments in Turkish must raise to be assigned case, I have
shown that it is the interaction of specificity and movement that give us the
alternations in Turkish relatives. The generalization is not one of subject/non-subject
asymmetry but rather a combination of the category of the subject, feature checking
and case assignment. An innovation in this chapter and in the previous one, is the
proposal that there is a +Wh-like feature on T. Further cross-linguistic research is
needed to support this claim. The idea that specifics are DPs and nonspecifics are
NPs, I think can be maintained for Turkish, although it seems clear that there will be
parametic variation as to how languages encode specificity. The assumption that the
EPP can only be satisfied by D is fairly well accepted; however, the proposal that NPs
are not subject to the Case Filter also needs further investigation. It may be that in
Turkish -features are uninterpretable, that is, -features need checking via
(structural) case. Ns may be lexical atoms like verbs, and prepositions, adjectives,
adverbs, etc. that do not need to have their features checked.
This chapter was intended to review the fundamental claims regarding
specificity and the categorial structure of nominals that are important to the overall
research proposal of this work. Many issues regarding the nature of DPs still need
clarification, and we will continue to address these in the proceeding chapters. In
Chapter 4, we look at how movement to [Spec, TP] is affected by the hierarchical
positions of nominals. In Chapter 5, we again address the structure of nominals, and
the syntactic reflex of a D° head.

110

Chapter 4: The EPP on T and Minimality


1 Introduction and Background



In this chapter, I present evidence for the EPP in Turkish. I then use the SR relative
clause as a diagnostic for movement to [Spec, TP]. I show that movement to [Spec,
TP] obeys Minimality and that the behavior of DP movement across NP subjects is
constrained along verb classes (as suggested by Perlmutter (1978), Burzio (1986),
among others), the structure of some of which induce Minimality effects.

1.1 Specificity, case, and displacement
102


Let’s review the behavior of arguments in Turkish. Assuming that adjuncts adjoin to
a verbal projection (either vP or VP), (1) is evidence that specific, case-marked DOs
must raise out of their thematic positions in VP.
103




102
As we saw in Chapter 3, Enç (1991) disambiguates definiteness and specificity noting that whereas
all definites must be specific, indefinites can be either specific or non-specific. The syntactic behavior
of Turkish nominals clearly falls along the specific/non-specific divide, as shown in (i). Two of the
women is indefinite but is unacceptable in an environment restricted to non-specifics.
(i) a. Bayan-lar-dan iki-si-ni tanıyor-um.
women-pl-ABL two-agr-ACC know-1s
‘I know two of the women’
b. * Bayan-lar-dan iki-si tanıyor-um.
women-pl-ABL two-agr know-1s
(ii) a. There are two women in the room.
b. *There are two of the women in the room.
103
The items marked as bad in (1) are in fact acceptable with marked stress and other contrastive or
focus intonations. The point, however, is that they are unacceptable as unmarked cases.

111

1) a. Ali dün /kaıkla /hızlı pasta(*-yı) yedi.
Ali yesterday /with a spoon/quickly cake(-ACC) ate
‘Ali ate (some) cake yesterday/quickly/with a spoon’

b. Ali bu pasta-yı dün /kaıkla /hızlı yedi.
Ali this cake-ACC yesterday /with a spoon/quickly ate
‘Ali ate this cake quickly/yesterday/with a spoon’

c. Ali pasta *dün /*kaıkla /*hızlı yedi.
Ali cake yesterday/with a spoon/quickly ate

In Chapter 2, we also saw that direct objects that raise must bear overt case, and those
that do not raise must be bare. Thus, we can use the presence or absence of overt case
as an indication of whether or not the object has raised from its base position. The
same correlation can be demonstrated for subjects, perhaps not in matrix clauses
because nominative case is the null morpheme, but certainly in embedded
constructions where the specific subject bears overt genitive case. The subject of so-
called “factive” clauses in Turkish has genitive case, as in (2). However, the non-
specific subject of the embedded existential construction in (3)a does not have case
morphology. Compare with (3)b in which the embedded subject has raised above the
locative, is case-marked, and must receive a partitive (i.e. specific) interpretation.

2) a. Ali-nin hasta ol-du-u-nu söyle-di-ler.
Ali-GEN sick be-NSR-3s-ACC say-PAST-3pl
‘They said that Ali was sick’

3) a. Yan-ın-da bir kız ol-du-u-nu söyle-di-ler.
side-his-LOC one girl be-NSR-3s-ACC say-PAST-3pl
‘They said that there was a girl by his side’

b. Bir kız-ın yan-ın-da ol-du-u-nu söyle-di-ler.
one girl-GEN side-his-LOC be-NSR-3s-ACC say-PAST-3pl
‘They said that (of the salient girls) one (of them) was by his side’


112

The conclusion is: case-marked arguments must have raised, and non-case-marked
ones cannot have raised. Let’s now look at arguments that raising is a consequence of
an EPP feature on v and T, although the focus throughout this chapter will be on T.
1.1.1 The EPP on T in Turkish
There are several versions of the EPP, but for now, I would like to adopt the version
that defines the EPP as a feature on a functional head that attracts a morphologically
contentful element to its specifier. More specifically, the EPP attracts a D feature,
and thus can only be satisfied by movement of a DP. EPP driven movement is
sensitive to minimality such that the EPP feature attracts the closest element from its
c-commanding domain that can satisfy it. I assume that all movement
104
of nominals
is to satisfy the EPP feature of the head into whose projection they move. We have
seen evidence that structural Case in Turkish must be assigned in a Spec-Head

104
My use of the term “movement” precludes scrambling. Importantly, the Turkish facts indicate that
scrambled elements do not satisfy the EPP. The SR form is unacceptable in (i), but if we embed sofa in
a locative PP/DP as in (iii), we can scramble the locative above the subject and then extract sofa
though TP licensing the SR form (ii). As shown in (iii), the PP/DP ‘on the sofa’s top’ has a
possessor/possessee structure, where sofa is the specifier or possessor (and bears genitive case). After
scrambling of the locative PP/DP (iva), the possessor first moves to [Spec, TP] and then to [Spec, CP],
triggering the SR form (ivb). The scrambled position of the PP/DP must be different from the EPP
position because if the entire PP/DP had moved to [Spec, TP] (rather than just the +Wh-expression
sofa-GEN from the specifier of the PP/DP), the SR form would not be licensed.
(i) *[bayan Ø
i
otur-an] divan
i

woman Ø sit-SR sofa
Intended: ‘the sofa that a woman is sitting on’
(ii) [Ø
i
üst-ün-de bayan otur-an] divan
i

top-AGR-LOC woman sit-SR sofa
‘the sofa that a woman is sitting on (top of)’
(iii) [
PP/DP
[
DP
divan-ın] üst-ün]-de
sofa-GEN top-AGR-LOC
‘on the sofa’ (Literally: ‘on the sofa’s top’)
(iv) a. [
TP
Spec-empty [
PP/DP
[
DP
divan-ın] üst-ün-de]
2
[
vP
bayan t
2
otur] T° ]
sofa-GEN top-AGR-LOC woman sit
b. [
CP
[
DP
divan-ın]
1
[
TP
[
DP
divan-ın]
1
[
PP/DP
[
DP
t ]
1
üst-ün-de]
2
bayan t
2
otur-an] C°]
sofa-GEN sofa-GEN top-AGR-LOC woman sit-SR
c. [
CP
t
1
[
TP
t
1
[
PP/DP
[
DP
t ]
1
üst-ün-de]
2
bayan t
2
otur-an] ] divan
i

top-AGR-LOC woman sit-SR sofa
Lavine and Freidin (2002) also show that for Russian and Ukrainian A-movement and scrambling have
different properties and consequences.

113

configuration. Within the generative framework, three potential factors have been
identified as motivating this displacement: Case (Boškovi 2002, 2005), agreement,
and the requirement that certain functional heads must have a syntactically realized
specifier phrases (Chomsky 1981, 1982, 1999). I have presented evidence that in
Turkish non-specifics are NPs, and specifics are DPs.
105
We have seen that DPs must
raise and be case-marked; “movement is driven by case” proponents could explain
this by saying that DPs must satisfy the Case Filter and so Case drives movement of
argument DPs to structural case-assigning heads. We will see shortly that this
argument is probably not correct for Turkish.
For the moment, let’s disentangle Case from displacement, and rely on a
notion of an EPP feature as “the thing that ‘causes’ movement”. Thus, for now let’s
adopt the view that the case-assigning functional heads T and v (as well as the
functional head C°) have an EPP feature that must be satisfied by a DP moving to the
specifier of that head, with case assignment occurring when the proper configuration
is achieved, as a free rider.
106

To demonstrate, in (4)a, the subject dog, to the right of the locative street, has
not raised and receives a non-specific interpretation, i.e. is an NP which cannot satisfy
T’s EPP. Assuming locatives mark the edge of VP
107
, the locative street must have
moved to a position higher than the base position of the subject, having been attracted
to [Spec, TP] by the EPP. Compare (4)a with (4)c where the subject must receive a
specific interpretation. In sentence (4c), dog is a DP which can be attracted by T’s

105
In fact, Chomsky (1995) considers Dº to be the locus of specificity.
106
Chomsky 1995, p. 282.
107
See Kural (1991).

114

EPP, hence its position to the left of the locative. Note that (4)b is unacceptable with
a non-specific interpretation for dog; an NP subject cannot raise over a locative.

4) a. Sokak-ta köpek havl-ıyor.
street-LOC dog bark-PRES.CONT.
‘A dog/dogs are barking in the street’

b. *Bir köpek sokak-ta havl-ıyor.
a dog street-LOC bark-PRES.CONT.
‘Some dog (or other) is barking in the street’

c. Köpek sokak-ta havl-ıyor.
dog street-LOC bark-PRES.CONT.
‘The dog is barking in the street’
*A/some dog is barking in the street’

The unacceptable or marginal examples
108
in (5) provide further support that NPs do
not check the EPP of T. Note that these sentences become acceptable with the
addition of a locative or temporal phrase, as in (6).

5) a. *[Bir tavuk] pi-iyor
a chicken cook-PRES
‘A chicken is cooking’

b. ??[Bir bardak] kır-ıl-mı
a glass break-PASS-PST
‘A glass is broken’

c. ?*[Bir adam] uyu-mu
a man sleep-PST
‘A man slept’

6) a. Burada [bir tavuk] piiyor.
here a chicken is-cooking


108
Examples from Kural (1992).

115

b. Burada [bir bardak] kırılmı.
here a glass was-broken

c. Burada [bir adam] uyumu.
here a man slept


I conclude from these sentences that T has an EPP feature which must be satisfied for
convergence. The adverbial bura-da ‘here’ in Turkish is actually a determiner-like
nominal bu ‘this’ with locative case -da. The nominal can take other cases as well,
for example bura-dan with ablative case means ‘from here’, and bura-ya with dative
case means ‘to here’. Thus, bura-da is a DP with locative case that is deleting T’s
EPP feature.
Of course, another interpretation is possible. If we assume the Inverse Case
Filter, we can reason that the sentences in (5) are unacceptable because T has not
discharged its Case feature, whereas the examples in (6) are felicitous because T’s
Case feature has been discharged. More specifically, in (6) T° is assigning structural
case to an inherently case-marked DP. It is difficult to know if T must assign case.
In Section 2, we will see that the only DP that cannot move to [Spec, TP] is one with
accusative case. Is this because structural case-marked elements are frozen for
further A-movement
109
or is it because case-stacking of structural case on structural

109
According to Chomsky (1995), whereas multiple satisfaction of the EPP and multiple agreement is
possible, multiple case checking is excluded under Last Resort. Case checking of a DP erases its case
features, so that if the DP moves to another case position (satisfying the EPP), it “offers no Case
feature to be checked, so the derivation crashes” (p. 284). In this account, a DP’s [– Interpretable]
feature, such as Case, is “frozen in place” once checked (p.280). This is because “checking [of
features] is deletion and is followed by erasure without exception” (p.281). Notice that this account
requires that T° (and v°) have case features that must be discharged (the Inverse Case Filter); otherwise
there would be no reason for the derivation to crash if the [-Interpretable] Case feature of the DP that
satisfied the EPP on these functional heads were erased. Furthermore, Chomsky vaguely suggests that
some version of the uniformity condition (Chomsky 1986: pp.193-194) bars A-movement from an
inherent Case position (fn. 55). Clearly, the Turkish facts show this conjecture to be incorrect.

116

case is disallowed, perhaps for PF reasons?
110
Let’s look at the implications of each
assumption.
In adopting an EPP story, I forego the need for the Inverse Case Filter. I
assume that T assigns Case to the DP in its Spec but that movement of the DP is
triggered by the EPP. In this story, if T’s EPP were satisfied, and T did not check the
case of the element in its Spec, the derivation would still converge. The problem is
we now have no way of preventing an accusative DP from moving to [Spec, TP]. We
would need to invoke [±Interpretable] features such that once a DP is structurally
case-marked, its D feature is no longer available to the EPP, i.e. it is “frozen”. This
has the unhappy result of conflating the EPP and Case. Furthermore, it is not clear
how structural case marking would make D features invisible for Attract by the EPP.
In a story where T does not need to assign case, there is no way to prevent an
accusative-DP from moving to T.
On the other hand, adopting the Inverse Case Filter will mean that T will
always have a case feature it must discharge. This gives us two ways to prevent the
accusative object from further A-movement: an accusative expression will be frozen
because of Last Resort, or T will have to assign case to whatever element is in its
Spec with the consequence that structural case stacking leads to a crash.
111
The facts
in (5) and (6) are a problem for the pure Case drives movement account. The locative
that moves to [Spec, TP] already has case, so it does not need to move for case.

Chomsky (2000) also maintains that “once Case of is checked, is “frozen”; it cannot enter into
further agreement relations” (fn. 36), although Icelandic facts regarding Quirky Dative raising to check
T’s EPP feature are mentioned. But, Chomsky makes clear that the Quirky Dative in [Spec, TP] “does
not satisfy the requirement that the features of T can be checked only by Nominative”. Either default
inflection or long distance Agree with a lower Nominative is required for convergence (p. 11).
110
Chomsky (1995) also makes reference to “Case conflict” (p. 285). See fn. 114 as well.
111
In Chapter 7, we see that the latter of these options seems to be the better account.

117

Certainly, if one rejects the EPP, the fact that a locative must move if and only if the
subject is non-specific is theoretically unstatable without recourse to the Inverse Case
Filter.
An argument against the Inverse Case Filter is found in Lavine and Freidin
(2002) (L&F). They argue that Russian and Ukrainian Accusative Unaccusative
Constructions have a “defective” T. A defective category is defined as in (7).

7) A category that lacks a full set of -features is defective.


Examples of defective categories in English are infinitival T and passive v which are
both -incomplete appear in (8)a and (8)b respectively.

8) a. We expect Len [
TP
T-DEF to finish his book this summer].

b. He was [
vP
v-DEF attacked by the visitor].


The idea is that a T that is -incomplete cannot enter into an Agree relation with a
nominal and also cannot assign nominative case. In the Russian and Ukrainian
examples below, the T° cannot assign nominative case, and cannot show
agreement.
112
The Russian constructions usually contain two nominal expressions, an
accusative direct object and an instrumental. One of these nominals must always

112
The authors (L & F) point out that they follow Chomsky (2000, 2001), Pesetsky and Torrego
(2001), George and Kornfilt (1981), among others, in assuming that nominative case in Slavic (as well
as in English) is licensed by agreement. They admit that this correlation does not hold universally
across all languages, citing Modern Greek and Japanese as examples of languages where T’s -
features and nominative case can be checked independently.

118

appear preverbally, as in (9). In the Ukrainian passive-like construction, the
accusative nominal must always move to preverbal position, as in (10).

9) a. Soldata ranilo pulej
soldier-ACC wounded-[–AGR] bullet-INST
‘A soldier was wounded’

b. Podvaly zatopilo livnem
basements-ACC flooded-[–AGR] downpour-INST
‘Basements were flooded by the downpour’

c. Volnoj oprokinulo lodku
wave-INST overturned-[–AGR] boat-ACC
‘A wave overturned a boat.’

10) Inozemcja bulo posadženo do v’jaznyci
foreigner-ACC was-[–AGR] placed-[–AGR] to prison
‘A foreigner was put into prison’


Let’s focus on the Russian case. Of interest to us in these constructions is the
following: First, the word order is discourse neutral, no matter which argument
fronts. Second, when T is -complete, the subject surfaces with nominative case.
Compare (9)a-b with (11)a-b.

11) a. Pulja ranila soldata.
bullet-NOM.F.SG. wounded-F.SG. soldier-ACC
‘A bullet wounded a soldier’

b. Liven' zatopil podvaly
downpour-NOM.M.SG. flooded-M.SG. basements-ACC
‘A downpour flooded basements’


119

Third, L&F take the fact that either the accusative expression or the instrumental can
front as evidence of accusative case being valued via ling-distance agree, and
conclude that Russian and Ukrainian, unlike Scandinavian, are not object-shift
languages. Here, we have evidence that a T that is not assigning case, nevertheless
attracts a nominal to its specifier. This is evidence against the Inverse Case Filter
because in the Accusative Unaccusative Construction expressions which already have
case move to a functional projection that does not seem to be assigning case.
Further evidence that Case and the EPP operate independently come from
Quirky Case subjects in Icelandic and Germanic case facts.

1.2 Evidence for the EPP in Icelandic and German
It looks like in Icelandic, as in Turkish, T° has an EPP feature which can be satisfied
by movement of a non-case-marked nominal to [Spec, TP] where it is assigned
nominative case as in (12) or by movement of an inherently case-marked nominal
expression to [Spec, TP] as in (13).
113
I assume that the Icelandic verb in (13),
batnaði ‘recover’ is unaccusative-like with the difference being that the expression
bearing the Benefactive/Experiencer theta role first-merges as the complement to V°
with inherent dative case. As this argument already bears case, it does not need to
raise to [Spec, TP] for case. Furthermore, nominative case in Icelandic has a distinct
morpheme. The dative expression in [Spec, TP] bears dative, not nominative, case. I
assume that T° has not assigned case in (13).
114


113
The Icelandic examples in this section are from Freidin and Sprouse (1991).
114
This is contra Burzio (2000) and Frank (2002). Burzio argues that (abstract) nominative case is
stacked upon Quirky subjects (datives, for example), and that because the nominative is assigned to the
‘outer’ shell of the DP, the agreement between subject and verb is with the outer shell only, not with
the lexical -features of the subject. In order to account for the agreement between the verb and the

120

12) Haraldur
i
las bókina sína
i

Herald-NOM read book his(+REFL)

13) Haraldi
i
batnaði veikin hjá bróður sínum
i

Harald-DAT recovered-from the-disease at-the-home-of brother his(+REFL)

Dative expressions behave in an identical fashion to the nominative subjects in raising
verbs as shown in (14). Note that whereas in (14)a the subject has been assigned
nominative case, in (14)b, the subject bears the inherent dative case which was
assigned in the embedded clause. Again, it seems clear that matrix T° has not
assigned case. These examples also demonstrate that, as in Turkish, an inherently
case-marked expression can move to [Spec, TP].

14) a. Haraldur
i
virðist [ t
i
hafa lesið bókina]
Herald-NOM seems to-have read book

b. Haraldi
i
virðist [ t
i
hafa batnað veikin]
Harald-DAT seems to-have recovered-from the-disease


object as shown in (i) from Sigursson (1996) and Taraldsen (1995).Jonas and Bobaljik (1993), Frank
adds the notion that checking of ç-features does not necessarily entail their deletion. Thus, in (i), T
first checks its EPP and ç-features against the dative subject, deleting only its EPP feature, and then
again checks, and now deletes, its ç-features against the object. This second checking operation is
optional, as evidenced by example (ii), with default verbal agreement. Although Frank provides a
lengthy account to explain the restriction against agreement between between a matrix verb across a
dative to the object of an infinitival, he fails to account for simpler examples such as (iii) from Jonas
and Bobaljik (1993), where the matrix verb agrees with the embedded subject. For a complete version
of his account, see Frank (2002: 138-153).
(i) Mér mistókust allar tilarunirnar.
me-DAT failed-3pl all the-attempts-NOM
‘I failed all the attempts.’
(ii) Mér virtust / virtist þær vinna vel.
me-DAT seemed-3pl/-3s they-NOM to-work well.
‘It seemed to me that they work well.’
(iii) Mér virðast t [ hertarnir vera seinir]
me-DAT seem-pl the-horses-NOM be slow
‘It seems to me that the horses are slow.’


121

Expressions merged with inherent case retain their case-marking even in passive
constructions. If passivization is an operation that bleeds accusative case-assignment,
there is no reason to assume that it would have any effect on inherently case-marked
DPs. Let’s assume that passivization is simply the absence of a vP projection in a
derivation with the result that the theta position for an external argument is
unavailable. In this structure, the EPP of T must still be satisfied. The sentences in
(15)a and (15)b select for a bare (non-case-marked) argument and a dative argument,
respectively. In (16)a, the direct object Harald (which needs case to avoid a Case
Filter violation) raises from its theta position to [Spec, TP] and is assigned nominative
case
115
, whereas in (16)b, the dative argument retains its case. Unlike in (15)a, there
is no evidence that nominative case has been assigned in (15)b; therefore, some other
feature must be implicated for the movement.

15) a. Egill drap Harald í gær
Egill-NOM killed Harald-ACC yesterday

b. Egill hjálpaði barninu
Egill-NOM helped the-child-DAT

16) a. Haraldur var drepinn í gær
Harald-NOM was killed-PPP yesterday

b. Barninu /*barnið var hjálpað
the-child-DAT/the-child-NOM was helped

We saw in Turkish that non-specifics cannot satisfy the EPP. The same phenomenon
seems to hold in Icelandic passive constructions. In (17), where the internal argument

115
In fact, there is evidence of long distance case assignment in Icelandic. Thus, it may be that T°
assigns nominative case in Icelandic via Agree prior to movement.

122

is non-specific, some other expression must satisfy the EPP. In (17)a, the time
adverbial ‘yesterday’ raises to [Spec, TP], and as shown in (17)b-c, where no other
expression exists in the clause, a pleonastic (or expletive) ‘it’ must be inserted to
satisfy T’s EPP feature. Note that neither of these strategies is acceptable when the
internal argument is specific, as shown in (18). I assume this is because the strategies
in (17) are both Last Resort, in order to save a derivation. When the derivation
contains a “proper” candidate that will satisfy the EPP (i.e. a specific argument), it is
the argument that must raise.

17) a. Í gær var hjálpað barni
yesterday was helped-PPP a-child-DAT

b. Það var hjálpað barni
it was helped-PPP a-child-DAT

c. *var hjálpað barni
was helped-PPP a-child-DAT

18) a. *Í gær var hjálpað barninu
yesterday was helped-PPP the-child-DAT

b. *Það var hjálpað barninu
it was helped-PPP the-child-DAT

Similar evidence appears in the German examples below.
116
Whereas T assigns
nominative case to canonical subjects, it does not seem to discharge/assign case to
inherently case-marked DPs that move to subject position. The German verbs
‘observe’ and ‘help’ subcategorize for an accusative Theme and a Dative Theme,
respectively, as shown in (19).

116
The German examples are from Freidin and Sprouse (1991), who following convention give the
examples in the form of subordinate clauses to abstract away from German matrix V2 effects.

123

19) a. daß der Polizist den Spion beobachtete
that [the policeman]-NOM [the spy]-ACC observed
‘that the policeman observed the spy’

b. daß der Polizist dem Spion half
that [the policeman]-NOM [the spy]-DAT helped
‘that the policeman helped the spy’


In German passive constructions, the formerly accusative object behaves as expected,
ending up with nominative case in subject position, as in (20)a. Contrast this with the
dative Theme in (20)b which retains its lexical (inherent) case marking, appearing in
subject position [Spec, TP] with dative case.
117
As shown in (20)b, nominative case
is not acceptable on a passivized dative expression.

20) a. daß der Spion beobachtet wurde
that [the spy]-NOM observed-PPP was

b. daß dem Spion /*der Spion geholfen wurde
that [the spy]-DAT/[the spy]-NOM helped-PPP was

One might argue that the subject in (20)b bears null nominative on top of the dative.
This is unlikely for the following reason. Verbs that have subjects with nominative
case can be coordinated in German, as in (21)a. However, as demonstrated in (21)b,
a passivized dative-DP subject is not possible in a coordinate structure with a
nominative-DP. I conclude from this that there is a case mismatch in the two DPs
which should not be the case if the dative bore (null) nominative case.

117
Franks and Levine (2005) point out that in Genitive under negation constructions in Lithuanian,
“[i]n the competition between structural and Lexical Case, lexical Case normally wins.” The same
facts hold for Russian. In Genitive under Negation constructions and Numerically Quantified NPs,
expressions that would normally bear structural case surface with genitive case, whereas inherently
case-marked expressions retain their lexical case.



124


21) a. daß der Spion Angst hatte und beobachtet wurde
that [the spy]-NOM fear had and observed-PPP was
‘that the spy was afraid and was observed’

b. *daß dem Spion Angst hatte und geholfen wurde
that [the spy]-NOM fear had and helped-PPP was
Intended: ‘that the spy was afraid and was helped’


The combination of these arguments suggests that the EPP is separate from Case.
Committing ourselves to the EPP, let us now turn to determining what elements
satisfy the EPP on T and under what circumstances. The evidence will show that the
Turkish subject relative (SR) clause form provides a reliable diagnostic for movement
to [Spec, TP].

2 Review of Turkish Relative Clauses


We saw that the subject relative (SR) requires the -An verbal morpheme with no
agreement as in (22), whereas a non-subject relative (NSR) as in (23) requires the
-DIK verbal morpheme with possessive agreement. Crucially, the NSR form requires
the (overt) clausal subject to bear genitive case morphology. Thus, in (23)a, the
subject must be case-marked and must receive a specific interpretation; the NSR
forms in (23)c and (23)d where the subject bears no overt case are unacceptable.

22) a. [Ø
i
divan-da otur-an ] adam
i

Ø sofa-DAT sit-SR man
‘the man who is sitting on the sofa’

b. *[ Ø
i
divan-da otur-dug-u] adam
i

Ø sofa-DAT sit-NSR-3s man

125

23) a. [adam-ın Ø
i
otur-dug-u ] divan
i

man-GEN Ø sit-NSR-3s sofa
‘the sofa that the man is/was sitting on’

b. *[adam Ø
i
otur-an] divan
i

man Ø sit-SR sofa

c. *[adam Ø
i
otur-dug-u] divan
i

man Ø sit-NSR-3s sofa
Intended: ‘the sofa a (non-specific) man is sitting on’

d. *[Ø
i
adam otur-dug-u] divan
i

Ø man sit-NSR-3s sofa
Intended: ‘the sofa a (non-specific) man is sitting on’

We also saw, as in (24)a, that when the subject is non-specific, i.e. is an NP under our
assumptions, the Subject Relative form is permitted even though a non-subject has
been relativized. In the minimally different (24)b the subject is specific, and thus a
DP, and the NSR verbal form is required. When the subject is a DP, the SR form is
barred, (24)d. Note though the position of the subject vis-à-vis the dative “top” (a
nominal expression that takes case) which I assume is an adjunct. As expected, when
the subject is an NP, it cannot raise, i.e. be in a structurally higher position than the
dative “top”. Compare the unacceptable (24)f with the acceptable (24)a.

24) a. üzeri-ne ögrenci yaslan-an araba
top-DAT student lean-SR car
‘the car that a student is leaning on’

b. üzeri-ne ögrenci-nin yaslan-dıg-ı araba
top-DAT student-GEN lean-NSR car
‘the car that the student is leaning on’

c. ögrenci-nin üzeri-ne yaslan-dıg-ı araba
student-GEN top-DAT lean-NSR car
‘the car that the student is leaning on’


126

d. *üzeri-ne o ögrenci(-nin) yaslan-an araba
top-DAT that student-GEN lean-NSR car

e. *o ögrenci(-nin) üzeri-ne yaslan-an araba
that student-GEN top-DAT lean-NSR car

f. *ögrenci üzeri-ne yaslan-an araba
student top-DAT lean-SR car


We formulated a generalization, or rather, a mnemonic, to handle the facts: the SR
form in Turkish relatives is licensed when the EPP of T attracts a +Wh DP to its
specifier. When a non-Wh expression occupies [Spec, TP], the SR form is barred.
Since the SR form is the instantiation of a +Wh element having moved into (and then
out of) [Spec, TP], we can use the acceptability of the SR form as a diagnostic of
attraction of a DP (in this case albeit a +Wh one) by the EPP on Tº.

3 The SR and Movement through [Spec, TP]


We are assuming that the SR form is the instantiation of a +Wh element having
moved through [Spec, TP]. Use of the SR form with NP subjects when extracting
non-subjects is not completely free, however. Using the acceptability of the SR form
as a diagnostic, let us examine when the SR form is acceptable and when it is not,
with the aim of discovering the constraints on the EPP of T
118
in Turkish.


118
This paper examines the EPP only in embedded contexts which may have different properties than
matrix contexts. For example, left dislocation is degraded in embedded environments.
i) a. Down the street ran the boys.
b. *He announced/believed/promised that down the street ran the boys.

127

3.1 Relative clauses with non-specific subjects and the SR
form

Subjects in existential constructions are non-specific. In an existential construction,
the subject, being an NP, cannot raise, and the EPP of T must target another DP. We
would thus predict that relativization of a non-subject DP in an existential
construction would license the SR form. This is indeed the case, as in (25). In the
marginal (25)b, the interpretation of the subject must be a specific type of
“incorporation”. Note that the example improves with the addition of “this” to the
subject as in (25)c.

25) a. incorporation ol-an dil-ler-de
incorporation be-SR language-pl-LOC
‘languages that have incorporation’
Literally: languages in which there is incorporation

b. ?incorporation-ın ol-dug-u dil-ler-de
incorporation-GEN be-NSR-3s language-pl-LOC
‘languages which have [this kind of] incorporation’

c. bu incorporation-ın ol-dug-u dil-ler-de
this incorporation-GEN be-NSR-3s language-pl-LOC
‘languages which have this [kind of] incorporation’

Let’s see what happens with different verb classes. The verb in examples (26) and
(27) is unaccusative. As these examples show, both RC forms are acceptable
depending on whether or not the subject is specific. The (a) examples with the SR
form have non-specific NP subjects, while the NSR form is required in the (c)
examples with DP subjects. The (b) sentences demonstrate the case on the relativized
element prior to movement: the element that was attracted by the EPP of T in (26)a is

128

a locative, and in (27)a, a dative. Crucially, the SR form is acceptable (when
extracting non-subjects) with unaccusative verbs when the subject is an NP.

26) a. yagmur yag-an bölge
rain rain-SR region
‘the region where it rains/where (typically) there is rain’

b. Bölge-de yagmur yag-ıyor.
region-LOC rain raining-pres.prog.-3s
‘It’s raining in that region.’

c. yagmur-un yag-dıg-ı bölge
rain-GEN rain-NSR-POSS-3s region
‘the region where it is raining/rained’

27) a. gemi yanas-an liman
ship sidle-SR harbor
‘the harbor that a ship is sidling up to’

b. Liman-a gemi yanas-ıyor.
harbor-DAT ship sidle-pres.prog.-3s
‘A ship is sidling up to the harbor’

c. gemi-nin yanas-tıg-ı liman
ship-GEN sidle-NSR harbor
‘the harbor that the ship is sidling up to’

This contrasts with unergative verbs, such as (28) and (29), where only the NSR form
is permitted. Notice that the subject must be specific (a DP) and receive overt case; a
non-specific, bare subject is unacceptable in either RC form.

28) a. *gençler dans ed-en salon
youth dance do-SR club
Intended: ‘the club where young people dance’

b. gençler*(-in) dans et-tig-i salon
youth-GEN dance do-NSR club
‘the club where young people dance’

129

29) a. *atlet-ler kos-an saha
athlete-pl run-SR track
Intended: ‘the track where athletes run’

b. atlet-ler*(-in) kos-tug-u saha
athlete-pl-GEN run-NSR-3s track
‘the track where athletes run’

So far, we see that locatives, genitive possessor DPs, and datives can satisfy T’s EPP
in unaccusatives with NP subjects; but, this is not acceptable with unergatives.
Let’s now turn to transitive constructions. In (30)a, the relativized element is
the possessor of the object DP, and as expected the non-subject relative (NSR) verbal
form is licensed when the subject is specific. Use of the SR form is also acceptable,
when the subject is non-specific, as in (30)b. Note that the non-specific subject being
an NP cannot raise from its base-generated position to a position higher than the
accusative direct object in [Spec, vP], (30)c. In (30)b, the relativized element, i.e. the
element that moved through [Spec, TP], was the genitive possessor girl of the
accusative object, girl’s leg. Thus, a genitive can satisfy the EPP on Tº.

30) a. [
CP
arı-nın [
VP
[
DP
Ø
1
bacag-ın-ı] sok-dug-u]] kız
1

bee-GEN leg-POSS-ACC sting-NSR-3s girl
‘the girl whose leg the bee stung’

b. [ [Ø
1
bacag-ın-ı] arı sok-an ] kız
1

leg-POSS-ACC bee sting-SR girl
‘the girl whose leg some bee stung’

c. *[arı [Ø
1
bacag-ın-ı] sok-an ] kız
1

bee leg-POSS-ACC sting-SR girl

d. *Ø
1
arı sok-an kız
1

Ø bee sting-SR girl
Intended: ‘the girl some bee stung’ [Actual: ‘the girl who stung some bee’]


130

Note that, as expected, the SR form is unacceptable when relativizing an accusative
object even when the subject is non-specific, (30)d. As we noted, an accusative
case-marked element cannot move to [Spec, TP]. Thus, a non-subject +Wh-DP can
raise to [Spec, TP] in a transitive construction as long as the DP does not bear
accusative case (and, of course, the move obeys minimality). This is in line with
Chomsky (1995, 2000); once structural case has been checked on a DP, it can no
longer undergo A-movement to a structural case assigning/checking position. Recall
that it is only structural case that causes this restriction; inherently case-marked
expressions are free to undergo A-movement, perhaps because they can get a second
case (Hong 2002).
119
Because in (30)b a genitive possessor is patterning like the
dative and locative DPs in these SR constructions, we must assume that, at least in
Turkish, genitive case on a possessor is inherent, not structural.
120
Recall that the RC
subject (i.e. in the NSR form with overt subject) bears genitive case. I assume that
the genitive on RC subjects in Turkish is structural case assigned either by T or by a

119
Although I have argued against the idea that T° has a case feature it must discharge, nothing in the
story I am presenting rests on this assumption, and I leave the option open that the Inverse Case Filter
may hold. I point to arguments in favor of one view or the other as a way of expanding the discussion.
120
Chomsky (1995) considers genitive case inherent because it is associated with theta-marking (p.
114). This would distinguish possessor genitive from clausal genitive on a subject. See also fn. 117. A.
Ince (p.c.) provides the following data as evidence that Turkish has two types of genitive case: inherent
and structural. In the NP-Deletion example in (i), genitive case remains on the possessor. In (iia), the
specific subject in the embedded clause bears genitive case. However, in the sluiced structure (iib),
genitive case on the subject is ungrammatical. But see fn. 118.
(i) Bu kim-in araba-sı. Ahmet’in arabası.
this who-GEN car-poss. Ahmet-poss
‘Whose car is this? Ahmet’s’
(ii) a. Ahmet birin-in sen-i ara-dıgı-nı söyle-di.
Ahmet one-GEN you-ACC call-COMP-ACC said-pst
‘Ahmet said someone called you’
b. Ahmet birin-in sen-i ara-dıgı-nı söyle-di. Ama kim(*-in) bilmi-yorum.
Ahmet one-GEN you-ACC call-COMP-ACC said-pst but who(*-GEN) know-pres-1s
‘Ahmet said someone called you. But I don’t know who’

131

V-T-C amalgam.
121
On the other hand, the possessor genitive assigned by D is
inherent case because it patterns with all other inherently case-marked elements in
being able to raise to T in unaccusatives with NP subjects.
122,123
The unergatives in (28) and (29) had animate subjects. One might wonder if
sentiency or animacy might be a factor. But, as demonstrated in the unacceptable
unergative examples in (31) and (32), unergatives with inanimate NP subjects also
bar movement of another element to T. Example (31) is an attempt to raise a locative
to [Spec, TP], and (32), an ablative.

31) *araba-lar gid-en cadde
cars go-SR street
‘the street that cars go on.’

32) a. *tas-lar düs-en tepe
rocks fall-SR hill
‘The hill rocks fall from’

121
See Hiraiwa (2001) for arguments that a possessor genitive and RC subject genitive are different
types of case and that the subject genitive case is assigned by a V-T-C amalgam.
122
Of course, if one wanted to assume that genitive case assigned by D° is also structural, one could
argue that nominative case on top of accusative leads to a PF crash. Since inherent or lexical case is
thematic, these may be interpreted at LF whereas structural case must be PF-interpretable. Since the
case assigned by the RC T° is genitive, movement of the structural genitive possessor to [Spec, TP]
does not lead to PF crash because of homomorphism. This would mean that both the accusative object
and its possessor are candidates for Attract by T’s EPP. The former leads to a PF crash whereas the
latter will lead to convergence. I must admit that I find the idea that the case assigned by D° is
structural tempting because it is not theta-related and does seem to require raising to a Spec-Head
configuration. Not surprisingly, I will not open the can of worms about an EPP feature on D.
123
Note that a possessor genitive as in the response to the question in (i), does not allow an additional
inherent case, ablative in (ii), or a structural accusative case, (iii). These facts suggest that possessor
genitives are structural. Note that structural case does not seem to allow a second case, structural or
inherent. This is consistent with Hong’s 2002 observations in Korean.
(i) Kim-in köpeg-i kaçtı? Sen-in.
who-GEN dog-AGR ran.away you-GEN
‘Whose dog ran away? Yours.’
(ii) *sen-in-den korkuyor-um
you-GEN-ABL fear.1s
‘I am afraid of yours’
(iii) *sen-in-i gördüm
you-GEN-ACC saw-1s
‘I saw yours’

132


b. Tepe-den taslar düsü-yor.
hill-ABL rocks fall-PRES.PROG.
‘Rocks fall from this hill.’


4 Toward an explanation


We are assuming that once structural case has been checked on a DP, it can no longer
undergo A-movement to a case-checking position.
124
We would assume that another
(non-structurally case-marked) DP in the structure should be able to satisfy the EPP
of T as long as the move obeys minimality. In a transitive construction, the EPP of v
must be satisfied before T° merges with vP. At this point in the derivation, the DO in
[Spec, vP] will already have been assigned structural case, and will no longer be a
candidate for Attract by the EPP of T.
125
The question is, will the accusative object
be invisible for Attract or will it serve as an intervener for movement of another
nominal element lower in the structure?
In (30)b, a transitive with an NP-subject, the genitive possessor of the
accusative DO raised to [Spec, TP]. However, as example (33) demonstrates, a
locative cannot move to [Spec, TP] over an accusative object. Thus we can conclude
that accusative object induces intervention effects for A-movement. This is
demonstrated in the tree in (34) for example (33).

124
I assume that case is assigned in a Spec-Head configuration. T can, trivially, assign case to a DP
without case; it can also assign nominative case to an inherently case-marked DP. This is consistent
with Hong’s (2002) observation about multiple case on DPs in Korean: structural case can be added
onto an inherently case-marked element, but stacking of structural case onto structural case does not
seem to be possible. See fn. 119.
125
But see fn. 122 where I proposed that perhaps the DP could move to just such a position, but the
derivation would crash at PF.

133

33) *[Ø
1
çocuk-lar-ı arı sok-an] orman
1

Ø children-ACC bee sting-SR woods
‘the woods where a bee/bees sting children’

34)
CP

woods
+Wh TP C°

t-woods
+Wh vP T°
+EPP
DP-children-ACC
NP-bee
VP v°

woods-LOC
+Wh DP V°
children sting





The unacceptable SR RC in (33) contrasts with the acceptable example in (35). How
do we account for the difference?

35) [Ø
1
çocuk-lar-a kurt saldır-an] orman
1

Ø children-DAT wolf attack-SR woods
‘the woods where a wolf/wolves attack children’

Recall that we had assumed that vº also has an EPP feature that must be
checked/deleted.
126
A DP direct object can satisfy v’s EPP, or in the absence of one,
another nominal element must move to [Spec, vP]. It is important to note that with
both Tº and vº, if they do assign (overt) case, the case-marked element must be the

126
A footnote in the previous chapter contained a discussion about the EPP on v° where I speculated
that the vP projection might be analogous to the TP projection. That is, T° has an EPP feature when
selected by C°, and v° has an EPP feature when selected by v*°, the head of a (perhaps A-bar) v*P
projection above vP.

134

one Attracted to the case-assigning head by the EPP. Thus, in (33), the direct object
must raise out of the VP, leaving the locative in VP. If the direct object does not
raise, the derivation will crash because of a Case Filter violation. In example (35), on
the other hand, the internal argument is assigned inherent dative case which does not
need to move for case. This leaves [Spec, vP] available for the +Wh-locative.
As demonstrated in the tree in (36), the dative internal argument children and
the locative woods are in the same minimal domain, and so equidistant from v.
(Actually, under one definition, the locative being in the specifier is closer.) We can
assume that the Turkish verb attack “saldıran” in (35) is quirky and does not
assign/check accusative case. In the derivation in (36), the locative woods moves to
[Spec, vP] and deletes v’s EPP feature in , but is not assigned structural case, and
so, can further undergo A-movement to [Spec, TP] as in .

36)
CP

woods
+Wh TP C°

t-woods
+Wh vP T° [+EPP]

woods-LOC
+Wh NP-wolf
VP v° [+EPP, -Case]

woods-LOC
+Wh PP/DP V°
children-DAT attack


The derivation in (36) has implications for the “movement for case” story. Let’s
assume that v° had a case feature that needed to be deleted or checked. If the dative
DP children moved to [Spec, vP], it would block the +Wh locative from raising to

135

[Spec, TP] and the SR form would be barred. If the locative woods moved to delete
v°’s case feature, it would be assigned structural accusative case on top of the
inherent locative case. This would be permissible; but now, once the +Wh-locative
had been assigned structural accusative case, it would be frozen for further
A-movement, and like other accusative objects, could not move to [Spec, TP],
bleeding the SR form. It seems that this example is another piece of evidence against
the movement is for case analysis.
One might ask whether the movement of the locative to [Spec, vP] in (36)
might not be scrambling. This is not likely as there is evidence that once scrambled,
an element is frozen. A constituent may raise out of a scrambled expression, but the
expression itself is locked in place for the rest of the derivation.
127


127
It is not possible to deal with the larger issue of scrambling in this thesis. Throughout this work,
however, we will observe effects and properties of scrambling. Crucially, we will see many examples
where a scrambled expression ameliorates minimality effects by carrying a +Wh-DP constituent
around an intervener. We also see evidence that a scrambled expression cannot move again because a
DP cannot scramble around an intervener and A-move, only a constituent can move once an expression
has scrambled. Thus, we are able to note three properties that emerge with respect to scrambling
which are important for the analysis in this work: 1) a scrambled expression does not delete an EPP
feature (because it does not move to the specifier of the functional head bearing an EPP feature), 2) a
scrambled expression is porous for movement (contra Chomsky (2005)), and 3) once an expression
scrambles, it cannot move again (See fn. 11 in Chapter 7 for speculations as to why this may be so.)
My assumption that scrambling freezes an expression was based on the fact that, once
scrambled, a DP cannot be relativized using the SR form. A scrambled DP adjoins below TP; it can
cannot move to [Spec, TP] and then to [Spec, CP] which would license the SR form. Tomo Fujii (p.c.)
points out, however, that the scrambled expression may not be frozen; the freezing effect may be
epiphenomenal to structural and semantic constraints. I assume the raising/promotion analysis of RCs,
where the RC head is an N° that is promoted from [Spec, CP] to a theta position in the matrix sentence.
This is consistent with the Turkish facts: Turkish RCs do not permit a possessor-possessee relative
head. The scrambled elements we are looking at in this work have just this configuration.
Furthermore, Japanese also has constructions such as [sofa’s top]+case, and [table’s under]+case ‘the
bottom of the table’, but although “top” and “bottom” bear case (and in Turkish they bear possessive
agreement with the possessor in the Spec) these terms to be semantically weak (i.e. they are pseudo-
PPs) and cannot be relative heads either in Turkish or Japanese. Even in English, in a context where
the child spilled juice on top of one of the sofas, you cannot say “*?the top that the child spilled juice
on”. However, in a context where a factory worker is painting tops of cars, it does seem okay to say
“the top that the worker painted (green)”. It seems more prudent then, to merely state that these larger
DP’s cannot be relativized, and that within the confines of this work, we are not able to determine
whether, once having scrambled, they can actually move to TP. There is no way to tease apart the
cause of the derivational crash when we try to relativize these scrambled expressions.

136

4.1 The base position of subjects and intervention by NPs
Let’s return now to our questions regarding the unacceptable unergative examples in
(28)-(29) and (31)-(32). Why was the SR form acceptable for unaccusatives but not
for unergatives? Previous research has suggested that the base position of the subject,
i.e. the expression that picks up the outermost theta position, in these verbs is
different,
128
i.e. that in an unaccusative structure, the subject is generated as the
complement of Vº, as in (37), whereas in an unergative, the subject is generated in
[Spec, vP], as in (38) for example (29)a. Note that the difference in these structures is
the position of the NP-subject which intervenes between the +Wh-expression and T in
(38), but not in (37).

37) [Ø
i
su ak-an] dam
i

water pour-SR roof
‘the roof water pours/drips from’


CP

roof
+Wh TP C°

t-roof
+Wh VP T° [+EPP]

roof-ABL
+Wh NP V°
water pours



Kural (1992), using WCO, scope and binding phenomena as diagnostics, argues that
scrambling in Turkish is A-bar movement, but crucially not to a C projection. There are suggestions
that there may be an A-bar projection above the verbal domain but below the inflectional domain (also
proposed by Baker & Stewart 2002). This is hierarchically equivalent to the position the scrambled
expressions here seem to occupy. Kural’s idea is appealing because the idea that scrambling is A-bar
movement to a projection immediately above vP provides us with a landing position as well as a reason
for the freezing effect.
128
Perlmutter’s (1978) “Unaccusative Hypothesis” and Burzio (1986) for evidence in Italian.

137

38) *[atlet-ler kos-an] saha
athlete-pl run-SR track
Intended: ‘the track where athletes run’

CP


TP C°

track
+Wh vP T° [+EPP]

NP-athletes
VP v° [-EPP, -Case]

track-LOC V°
+Wh run



Note also that I am making one more assumption about unergatives: that the v° head
does not have an EPP feature. This differs from the transitive construction in (35)
with the tree in (36) which had an internal argument that was assigned inherent dative
case. I had assumed that the v in this construction had an EPP feature but did not
assign case. The implication is that there are three v heads in Turkish: one with both
EPP and case features, one with only an EPP feature, and one with neither an EPP nor
a case feature.
I’d like to side-track momentarily to point out thatthe derivation in (36) is an
argument against the non-EPP story that movement to functional heads is for Case. If
we are right in assuming that an element that is assigned structural case is no longer a
candidate for A-movement, i.e. movement to another case assigning functional
projection, then the +Wh-locative in (36) could not have moved for case. For if it had
been assigned accusative case on top of its inherent locative case (which is
theoretically possible in the story here), it would be frozen and could not move to T to

138

trigger the SR form. Another possible explanation for the movement of the locative
to the vP is scrambling. The locative could scramble above the subject and then move
to [Spec, TP]. The facts throughout this thesis indicate that this option is not feasible.
The question that should be asked is why does the locative need to move to vP in the
first place? After all, the NP subject cannot satisfy the EPP on T. The differences
with respect to RCs in unergatives vs. unaccusatives provides us with insight to
answer such questions.
What we see in (38) is evidence for what Chomsky (2000) describes as a
defective intervention constraint (39).

39) Defective Intervention Constraint (DIC):

In the structure, u > µ > ¸ , where > is c-command, and µ and ¸ match
probe u, but µ is inactive, the effects of matching are blocked.

In terms of the DIC, in (38), the locative DP, track (¸), matches probe Tº (u). The
NP-subject, athletes (µ) is “inactive”, i.e. not a candidate to move to T, but it does
serve to block the locative from being attracted to T. The structure of (38) meets the
requirements of the DIC: µ (athletes) c-commands ¸ (track). Thus the NP subject is
an intervener for attraction of the locative by T’s EPP. This contrasts with (37) where
there is no interevener. We must conclude that the DIC plays a role in the unergative
examples: an NP, which itself cannot satisfy the EPP of T, blocks the movement of a
DP from its c-commanding domain.
129


129
Another way of explaining the intervention effects of NPs might be to propose that they are
unspecified, i.e. lack a value. Thus, while an NP may bear the kind of features that would be picked up

139

5 Versions of the EPP


Within the generative framework, there are three core versions of the EPP each with
different implications. The first, based on Chomsky (1981, 1982) is configurational:
clauses must have subjects. In the second version, based on Chomsky (1995), the
EPP is a strong D feature of a functional head and the subsequent pied-piping of the
phrase entire phrase. Recently, Chomsky (2000, 2001) relies on Match and Agree
operations for ç-feature checking, precluding the motivation for movement. In this
version, the EPP is a requirement that certain functional heads must have a specifier.
Having argued throughout this chapter that there is an EPP, the question is, is
the EPP (in Turkish) a feature or a structural requirement? If the EPP were merely a
structural requirement as argued by Lasnik (2001), we would not see intervention
effects. The data indicates that not only do c-commanding DPs intervene in the
raising to [Spec, TP], but even those expressions that cannot themselves satisfy the
EPP intervene. These include DPs such as accusative objects, and NPs such as the
subject of an unaccusative. If the EPP was only a requirement that the Spec position
of a functional head be occupied, these expressions would not induce intervention
effects. It is only when specific features must be checked can the head be selective in
admitting or barring movement to its specifier. If the EPP were not sensitive to
features, an NP subject of an unergative verb could not block another DP in its

by the EPP of an Attracting head, the features are unvalued and thus unable to delete the EPP on that
head. This idea is radically different from what I have been proposing thus far. In this version, NPs do
move, they are attracted by the EPP, but they fail to delete or check the EPP. Note that this version
requires the idea of unique Spec positions. Once an NP has been attracted by the EPP, the derivation
will crash; a DP cannot come along and check the EPP on a head by merging in a second Spec
position.

140

c-commanding domain from moving to T. We are able to deduce that the NP subject
is an intervener for movement because it is possible to skirt the NP via scrambling.
We will see this operation in greater detail in later chapters, but at this point note that
the unacceptable (31) repeated as (40), is acceptable in (41). As demonstrated in
(41)b, the DP/DP phrase on the street with the structure [street’s top-AGR]-LOC
functions as a carrier of the +Wh-street around the NP subject, cars. After, street has
gotten around the subject, it is free to move to [Spec, TP] licensing the SR form. If
the EPP were merely a structural requirement, what would prevent an NP from
satisfying it by moving to [Spec, TP]. Obviously, there is some featural requirement
that an NP cannot satisfy.

40) *araba-lar gid-en cadde
cars go-SR street
‘the street that cars go on.’

41) a. [[Ø
1
üst-ün]-de araba gid-en] cadde
1

top-AGR-LOC car go-SR street
‘the street that cars go on (top of)’

b. [
CP
[
TP
[
DP/PP
Ø
1
üst-ün]-de
k
[
vP
araba t
k
gid-en] cadde
1

top-AGR-LOC car go-SR street



6 Conclusion


This chapter has looked at the consequences of and constraints on the EPP of T in
Turkish. Although limited to subordinate clauses, the findings prove interesting
because they support previous observations about the syntactic structure of different

141

verb classes, case-marking of arguments and intervention effects. Using the
acceptability of the SR relative clause form as a diagnostic of movement to [Spec,
TP], we have seen intervention effects of Attraction by the EPP of T. We have seen
arguments for an EPP feature which can only be satisfied by a DP (a specific nominal
expression). Nominals that cannot satisfy T’s EPP feature do, however, serve as
interveners for the raising of other DPs in their c-commanding domain. In clauses
with an NP-subject, some other expression, a DP, must raise to delete T’s EPP feature
for convergence. NP-subjects and accusative objects induce intervention effects; both
are barred from moving to T, the former cannot satisfy the EPP and the latter cannot
move to a structural case position. Although neither can satisfy the EPP on T, they
block movement of a lower DP.
Approaches based on theta-roles are also not adequate. Structural hierarchy
rather than grammatical functions such as subject or object determines the ability of a
DP to raise to T. An NP subject of an unaccusative verb because of its position as the
complements of V° does not block another DP raising to T. Any inherently case-
marked expression can be relativized using the SR form in unaccusatives because the
subject will not intervene. By contrast, the NP subject of an unergative verb, which is
generated in a structurally higher position, does block the raising of DPs from its c-
commanding domain. Similarly, in transitive constructions, a DP-direct object raises
to vP. Although the accusative object itself cannot satisfy T’s EPP feature, it serves
to block the raising of a lower DP such as a locative. A genitive possessor from
within the accusative object, however, can raise to T. This is because neither the
accusative object nor the NP-subject c-commands this possessor DP. These facts

142

give support to Chomsky’s (2000) Defective Intervention Constraints which stipulate
that elements that cannot themselves delete features serve to block other elements in
their c-commanding domain from checking those features. In sum, the EPP seems to
be a featural requirement; as long as there is no intervening c-commanding nominal,
any DP can check T’s EPP feature with the proviso that it not be structurally case-
marked (i.e. not bear accusative case).


143

Chapter 5: Human/Non-human Distinctions in Turkish



1 Background



In Chapter 4, we saw that the movement of a DP to T obeys Minimality. The
example in (1) has an unaccusative verb. This means that the “subject”
130
is the
complement of V° and other nominal expressions are generated higher than the
subject. When relativizing a non-subject, the SR form is generally not acceptable in
clauses with unergative, or intransitive, verbs. As demonstrated in the tree in (1)b for
(1)a, the NP-subject cars intervenes between the Wh-element and T. There is no way
to express this clause with a non-specific subject; it must be stated with a specific
subject using the NSR form, as in (2).
131


1) a. *[araba(-lar) gid-en] cadde
car(s) go-SR street
‘the street that cars go on.’





130
By subject I mean the expression that merges into the outermost theta position of the verb. Thus,
the subject of an unaccusative verb is its complement, while the subject of both transitive and
intransitive verbs is the expression merged in the theta position in [Spec, vP]. Note that this excludes
inherently case-marked expressions which, analogous to PPs, are lexically theta-marked.
131
Note that example (6) must do double-duty, and is used to mean both “the specific cars” and “the
cars in general”. As we will see later in this chapter, another way of denoting non-specificity in
Turkish is by use of a partitive construction, so example (i) is also a possibility.
(i) [araba-lar-dan git-ti-i] cadde
car-pl-ABL go-NSR street
‘the street that (some (of the)) cars go on’


144

b.
CP

track
TP C°

t-street
+uWh vP T°

NP-cars
VP v°

DP-street V°
+Loc go



2) [araba(-lar)-ın git-ti-i] cadde
car(s)-GEN go-NSR street
‘the street that the cars go on’

I have argued that (1)a is a Minimality violation; the NP subject c-commands the
locative and is an intervener. There is another derivation that literally gets around the
Minimality issue by letting cadde ‘street’ become a “free-rider” on a scrambled
expression. A larger expression, for example the DP/PP [on the street’s top] with
street as the specifier as in (3)a and (3)b, can scramble around the subject, (3)c. The
specifier street is now free to move out of the scrambled expression raising to [Spec,
TP] without intervention from the subject as in the now acceptable SR clause in (3)d.

3) a. [
DP/PP
[
DP
cadde-nin] üst-ün]-de
street-GEN top-AGR-LOC
Lit: ‘on the street’s top’

b. [araba [
VP
[
DP
[
DP
cadde-nin] üst-ün]-de gid-er]].
car street-GEN top-AGR-LOC go-AOR
Lit: ‘Cars go on the street’s top’ (i.e. ‘Cars go on/over the street’)



145

c. [ [
DP
[
DP
cadde-nin] üst-ün-de]
1
[
vP
araba [
VP
t
1
gid-er ]
street-GEN top-AGR-LOC car go-AOR
‘It is cars that go on/over the street’

d. [
CP
... [
DP
[
DP
Ø
2
] üst-ün-de]
1
[
vP
araba [
VP
t
1
gid-en] cadde
2

top-AGR-LOC car go-SR street
‘the street that cars go on (the top of)’

We see the same phenomena with transitive verbs. First, notice that as shown in (4)b
for the unacceptable example in (4)a, an accusative direct object cannot move to T.

4) a. *[kedi tırmala-yan] çocuk
cat claw-SR child
Intended: ‘the child that a cat clawed’ [Actual: ‘the child that clawed a cat’]

b. CP


TP C°

+WH-DP-child+ACC
vP T°

+Wh-DP-child+ACC
NP-cat
VP v°

+Wh-DP-child V°
claw



It is not surprising that a structurally case-marked element cannot move to another
structural case-assigning position. On the other hand, we have seen that Turkish
allows inherently case marked elements in the [Spec, TP] position. The relativized
expression in (5)a bears inherent locative case prior to movement to [Spec, TP], as
shown in the sentence in (5)b, and in (6)b we see that the relativized expression in
(6)a bears inherent ablative case prior to promotion.


146

5) a. [mısır yetis-en] tarla
corn grow-SR field
‘the field where corn grows’

b. (Bu) tarla-da mısır yetisiyor.
(this) field-LOC corn grows
‘Corn grows on (this) field’

6) a. [fare çık-an] delik
mouse come.out-SR hole
‘the hole that mice come out of’

b. Delik-den fare çıkı-yor.
hole-ABL mouse come.out-pres.prog.
‘A mouse/mice is/are coming /came out of the hole.’


Returning to the issue of Minimality, we have so far determined that i) an intervening
NP subject blocks raising of a +Wh-expression to T; ii) although an accusative direct
object is higher than an in-situ subject, the object is barred from moving to T; and iii)
inherently case-marked expressions are permitted in [Spec, TP]. The example in (7)a
demonstrates that a genitive possessor of an accusative object can also move to [Spec,
TP] as evidenced by the acceptability of the SR form. The derivation of (7)a appears
in (7)b. Once the direct object has raised to [Spec, vP], the subject no longer
intervenes between the object and T. Whereas the accusative object itself, the DP
[child’s arm] is barred from moving to [Spec, TP], the possessor child is free to raise
out of the DP to T, as in move .

7) a. [ [Ø
1
kol-u]-nu kedi tırmala-yan] çocuk
1

arm-POSS-ACC cat claw-SR child
‘the child whose arm a cat clawed’



147

b. CP


TP C°

+Wh-child
vP T°

DP+ACC
NP-cat
VP v°

DP V°
claw
+Wh-child
+GEN arm D°




Crucially, the direct object [child’s arm] is not +Wh, only the element in its Spec,
child, is. Thus, it is only the genitive possessor of the accusative object that raises in
from the vP to [Spec, TP] after which it again raises to [Spec, CP], triggering the
SR form.
The conclusion is that as long as the relativized expression can get around the
subject, and the subject itself does not have to, indeed, cannot raise to T, the
relativized expression can move to [Spec, TP] and license the SR form.


2 The Problem of Human Subjects


Having determined that the SR form is only licensed when the relativized expression
can move to, and through, [Spec, TP], we saw that this move must obey Minimality.
We also saw that raising of a possessor out of an accusative object which itself has
raised above a non-specific NP-subject avoids a Minimality violation. The problem

148

is that whereas this strategy generally holds, it results in unacceptability when the
subject is +human. The near-minimal sets (8) through (10) are transitive
constructions where the relativized element is the possessor of the accusative direct
object. However, the SR form is acceptable only when the subject is –human; a
+human subject results in unacceptability of the SR form.

8) a. [[Ø
1
kayıg-ı]-nı nehir sürükley-en] balıkçı
1

boat-POSS-ACC river drag-SR fisherman
‘the fisherman whose boat a river dragged’

b. *[[Ø
1
kayıg-ı]-nı köylüler kıyı-ya sürükley-en] balıkçı
1

boat-POSS-ACC villagers shore-DAT drag-SR fisherman
‘the fisherman whose boat villagers dragged onto the shore’

9) a. [[Ø
1
ev-i]-ni fırtına yık-an] aile
1

house-POSS-ACC tornado raze-SR family
‘the family whose house a tornado razed’

b. [[Ø
1
ev-i]-ni belediye yık-an] aile
1

house-POSS-ACC municipality raze-SR family
‘the family whose house a municipality razed’

c. *[[Ø
1
ev-i]-ni isçi/asker(-ler) yık-an] aile
1

house-POSS-ACC worker/soldier(-pl) raze-SR family
‘the family whose house (a) worker(s)/soldier(s) razed’

10) a. [[Ø
1
çocug-u]-nu arslan yiy-en] anne
1

child-POSS-ACC lion eat-SR mother
‘the mother whose child a lion/lions ate’

b. *[[Ø
1
yavru-su]-nu avcı(-lar) yiy-en] geyik
1

young-POSS-ACC hunter(s) eat-SR deer
‘the deer whose young a hunter/hunters ate’

c. [[Ø
1
yavru-su]-nu avcı-lar-ın ye-di-i] geyik
1

young-POSS-ACC hunter-pl-GEN eat-NSR-3s deer
‘the deer whose young the hunters ate’


149

As demonstrated in (11), the difference does not lie in animacy. The subjects in both
(11)a and (11)b are animate; but in the unacceptable (11)b, the subject is +human.
Furthermore, (11)b becomes acceptable in the NSR form (11)c, as long as the subject
is specific, and has moved to T
132
. We can assume this to be so because the subject
bears genitive case, which it cannot get unless it has raised to T. Support for this
assumption appears in (12).

11) a. [pro bir kız-ın yan-ın-da ol-du-u]-nu gör-dü-ler.
one girl-GEN side-his-LOC be-NSR-3s-ACC see-PAST-3pl
‘They saw that (of the salient girls) one (of them) was by his side.’

b. pro [yan-ın-da bir kız ol-du-u]-nu gör-dü-ler.
side-his-LOC one girl be-NSR-3s-ACC see-PAST-3pl
‘They saw that there was a girl by his side.’

Subjects of embedded clauses in Turkish generally bear overt genitive case. We
assumed case is assigned in a Spec-Head configuration and that the EPP of a case
assigning head attracts a DP to its Specifier. We also assumed that non-specifics are
NPs that neither satisfy the EPP nor require case. We assumed that locatives mark
the edge of the VP (Kural 1992). In (11)a, the embedded subject of the copular

132
Note that the accusative direct object in the NSR form is higher than the subject in these examples.
I argue that the genitive subject in the NSR form is in [Spec, TP], so I assume that the possessor-
possessee DP-object scrambles out of the vP, to a position above the subject. (There may be a
projection v*P above vP but below TP as suggested by Lasnik (1998, 2002)and Chomsky (2005) to
which the scrambled expression adjoins.) It is beyond the scope of this work to discuss this scrambling
further, but I will point out that the unscrambled version of the NSR form is also acceptable as in (i)
for example (13c) because the Wh-expression geyik undergoes A-bar movement from with the
accusative DP in [Spec, vP]. As expected, this order is not possible with the SR form because the Wh-
geyik is A-moving to [Spec, TP] and the subject intervenes.
(i) [avcı-lar-ın [
vP
Ø
1
yavru-su]-nu ye-di-i] ] geyik
1

hunter-pl-GEN young-POSS-ACC eat-NSR-3s ] deer
‘the deer whose young the hunters ate’

150

clause has raised above the VP and bears genitive case.
133
It must also receive a
specific interpretation. This contrasts with (11)b, where the embedded clause is an
existential construction. In (11)b the locative is above the non-specific subject which
bears no case morphology. A non-specific subject must remain in-situ and must be
bare, whereas genitive case on the subject can be viewed as evidence that the subject
is in [Spec, TP]. Returning then to the acceptable (10)c, because the subject bears
genitive case, it must be in [Spec, TP], it must receive a specific interpretation, and
based on our assumptions, it must be a DP.
No other animacy heirarchy among the arguments is playing a role in the
grammaticality of the examples. Let’s look at several alternatives. In examples (8)
and (9) above, the direct object is composed of a human possessor and an inanimate
possessee. In (10)a, possessor and possessee are both human, and in (10)b, they are
both animate. In (12) below, the possessor is human and the possessee is animate.
And, in (13), the possessor is inanimate and the possessee is human. In all these
examples, the same pattern obtains. The SR form becomes unacceptable when the
non-specific subject is +human.
134
As we see (12)c-d, the NSR form is acceptable for
both human and non-human subjects with the stipulation that the subject be specific.

133
I assume that the subject of a copular structure merges as a complement of V° and that locatives
adjoin to VP. Thus a locative in a copular sentence merges above (and linearly to the left) of the
subject.
134
Zimmer (1987) notes that the RC with the human subject in (i) is quite marginal whereas (ii) is fine.
He suggests that this is “presumably because humans are inherently more individualized and topic-
worthy than non-humans.”
(i) ??bir doktor otur-an ev
one doctor live-SR house
‘a house in which a doctor lives’
(ii) bir köpek bul-un-an ev
one dog find-PASS-SR house
‘a house in which there is a dog’ (Literally: ‘a house in which one finds a dog’)

151

12) a. [[Ø
1
ineg-i]-ni tren ez-en] köylü
1

cow-POSS-ACC train run.over-SR peasant
‘the peasant whose cow a train ran over’

b. *[[Ø
1
ineg-i]-ni soför ez-en] köylü
1

cow-POSS-ACC driver run.over-SR peasant
Intended: ‘the peasant whose cow a driver ran over’

c. [[Ø
1
ineg-i]-ni soför-ün ez-di-i] köylü
1

cow-POSS-ACC driver-GEN run.over-NSR-3S peasant
‘the peasant whose cow the/*a driver ran over’

d. [[Ø
1
ineg-i]-ni tren-in ez-di-i] köylü
1

cow-POSS-ACC train-GEN run.over-NSR-3S peasant
‘the peasant whose cow the train ran over’

To rule out as much as possible the other factors that may be responsible for the
difference, I include the minimal pairs in (13). In Turkish, the words ‘publisher’ and
‘publishing house’ are similar to the English in that they contain the same root plus a
morpheme. In Turkish, ‘publish’ yayın plus the ‘-er’ morpheme -cı, denote the
+human noun ‘publisher’ yayıncı, and ‘publish’ plus the word ‘house’ ev denote the
non-human organizational ‘publisher’, or ‘publishing house’, yayınevi. As
demonstrated in (13)a, whereas ‘publishing house’ is acceptable as the subject of the
SR clause, the +human ‘publisher’ is not. Both are possible as clausal subjects, as
shown in (13)b, but only if specific, and case-marked.

13) a. [[Ø
1
yazar-ı]-nı yayınevi /*yayıncı aray-an] makale
1

author-POSS-ACC publishing.house/publisher search-SR article
‘the article whose author publishing.houses/*publishers are looking for’




Although I do not find this explanation adequate—indeed, the purpose of this chapter is to find a
principled syntactic explanation as to why this may be so—it demonstrates that the phenomena has
been previously remarked upon.

152

yayınevi-nin
b. [[Ø
1
yazar-ı]-nı /yayıncı-nın ara-dı-ı] makale
1

author-POSS-ACC publishing.house-GEN search-NSR-3S article
/publisher-GEN
‘the article whose author the publishing.house/the publisher is/was looking for’


3 Toward a Solution


In the previous section, we saw that in otherwise acceptable transitive SR relative
clauses, a +human subject leads to unacceptability. Interestingly, there are other
instances in Turkish where a +human nominal behaves differently from its non-
human counterpart.
135
As shown in (14), in matrix sentences, a non-human non-
specific subject is restricted to the immediate preverbal position, i.e. must remain in
situ. The examples in (15) show that this restriction does not hold for +human non-
specific subjects.

14) a. Agaç-tan bir elma düs-tü
tree-ABL one apple fall-PST
‘An apple fell from a tree’

b. *Bir elma agaç-tan düs-tü
one apple tree-ABL fall-PST

15) a. Agaç-tan bir çocuk düs-tü
tree-ABL one child fall-PST
‘A child fell from a tree’

b. Bir çocuk agaç-tan düs-tü
one child tree-ABL fall-PST

135
Most of the examples (14)-(20) are from Erguvanlı (1984) which includes the following observation
based on animacy: [-animate] indefinite subjects are restricted to the immediate left of the verb 1- in
intransitive sentences and 2- in transitive sentences with a definite [+animate] DO. I show that the
contrast is more aptly described in terms of human/non-human features and specificity, rather than
animacy and definiteness.

153

Non-specific non-human direct objects cannot have case (16)b (or a plural marker
(16)c). Note that overt case on ‘apples’ in (16)d-e)
136
yields a specific interpretation.
This contrasts with +human direct objects which must bear accusative case, as shown
in (17)a-c.
137
A definite or specific interpretation is achieved by means of a
demonstrative, (17)d-e.

16) a. Ben elma sev-er-im
I apple like-AOR-1s
‘I like apples’

b. *Ben elma-lar sev-er-im
I apple-pl like-AOR-1s

c. *Ben elma-lar-ı sev-er-im
I apple-pl-ACC like-AOR-1s

d. Ben bu elma-lar-ı sev-er-im
I these apple-pl-ACC like-AOR-1s
‘I like these apples’

e. Ben elma-lar-ı sev-di-m
I apple-pl-ACC like-PST-1s
‘I liked the apples’

17) a. *Ben insan sev-er-im
I human like-AOR-1s

b. *Ben insan-lar sev-er-im
I human-pl like-AOR-1s


136
Use of the aorist tense in (16)d would have been odd without the demonstrative ‘this’, i.e. the
sentence would mean ‘I like the apples (we grow rather than the pears)’. I include the past tense
example in (16)e to demonstrate that the demonstrative is not necessary to denote specificity.
137
Although Erguvanlı (1984) did not include it, the non-plural example with accusative case is also
unacceptable with the unmarked interpretation of ‘I like apples’. (i) has the interpretation of ‘I like
apples [not some other fruit]’ and can only be uttered as a contrastive response to someone else’s
utterance about some other fruit.
(i) Ben elma-yı sev-er-im
I apple-ACC like-AOR-1s


154

c. Ben insan-lar-ı sev-er-im
I human-pl-ACC like-AOR-1s
‘I like people’

d. Ben bu insan-lar-ı sev-er-im
I these human-pl-ACC like-AOR-1s
‘I like these people’

e. Ben bu insan-lar-ı sev-di-m
I these human-pl-ACC like-PST-1s
‘I liked these people’


Although Turkish has relatively free word order, the transitive constructions in (18)
demonstrate that a non-human object can never precede a +human subject, even a
non-specific one. The exception is (18)f where the object is in Topic position, i.e.
must be D-linked.

18) a. [bir adam] [bir bahçe] suluyor
one man one garden watering
‘A man is watering a garden’

b. [bir adam] [bir bahçe]-yi suluyor
one man one garden-ACC watering
‘A man is watering a (specific) garden’

c. [bir adam] [bahçe]-yi suluyor
one man garden-ACC watering
‘A man is watering the garden’

d. *[bir bahçe]-yi [bir adam] suluyor
one garden-ACC one man watering

e. *[bir bahçe] [bir adam] suluyor
one garden one man watering

f. bahçe-yi [bir adam] suluyor
garden-ACC one man watering
‘The garden, a man is watering (it).’



155

As shown in (19), a non-human non-specific subject cannot precede a +human
specific object.
138
No such restriction applies to non-human specific objects.
139


19) a. Ali-yi ev-de bir sürpriz bekliyor
Ali-ACC home-LOC one surprise waiting
‘A surprise is waiting for Ali at home’

b. *bir sürpriz Ali-yi ev-de bekliyor
one surprise Ali-ACC home-LOC waiting

20) a. Yol-u bir araba tıkamıs
road-ACC one car blocked
‘A car has blocked the road’

b. bir araba yol-u tıkamıs
one car road-ACC blocked
‘A car has blocked the road’

Let’s take stock:
i. Whereas –human non-specific subjects must remain in situ, +human non-
specific subjects can (in fact, it seems must) raise from their base positions.
ii. Whereas non-specific –human direct objects cannot have case (or a plural
morpheme), +human direct objects must bear overt case.
iii. A non-human object can never precede a +human subject.
iv. A non-human non-specific subject cannot precede a +human specific object.


138
I will shortly demonstrate that a non-specific human nominal is not possible in Turkish, and will
argue that a partitive construction is used to denote non-specificity for humans.
139
I am not sure what position bir araba ‘a car’ occupies in (20)b. I assume that in (20)a, it is in its
base position. Although I have not said anything throughout about Agr projections, it may be the
AgrO projection in the verbal domain that causes the difference. As we will see later in this chapter,
+human expressions have ç-features, or rather are ç-complete. Perhaps, it does not matter whether it
is the object or the subject that checks the Agr features in (20); however, in (19) the +human object
must be the highest expression in the verbal domain so that its ç-features can be checked by AgrO.
Being marked with structural accusative case, the human direct object cannot have its ç-features
checked by AgrS because it is barred from moving to T. This point is purely speculative, though.

156

These facts combined lead us to conclude that +human arguments must be DPs.
Recall that I had proposed that only DPs can raise and receive Case. Extending this
assumption to the +human facts, we must conclude that +human nominals can never
be NPs, they must be DPs. Thus, just like DPs +human arguments must bear case,
(ii) above, and just like DPs, they must raise to case assigning positions, (iii) and (iv)
above. The facts in (i) also follow if we assume that +human subjects are DPs: they
must raise from their base positions, and be case-marked. Before I explain the nature
of the +human DP, let’s first look at more evidence and the consequences of such an
assumption, including the issue of specificity.
The assumption that +human arguments are obligatorily DPs is supported by
evidence from Quantifiers and Wh-phrases. According to Kural (1992) +human
direct object Quantifiers (QPs) and Wh-phrases must bear accusative case.
140
Let’s
look at the contrast between who and what, and someone and something, in the
examples below.
As shown in (21) and (22), the +human Wh-expression who must always bear
accusative case when it is a direct object, no such requirement holds for what.

21) a. pro kim-i unut-tu-n?
pro who-ACC forget-PST-2s
‘Who did you forget?’

b. *pro kim unut-tu-n?
pro who forget-PST-2s




140
from Kural (p.c.)

157

22) a. pro ney-i unut-tu-n?
pro what-ACC forget-PST-2s
‘What (specific thing) did you forget?’

b. pro ne unut-tu-n?
pro what forget-PST-2s
‘What did you forget?’


Nominative case is null, but we can assume the same holds for who when it is a
subject, i.e. it raises to T and bears the Ø nominative morpheme. This can be verified
by embedding the question in a complement clause. In (23), the +human wh-subject
kim ‘who’ must bear overt case. No such requirement exists for the non-human
wh-subject ne ‘what’ in (24)
141
.

23) a. *pro [kim gel-dig-i]-ni gör-dün?
pro who arrive-NSR-3s-ACC saw-PST-2s
‘Who did you see arrive?’

b. pro [kim-in gel-dig-i]-ni gör-dün?
pro who-GEN arrive-NSR-3s-ACC saw-PST-2s
‘Who did you see arrive?’



141
The example in (23)b is an interrogative even though the matrix verb does not bear the Question
morpheme –mI. This contrasts with (24), which can only be a question if the matrix verb has a
Q-marker, as in (i).
(i) a. pro [ne kırıl-dıg-ı]-nı biliyor-mu?
pro what broke-NSR-3s-ACC know-Q-3s
Does he know what broke?’
b. pro [ne-yin kırıl-dıg-ı]-nı biliyor-mu?
pro what-GEN broke-NSR-3s-ACC know-Q-3s
‘Does he know what (specific thing) broke’
When a Q-morpheme is added to (26b), the interpretation becomes ‘Did you see who arrived?’ I will
not speculate here as to what may motivate this difference. In the ECM example in (ii) with the verb
san ‘believe/think’, the bare what takes matrix scope, making (ii) an interrogative in contrast to (27).
(ii) Ne kır-ıl-dıg-ı-nı san-ıyor?
what break-NSR-3s-ACC believe-3s
‘What does he think/believe broke?’

158

24) a. pro [ne kırıl-dıg-ı]-nı biliyor
pro what broke-NSR-3s-ACC know-3s
‘He knows what broke’

b. pro [ne-yin kırıl-dıg-ı]-nı biliyor
pro what-GEN broke-NSR-3s-ACC know-3s
‘He knows what (specific thing) broke’


The same requirement holds for Quantifiers. In (25), we see that the +human QP
someone must also obligatorily bear case, whereas something need not (26). And, as
expected the QP someone must bear genitive case when it is the embedded subject
(27)a whereas no such requirement exists for something (27)b.

25) a. pro biri-ni unut-tu-n.
pro someone-ACC forget-PST-2s
‘You forgot someone’

b. *pro biri unut-tu-n.
pro someone forget-PST-2s

26) a. pro birsey-i unut-tu-n.
pro something-ACC forget-PST-2s
‘You forgot some (specific) thing’

b. pro birsey unut-tu-n.
pro something forget-PST-2s
‘You forgot something’

27) a. pro [biri*(-nin) gel-dig-i]-ni biliyor
pro someone-GEN come-NSR-3s-ACC know-3s
‘He knows someone came/is coming’

b. pro [birsey(-in) kırıl-dıg-ı]-nı biliyor
pro something(-GEN) broke-NSR-3s-ACC know-3s
‘He knows some (specific) thing broke’


159

The QP and Wh-facts support our earlier conclusion that +human arguments must
always bear case.
142
We saw that Case in Turkish is assigned in a Spec-head
configuration. Because +human arguments are always case-marked, we can conclude
that they must obligatorily have raised to a case-assigning head.
We saw that the SR relative clause is licensed for non-subject extraction when
the subject remains low in the structure, leaving [Spec, TP] vacant for the Wh-
expression. Repeating the tree in (7) as (28), note that we had assumed the non-
specific subject cat was an NP that must remain in-situ. If the subject were a DP, it
would have to raise to T and get case. When [Spec, TP] is occupied by a non-Wh-
element, the NSR is triggered, as shown in (29).



142
Similar facts can be found cross-linguistically. For example, in Persian (an SOV Indo-European
language), the direct object case morpheme –ra is obligatory on Wh-words and Quantifiers denoting
humans (ii) but is not permitted for non-human nominals unless they are D-linked and specific (i).
(i) a. che /yek chizi did-im b. che-ra /yek chizi-ra did-im?
what/one thing see-1p what-ACC/one thing-ACC see-1p
‘What did we see?’ ‘What (specific) thing did we see?’
‘We saw something.’ ‘We saw some (specific) thing.’
(ii) a. *ki /*kesi did-im b. ki-ra /kesi-ra did-im?
who/someone see-1p who-ACC/someone-ACC see-1p
‘Who did we see?’ ‘We saw someone.’
In Hindi (P. Chandra, p.c.) ‘something’ as a direct object can optionally be marked with the specificity
marker –ko, as in (i). On the other hand, a direct object ‘someone’ without the –ko marker is
unacceptable, (ii).
(i) a. her subeh mc kuch cizc bhul jati hu
every morning I somethings forget PROG be
‘every morning I forget something’
b. her subeh mc kuch cizc bhul jati hu
every morning I somethings forget PROG be
‘every morning I forget some (specific) thing’
(ii) a. *her partime mc kIsI bhul jati hu
every party I someone forget PROG be
b. her partime mc kIsI-ko bhul jati hu
every party I someone-KO forget PROG be
‘every party I forget someone’
In Afrikaan, direct objects raise when they are specific. A non-human universal quantifier direct
object can raise or remain in situ, but a human direct object must obligatorily raise (T. Biberauer, p.c.).

160

28) a. [ [Ø
1
kol-u]-nu kedi tırmala-yan] çocuk
1

arm-POSS-ACC cat claw-SR child
‘the child whose arm a cat clawed’

b. CP


TP C°

+Wh-child
vP T°

DP+ACC
NP-cat
VP v°

DP V°
claw
+Wh-child
+GEN arm D°




29) a. [ [Ø
1
kol-u]-nu kedi-nin tırmala-dı-ı] çocuk
1

arm-POSS-ACC cat-GEN claw-NSR child
‘the child whose arm the cat clawed’

b. CP


TP C°

DP-cat +GEN
vP T°

DP+ACC
t- DP-cat
VP v°

DP V°
claw
+Wh-child
+GEN arm D°





Thus, the SR form is not possible for non-subjects when the clausal subject is a DP.
By assuming that +human nominals are obligatorily DPs, we can account for the fact

161

that the SR form is banned when the subject is human: [Spec, TP] will always be
occupied by the +human DP-subject. Consequently, the +Wh relativized element will
not be able to move to [Spec, TP], and the SR form will be barred.


4 Explaining the Behavior of Human Nominals


We have determined that +human nominals in Turkish have the hallmarks of DPs.
They must be case-marked and they must raise to functional projections. I suggest
that +human nominals in Turkish must always first merge with a null pronoun which
gives the nominal its ç-features.
143
Adopting Postal (1966), I assume that D˚ and
pronouns are essentially the same thing, a bundle of features, specifically ç-features.
Adopting a version of Longobardi (1994), I propose that +human nominals in Turkish
have the structure as in (30).

143
Interestingly, number agreement on the verb in Turkish is permitted only when the subject is
+human. Plural agreement with an overt non-human subject usually denotes a plurality of events, as in
(iiib) and (ivb).
i. a. yolcu-lar geldi b. yolcu-lar geldi-ler
traveler-pl came traveller-pl came-pl
‘the travelers came’ ‘the travelers came’
ii. a. kus-lar ötüyor b. *kus-lar ötüyor-lar
bird-pl chirping bird-pl chirping-pl
‘the/some birds are chirping’
iii. a. köpek-ler hırladı b. köpek-ler hırladı-lar
dog-pl growled dog-pl growled -pl
‘the/some dogs growled’ ‘the/some dogs growled (on different occasions)’
iv. a. otobüs-ler geldi b. otobüs-ler geldi-ler
bus-pl came bus-pl came-pl
‘the buses came (at one time)’ ‘the buses came (series of comings/*all at one time)’
Thus, it may be that contrary to what is believed, Turkish actually has a “poor” agreement system and
that the verbal agreement that does exist is triggered by the presence of (or perhaps required for
interpretability by) ç-features which appear only on +human nominals. Kornfilt (2005) suggests that
in Turkish “agreement has pronominal features”. It seems reasonable to assume that ç-features must
obligatorily enter into an Agree relation, and that the null 3
rd
person singular verbal agreement is in
fact no agreement at all for non-human arguments, and is default or defective agreement for +human
arguments.


162

30) DP


NP+human D° (= Ø-pronoun=-features)
athlete


In Chapter 3 on Specificity, we looked at some of Longobardi’s arguments. Let’s
revisit some of them here. First, we saw that Italian sentences require a D in the
preverbal subject position, as in (31) and (32).

31) a. *Acqua viene giù dalle colline.
water comes down from.the hills

b. Viene giù acqua dalle colline.
comes down water from.the hills

32) *(Un/Il) grande amico di Maria mi ha telefonato.
(a/the) great friend of Maria called me up

However, an article is optional when the subject is a proper name, as in (33). An
exception is (34), when the proper name is the last name of a female, the article is
obligatory.

33) a. Gianni mi ha telefonato.
Gianni called me up

b. Il Gianni mi ha telefonato.
the Gianni called me up

34) La Callas/*Callas ha cantato.
the Callas/Callas sang


163

Longobardi assumes that in sentences such as (33)a) above, with a determinerless
proper name as subject, there is in fact a null D head, and that there has been N to D
movement. Thus, (33)a) is not an exception to the requirement of a DP as subject.
(33)a) and (33)b) are almost identical except that in (33)a), there has been head
movement of N to a null D. The structural difference in the subjects is shown in (10).

35) a. Gianni ... b. Il Gianni ...
DP DP


D° NP D° NP
Ø Il
N° N°
Gianni Gianni


It becomes understandable why an overt determiner is obligatory for last names of
females. Whereas a name like Marie carries a gender feature, the feminine gender
feature is not carried on the last name, which is neutral. I assume that this gender
feature is borne by D, with the result that there is too much ç-feature information on
the D for it to be null (or vacuous).
For Turkish, I had suggested a strong hypothesis: that only DPs can be
arguments, and that by extension NPs can only be nominal predicates
144
. I had
adopted Mandelbaum’s (1994) proposal that predicate NPs are basically adjectival.
Translating this idea into an “event-ish” semantic interpretation, a sentence with an
NP subject like cat in (36)a), would be an event of ‘cat-scratching’ which would have

144
Higginbotham (1987) proposed that an argument is “saturated” and can thus be assigned a theta
role. By extension Szabolcsi (1987), Abney (1987) and Longobardi (1994) have argued that NPs are
nominal predicates (unsaturated) and do not bear a theta-role and DPs are arguments that do bear a
theta role. Stowell (1989) has shown that NPs are non-referential, whereas DPs are referential.

164

his arm as the Theme, as shown in (36)b)
145
. This differs from the sentence in (37)a)
with a DP subject in that we now have an external argument cat. Thus, the semantic
interpretation for (37)b) would be there is an event of ‘scratching’ which has cat as
the Agent and his arm as the Theme.

36) a. [pro kol-u]-nu kedi tırmala-dı
arm-POSS-ACC cat scratch-PST
‘A cat (i.e. some cat or other) scratched his arm’

b. ∃e[∃x:Cat (x) [Agent (e, x) & Scratching (e)] & Theme (e, his arm)]

37) a. kedi [ pro kol-u]-nu tırmala-dı
cat arm-POSS-ACC scratch-PST
‘The cat scratched his arm’

b. ∃e [Agent (e, the cat) & Scratching (e) & Theme (e, his arm)]

The question remains as to why a +human subject is not allowed by the grammar to
be predicative or adjectival as in (36)a. Does the restriction reside in the syntax or the
semantic component? I suggest that it is the syntax that drives this constraint, or
more aptly, it is the lexicon. My proposal is that in Turkish (and perhaps in other
languages as well
146
), nominals in the lexicon marked as having +human features

145
I point out again that I reject the idea that the non-specific subject incorporates into the verb. See
Chapter 3 and reasons cited there.
146
Freeze (1992) includes interesting cross-linguistics data for existentials. He also includes the
following data from English in regards to the interaction between +human subjects and locations and
the verb ‘have’. When the subject of ‘have’ is non-human, the theme must be inalienably possessed
(or characteristically treated as such), as in (i). No such restriction holds for human subjects, (ii).
(i) a. The tree has branches.
b. The flour has weevils.
c. The tree has a nest *(in it.) = There is a nest in the tree.
d. The flour has a ring *(in it). = There is a ring in the flour.
(ii) a. The boy has a needle.
b. The boy has a cousin/nose.
c. The boy has fleas.

165

have a selectional requirement for ç-features. Pronouns are inherently human
because they contain, in fact, are ç-features.
147
As for non-pronominals, I assume
that a +human nominal must first merge with ç-features in the lexicon before it can
enter a derivation.
148
The +human element that enters the derivation, then, is a
composite of human+ç-features which is identical to human+(null)pronoun, which if
we take Postal’s work seriously is equivalent to human+D. This human+D
expression, as with all DPs, is prevented (either by the syntax or the LF interface)
from being predicative/adjectival, and so must first-merge into a theta position.
149


Note that whereas a non-human location permits existential predication with a theme, the human
location prefers to be the subject of a ‘have’ construction, (iii). In fact, as shown in the
existential/’have’ constructions in (v) and (vi), human “locations” require the ‘have’ construction in
contrast to non-human locations which can only appear in existential or copular constructions.
(iii) a. There is a nest in the tree.
b. There is a ring in the flour.
c. There is a needle on the boy. ( = The boy has a needle.)
(iv) a. *A book is with/at/by Lupe.
b. *There is a book with/at/by Lupe.
c. Lupe has a book.
(v) a. A mongoose is on the shelf.
b. There is a mongoose on the shelf.
c. *The shelf has a mongoose.
Although discussing the nature of human/non-human nominals in English is beyond the scope of this
thesis, the data here demonstrates that +human locations are not acceptable in existential “there”
constructions, and that they must raise to [Spec, TP] of the verb ‘have’. As for possession, if we
assume a thematic heirarchy with inalienable possession being the lowest, one could argue that the data
on possession in (i) and (ii) suggests that human possessors can raise within this heirarchy, but that
non-human possessors can not.
147
Note that the 3
rd
person singular pronoun in English it cannot refer to humans: *It came. I assume
this is so because this pronoun lacks ç-features, which is consistent with what I am proposing for
Turkish. Note that this excludes the 3
rd
person plural pronoun, they. It is noteworthy that number
features seem to add structure to a nominal, for example a NumP, that has consequences in Turkish.
Thus, in the good SR examples where the RC subject is an NP, changing the subject to a plural leads to
degrading or unacceptability. Thus, absence of ç-features means absence of person and number
features.
148
Of course, it is also possible that a +human nominal must merge with a null-pronoun, i.e. ç-
features, in overt syntax forming a DP which then merges into a theta position. I am not committed as
to the exact nature of the selectional requirement and how it is satisfied; it is only that such a
requirement exists that is germane here.
149
Higginbotham (1987) argues that the copula in “John is the director” is different from the copula in
“John is a director” in that the former expresses identity and is referential whereas the latter is
predicative. In addition, he argues that in English definite descriptions can be predicative as the
embedding of [John (is) the man] without an overt copula in (ia) demonstrates.
(i) a. I consider [John [the man (for the job)]]

166

Being a DP, it can satisfy the EPP, and it must receive case. As we will shortly see, a
DP must also enter into an Agree relation.
Before we return to Turkish, I should point out that the nature of D° must be
parametric cross-linguistically. A null D° in Italian may have default 3
rd
person,
singular (perhaps masculine) ç-features, which would explain why a proper name
may raise to D° in (10)a. It would also explain why an overt determiner is required
for female last names in Italian as we saw in (34); the feminine feature on the proper
name (assuming it exists only on last names of females) would create a feature
mismatch on D°.
Returning to the Turkish examples in (12) repeated as (38), we are now in a
position to account for the unacceptable SR form in (38)b. Recall that the SR form is
licensed when a +Wh-expression moves to [Spec, TP]. For a non-subject to move to
[Spec, TP], the subject had to be an NP so that it would not need case. In (38)b, the
non-Wh-subject oför ‘driver’ is +human. This means that it must have first merged
with a null-pronoun (or D°). As a DP, it must satisfy the Case Filter, and so must
move to [Spec, TP] in order to receive case. The unacceptability of (38)b is due to a
Case Filter violation. In (38)c, the +human subject has raised to [Spec, TP] and
received case. I assume the accusative object [
DP
[
DP
peasant’s] cow] has scrambled
over the subject, after which the possessor [
DP
peasant] raises from the specifier of the
DP to [Spec, CP]. A non-human subject, on the other hand, may be either an NP, as

b. *I consider [John [the man (standing over there)]]
I would argue that (ia) actually contains a small clause, as in “the man (who is best) for the job”. Thus
I will stick to the strong hypothesis that a DP cannot be a predicate.


167

in (38)a, leaving [Spec, TP] vacant for the Wh-expression, or a DP, as in (38)d, in
which case it must raise to [Spec, TP] to be assigned case.

38) a. [[Ø
1
ineg-i]-ni tren ez-en] köylü
1

cow-POSS-ACC train run.over-SR peasant
‘the peasant whose cow a train ran over’

b. *[[Ø
1
ineg-i]-ni soför ez-en] köylü
1

cow-POSS-ACC driver run.over-SR peasant
Intended: ‘the peasant whose cow a driver ran over’

c. [[Ø
1
ineg-i]-ni soför-ün ez-di-i] köylü
1

cow-POSS-ACC driver-GEN run.over-NSR-3S peasant
‘the peasant whose cow the/*a driver ran over’

d. [[Ø
1
ineg-i]-ni tren-in ez-di-i] köylü
1

cow-POSS-ACC train-GEN run.over-NSR-3S peasant
‘the peasant whose cow the train ran over’

We saw in (25) and (26), repeated as (39)and (40) respectively, that there is a contrast
with respect to case-marking and specificity based on whether an argument is
+human. A case-marked non-human argument, as in the accusative ‘something’
birey-i in (40)a, must receive a specific interpretation (a specific “thing”). This is
expected from what we know about overt case and specificity. However, the bare
someone as a direct object is unacceptable in (39)b, and the case-marked someone
does not denote a specific person. How is it that a case-marked expression can yield
a non-specific interpretation, in this case, a non-specific human?

39) a. pro biri-ni unut-tu-n.
pro someone-ACC forget-PST-2s
‘You forgot someone’

b. *pro biri unut-tu-n.
pro someone forget-PST-2s


168

40) a. pro birsey-i unut-tu-n.
pro something-ACC forget-PST-2s
‘You forgot some (specific) thing’ or ‘You forgot a/one thing’

b. pro birsey unut-tu-n.
pro something forget-PST-2s
‘You forgot something’


I suggest non-specificity for human nominals is achieved in Turkish by means of a
partitive construction. Support for this proposal comes from the fact that the word
someone in Turkish is in fact bi-morphemic: one plus possessive agreement, as shown
in (41)a. I assume that the structure of bir-i is actually “of them, one” as in (41)b,
with a null pro denoting the of them in the specifier of D°. This null pro triggers the
possessive agreement on bir ‘one’. This DP-NP structure is analogous to an
existential IP with the NP-one in the restrictor and the pro defining the scope.
Because D° is a case-assigner, the pro in the Spec of the DP is assigned case. The NP
does not require case, but the entire DP must be case-marked. The partitive itself is
specific, i.e. a DP, but the expression in its restrictor can be non-specific.

41) a. Bir-i gel-di.
one-3s come-PST
‘Someone came’

b. DP

pro
NP D° = Ø-pronoun
one+agr +human


169

There are two partitive constructions in Turkish: one with genitive case on the
superset nominal, and the other with ablative case, as in (42)a and (42)b
respectively
150
.

42) a. Ali kadın-lar-ın iki-si-ni tanıyordu.
Ali woman-pl-GEN two-AGR-ACC knew
‘Ali knew two of the women’

b. Ali kadın-lar-dan iki-si-ni tanıyordu.
Ali woman-pl-ABL two-AGR-ACC knew
‘Ali knew two of the women’


Let us look closer at the derivation of these partitives. Beginning first with (42)b, I
assume that the structures begin by merging the superset DP-women with the subset
NP two X’s, where X is of the same type as the superset, as in (43)a where the index
merely denotes “of the same type”. In (42)b the DP-women is a PP-like expression
“from/of the women” and as with most Turkish PPs is inherently case-marked, in this
case with ablative case. As shown in (43)b the merge of the PP/DP and the NP
projects to an NP which merges with a D°. D° has an EPP feature and attracts the
ablative PP/DP-women to its specifier triggering agreement on the NP
151
, as in (43)c.
Note that I refer to the phrase “of the women” as a PP/DP-women. This is because
the structure of this expression is actually a DP with ablative case; there is no PP
projection above the DP. In this structure, there is no issue of minimality in the
movement of the PP/DP to [Spec, DP]. In an English PP, the DP would be raising

150
Examples from Enç (1991).
151
I remain agnostic as to where and how agreement is actually triggered. There may be an Agr
projection (both in the nominal and verbal domains), but I have ignored this completely as it would
take me too far afield. What is relevant, is that an agreement morpheme shows up on the
phonologically contentful N°. Kornfilt (2005) offers an account of Agr in these constructions, but the
analysis is at right angles with what I am proposing here.

170

from the complement position of P°, and thus be c-commanded by the NP two.
However, the Turkish PP has the structure of a DP, i.e. it is not embedded in a PP
projection. This is shown in the tree in (44) for the derivation in (43).

43) a. [
PP/DP
women
1
]-ABL + [
NP
two X
1
]

b. [
NP
[
PP/DP
women
1
-ABL] + [
NP
two X
1
]] + D°

c. [
DP
[
PP/DP
women
1
-ABL] [
NP
[
PP/DP
women
1
-ABL] + [
NP
two X
1
]] + D°+AGR]



44) DP

women-ABL
NP D°= Ø-pronoun
+human
DP NP
women-ABL two+agr



The derivation of the partitive in (42)a is almost identical to (43)-(44) except that the
PP/DP women does not have inherent case and needs to be assigned structural case.
Looking at the derivation for (50)a in (45), we see that a possessive PP/DP (of)
women merges with the NP two X’s in (45)a. This structure then merges with D°, and
the PP/DP (of) women raises to [Spec, DP], triggering agreement. In this derivation,
D° assigns genitive case to the DP in its Spec.
152


45) a. [
PP/DP
(of) women
1
] + [
NP
two X
1
]

b. [
NP
[
PP/DP
(of) women
1
] [
NP
two X
1
]] + D°



152
This is basically a possessor construction. It may be that there is a null of in the PP/DP (of) women.
A possibility that I find appealing is that the specifier of NP may be a possessor theta position and that
an element merged here picks up a possessor theta role.

171

c. [
DP
[
PP/DP
(of) women
1
-GEN] [
NP
[
PP/DP
(of) women
1
] [
NP
two X
1
]] D°+AGR]



Let’s now return to the sentence in (39)a repeated as (46). Note that the structure of
the direct object is actually polymorphemic, consisting of one plus possessive
Agreement plus the accusative morpheme. I propose that this structure is identical to
the partitives we saw above, specifically the derivation in (45). The derivation in (47)
shows the structure of bir-i-ni ‘someone’. The superset denoted by (of) them+human,
is actually a null pronoun, or more specifically a D° with ç-features, as in (47)d. The
NP one X picks up its identity from whatever the identity of the superset is, but it does
not itself need to bear the same features. Thus one X can remain an NP. In this way,
we can explain the non-specific interpretation of someone+case; the larger partitive
construction is a DP which gets case, but the NP bir is non-specific. Thus, in (46),
the direct object is a partitive with the non-specific (NP) lexical item one embedded
in a DP shell that serves as its scope. Non-specificity obtains because someone, i.e.
[
NP
one X] remains in the restrictor. Of course, the partitive, being a DP, requires
case; (39)b repeated as (48), is therefore a Case Filter violation.

46) pro bir-i-ni unut-tu-n.
pro one-3s-ACC forget-PST-2s
‘You forgot someone’

47) a. [
PP/DP
(of) them+human
1
] + [
NP
one X
1
]

b. [
NP
[
PP/DP
(of) them+human
1
] [
NP
one X
1
]] + D°

c. [
DP
[
PP/DP
(of) them+human
1
-GEN] [
NP
[ t ] [
NP
one X
1
]] D°+AGR]


d. [
DP
[
PP/DP
D°=ç=null pronoun X
1
+human-GEN] [
NP
[ t ] [
NP
one X
1
]] D°+AGR]

172

48) *pro bir-i unut-tu-n.
pro someone-3s forget-PST-2s

To be sure we are on the right track, let’s look closer at the Turkish “one” bir.
Perlmutter (1969) demonstrates that the indefinite article in English behaves like the
numeral one. Yükseker (2003) argues that bir ‘one’ in fact behaves differently from
the other numerals. First, as demonstrated in (49), numerals (other than one) can
alternate in word order with adjectives with no semantic reflex. On the other hand,
bir denotes numericity when separated from the nominal it modifies, as in (50).

49) a. iyi yeni iki kitap
good new two book
‘two good new books’

b. iki iyi yeni kitap
two good new book
‘two good new books’

c. iyi iki yeni kitap
good two new book
‘two good new books’

50) a. iyi yeni bir kitap
good new one book
‘a/*one good new book’

b. bir iyi yeni kitap
one good new book
‘*a/one good new book’

c. iyi bir yeni kitap
good one new book
‘*a/one good new book’


173

Second, numerals can co-occur with a demonstrative as long as the nominal is case-
marked, as in (51)
153
, whereas bir is incompatible with a demonstrative as shown by
(52). In fact, bir seems to be incompatible with specificity, as shown by the
unacceptability of the accusative case on (53)b. Example (53)c demonstrates an
exception where bir is focused and pronounced with stress and can appear with an
overtly accusative, i.e. specific, object.
154


51) a. Patricia bu kitab-ı oku-du
Patricia this book-ACC read-PST

b. *... bu kitap ...
this book

c. ... iki kitap ...
two book

d. *... iki kitab-ı
155
...
two book-ACC

e. ... bu iki kitab-ı ...
this two book-ACC

52) a. *Patricia bu bir kitab-ı oku-du
Patricia this one book-ACC read-pst

b. *... bu bir kitap ...
this one book

53) a. Patricia bir kitap oku-du
Patricia one book read-PST
‘Patricia read a book’

b. *... bir kitab-ı ...
one book-ACC

153
Numerals seem to be [–specific] because they lead to unacceptability when modifying a case-
marked nominal without an overt demonstrative element, as in example (51)d.
154
This issue of focus will be addressed later as part of a larger discussion.
155
This example is acceptable when in response to a D-linked set, but this is orthogonal at this point.

174

c. Exception: Patricia her gün BIR gazetey-i oku-r
Patricia every day ONE newspaper-ACC read-AOR
‘Patricia reads one particular newspaper every day’

Finally, whereas numerals exhibit adjectival properties as part of a larger DP, (54)a,
bir ‘one’ is predicative; it does whatever the indefinite article “a” or the indefinite
quantifier “some” do in English.

54) a. [
DP
bu iki yeni kitap]
this two new book

b. [
DP
bu(*-nlar) iki yeni kitap(*-lar)]
this(*-pl) two new book(*-pl)

c. *[
DP
bu bir yeni kitap]
this one new book

d. *[
TP
[
DP
bu] [
VP
iki yeni kitap] ]
this two new book

e. [
TP
[
DP
bun-lar] [
VP
iki yeni kitap] ]
this-pl two new book
‘These are two new books’

f. [
TP
[
DP
bu] [
VP
bir yeni kitap] ]
this one new book
‘This is one new book’

g. [
TP
[
DP
bu] [
VP
yeni bir kitap] ]
this new one book
‘This is a new book’

For expository purposes I am going to refer to bir denoting non-specificity (translated
as the indefinite article in English in the glosses above) in lower case letters and the
numeral BIR ‘one’ in upper case letters. I conclude from the data that once bir raises
from a position as sister to N°, it is interpreted as the numeral BIR. To be more

175

precise, I assume that bir, in fact, cannot raise at all; it is the head of a phrase, OneP,
which is syntactically an NP, and thus non-specific.
156
This contrasts with the
numeral BIR that adjoins to NP, and is adjectival. The difference is shown in the
diagrams in (55).

55) a. bir ‘a new book’ b. the numeral BIR ‘one new book’

OneP NP

new BIR NP
One° N°
bir book new

book

Pursuing our analysis of partitives above, we might say that (55)a is analogous to
[one X
1
] [book
1
] whereas (55)b is simply [one [new [book]]].
I also assume that One° is the antithesis of a pronoun. A pronoun is a D° and
is composed of ç-features, and is the locus of specificity. D-features raise in the
syntax. One° is devoid of features, and serves as a “drag” on nouns: by this I mean
that 1) it is an NP and cannot satisfy the EPP (OneP and its complement N cannot
raise), and 2) it induces minimality effects and blocks raising from its c-commanding
domain. In this way, One° causes syntactic (non-)specificity effects which will be
interpreted in the semantic component.
Repeating (41) as (56)a, let’s use our analysis of the possessive partitive we
saw in (47), to build the structure of the DP subject with the non-specific

156
I adopt the view that phrasal labels are for expository/mnemonic purposes only and that the syntax
only cares about categories as identified by their features or properties. Thus, as far as the syntax is
concerned OneP is an NP.

176

interpretation ‘someone’. First, note that the subject biri is indeed a DP with null
nominative case because when we embed the sentence, the subject must be
obligatorily case-marked, as in (56)b.

56) a. Bir-i-Ø gel-di.
one-3s-NOM come-PST
‘Someone came’

b. Susan [bir-i*(-nin) gel-dig-i]-ni duy-mus.
Susan one-3s-GEN come-COMP-3s-ACC hear-EVID
‘Susan heard that someone had come’


In (56)a, I assume the subject bir-i is a partitive structure where the superset is a pro
with the meaning akin to ‘of the group’. As we saw, partitives can be denoted in
Turkish by either a nominal expression with ablative case, onlar-dan, as in (57), or a
possessive with genitive case, onlar-ın, as in (58).

57) on-lar-dan bir-i (dir)
3-pl-ABL one-3s (copula)
‘It/He is one of them’

58) on-lar-ın bir-i (dir)
3-pl-GEN one-3s (copula)
‘It/He is one of them’

I assume that D° is a functional head that assigns case and also has an EPP feature.
As with other case assigning heads a non-case-marked DP may move to its specifier,
to be assigned case, or an inherently case-marked expression may move there to
satisfy the EPP. This is shown in (59)a for (57) where the ablative moves to [Spec,
DP] to check the EPP feature of the (null pronoun) D°, and (59)b for (58) where the

177

3
rd
person plural pronoun them without case raises to [Spec, DP] and receives genitive
case from D°. In both instances, the movement triggers agreement on bir.

59) a. DP

them+ABL
OneP D° = Ø-pronoun

DP One°
them+ABL bir




b. DP

them+GEN
OneP D° = Ø-pronoun

DP One°
them bir


In both these structures, the effect is such that the superset group identifies the scope
and the OneP delineates the restrictor. Because bir is in the restrictor, it receives a
non-specific reading.
In (56)a, with subject bir-i, the superset DP “(of) them” remains
unpronounced, a reflection of the “Avoid Pronoun Principle” which operates quite
systematically in Turkish. The superset in (56)a could also be pronounced as in (60),
but would entail a D-linked group.

60) On-lar-dan/ın bir-i-Ø gel-di.
3-pl-ABL/GEN one-3s-NOM come-PST
‘One of them came’


178

Thus, I assume the superset “(of) them” of (56)a is pro. Furthermore, I will assume
the possessive structure for pro + bir because there is evidence that this partitive
structure requires the subset to be of the same type.
157
We had originally said that the
expression denoting the superset could be first-merged as the complement of the
subset (in this case One°) or as the Spec of One°.
158
I revise this view somewhat and
suggest that in the ablative construction, the superset ablative adjoins to the subset-
NP, and in the possessor partitive, the superset first merges as the specifier of the
subset-NP. Thus, as shown in (61), the pro denoting the superset “of them” merges
into [Spec, OneP], after which OneP merges with D°, and “of them”-pro raises to
[Spec, DP] to be assigned genitive case, triggering possessive agreement on One.

61) pro(on-lar-ın) bir-i

DP

pro-GEN
OneP D° = Ø-pronoun

pro(3pl)
( =“(of) them”) One° +Agr
bir -i



157
The genitive possessor requires agreement and the superset-subset relationship must be one that
allows normal “possessivity”. As shown in (i), the ablative can denote a group to which one belongs,
but (ii) indicates that the genitive possessor induces familial identity. Because [bir X] must pick its
identity from the pro superset, I assume the partitive construction has the genitive possessor as the
superset.
(i) çocuk-lar-dan iki kız
child-pl-ABL two girl
‘two girls of (the group of) children’
(ii) çocuk-lar-ın iki kız*(-ı)
child-pl-GEN two girl-3s
‘the children’s two girls (as in daughters)’
158
See fn. 156.

179

The result is that by being in the restrictor of this construction, bir remains non-
specific, while its superset possessor defines the scope. At the same time, the
requirement that +human nominals be DPs is satisfied.


5 Summary


We began with a problem: in Turkish, +human subjects triggered unacceptability in
relative clauses that were otherwise acceptable. Our analysis showed that +human
nominals could not be NPs. We saw evidence from other data that supported this
conclusion. We then looked at the non-specific +human QP, biri, ‘someone’, and
saw that the grammar required it to be obligatorily case-marked. We had determined
that overt case denotes specificity in Turkish, so how was the non-specific reading
obtained? The polymorphemic structure of bir-i indicated that it was part of a
partitive structure. By proposing an account where the bir remains in the restrictor of
a complex DP, the contradiction between overt case and non-specificity was resolved.
In addition to the theoretical implications of human-non-human distinctions in
specific grammars and in Universal Grammar, in general, on a pragmatic level, this
study serves to identify another variable that needs to be considered when evaluating
grammaticality. For example, Kennelly (1997) states that “Time expressions do not
function as a ‘Locative’ argument in terms of relativizing using the [SR] strategy.
*Adam gel-en gün. ‘The day when a man came.’” (p.64) Based on what we saw in
this chapter, the ungrammaticality of this example is predictable because the subject
is +human and must move to [Spec, TP] for case. However, examples with non-

180

human subjects are perfectly grammatical with the same time expression as the
relative head, as shown in (62).

62) a. [yagmur/kar yag-an] günler
rain/snow rain-SR days
‘the days it rained/snowed

b. [bomba patlay-an] gün
bomb explode day
‘the day a bomb exploded’

c. [çöp al-ın-an] gün
trash take-PASS-SR day
‘the day the trash is taken’



6 Contrastive Focus and Human Subjects


We saw that extraction from possessive accusative objects using the SR form is
unacceptable when the RC subject is human. Compare (63) with (64). Of course, the
NSR form is acceptable for both these examples with the condition that the subject be
specific, as shown in the examples in (65), where the clausal subjects, tren ‘train’ and
öför ‘driver’ must be overtly case-marked with genitive case and must receive a
specific interpretation.
159



159
As we have seen in other cases, the scrambled position above the subject in [Spec, TP] is also
available for the accusative object. Thus the examples in (i) are also acceptable, but note that the case-
marking and interpretation (modulo the discourse effect of the scrambling) is identical.
i. a. [[Ø ineg-i]-ni tren-in ez-dig-i] köylu
cow-AGR-ACC train-GEN run.over-NSR peasant
‘the peasant whose cow *a/the train ran over’
b. [[Ø ineg-i]-ni söför-ün ez-dig-i] köylu
cow-AGR-ACC driver-GEN run.over-NSR peasant
‘the peasant whose cow *a/the driver ran over’

181

63) [[Ø ineg-i]-ni tren ez-en] köylu
cow-AGR-ACC train run.over-SR peasant
‘the peasant whose cow a train ran over’

64) *[[Ø ineg-i]-ni söför ez-en] köylu
cow-AGR-ACC driver run.over-SR peasant
‘the peasant whose cow a driver ran over’

65) a. [tren-in [Ø ineg-i]-ni ez-dig-i] köylu
train-GEN cow-AGR-ACC run.over-NSR peasant
‘the peasant whose cow *a/the train ran over’

b. [söför-ün [Ø ineg-i]-ni ez-dig-i] köylu
driver-GEN cow-AGR-ACC run.over-NSR peasant
‘the peasant whose cow *a/the driver ran over’

But note that the unacceptable (64) becomes acceptable when the +human subject is
contrastively focused, as in (66).
160


66) [[Ø ineg-i]-ni BU öför ez-en] köylu
cow-AGR-ACC THIS DRIVER run.over-SR peasant
‘the peasant whose cow THIS DRIVER (rather than that one) ran over’

In the previous section, we determined that a +human nominal must be a DP and
must therefore always raise and be overtly case-marked. A +human subject cannot
remain in-situ. But this seems to be exactly what is happening in example (66), as
well as example (67) with a dative object, and example (68). The SR form with a
+human subject is unacceptable unless the subject is contrastively focused.

67) a. *[[Ø gemi-si]-ne kaptan çarp-an] adam
ship-AGR-DAT captain run.into-SR man
‘the man whose ship a captain ran into’


160
Issever (2003) shows that focus and contrastively focus are distinct phenomena in Turkish that
exhibit different properties.

182

b. [[Ø gemi-si]-ne BU kaptan çarp-an] adam
ship-AGR-DAT THIS captain run.into-SR man
‘the man whose ship THIS CAPTAIN (rather than that one) ran into’

68) a. *[[Ø
1
ev-i-ni isçi/askerler yık-an] aile
1

house-AGR-ACC worker/soldiers raze-SR family
‘the family whose house (a) worker(s)/soldiers razed’

b. [[Ø
1
ev-i-ni BU içi/askerler yık-an] aile
1

house-AGR-ACC THIS worker/soldiers raze-SR family
‘the family whose house THIS WORKER/SOLDIERS (rather than that/those) razed’

We know that the NSR form is acceptable (in fact, required) for non-subject
extraction with human subjects, and contrastive focus is also possible on the NSR
clauses, as in (69) through (71) below.

69) [[Ø
1
ineg-i]-ni BU öför-ün ez-dig-i] köylu
1

cow-AGR-ACC THIS driver-GEN run.over-NSR-3s peasant
‘the peasant whose cow THIS DRIVER (rather than that one) ran over’

70) [[Ø
1
gemi-si]-ne BU kaptan-ın çarp-dıg-ı] adam
1

ship-AGR-DAT THIS captain-GEN run.into-NSR-3s man
‘the man whose ship THIS CAPTAIN (rather than that one) ran into’

71) [[Ø
1
ev-i-ni BU içi-nin/askerler-in yık-tıg-ı] aile
1

house-AGR-ACC THIS worker-GEN/soldiers-GEN raze-NSR-3s family
‘the family whose house THIS WORKER/SOLDIERS (rather than that/those) razed’

Returning to the SR examples, we know that the SR form is licensed when [Spec, TP]
is available for the +Wh-expression. We had determined that human subjects cannot
remain low in the structure, i.e. they must always raise to [Spec, TP], so non-subject
SR clauses are expected to be bad because [Spec, TP] will always be occupied by the
non-Wh human subject. In the good SR examples (66), (67)b and 68)b, the

183

relativized expression must have moved to [Spec, TP]. How is it that contrastive
focus on a human subject allows it to remain out of [Spec, TP]?

There are three possible answers:
1- the subject is case-marked by T under long distance agree and the contrastive
focus takes care of the movement at LF.
2- Contrastive Focus allows default case on the subject or perhaps no case.
3- Case is a PF phenomenon for visibility.

Let’s look at these one at a time. The problem with the first option is that so far we
have not assumed Long Distance Case Assignment. In fact, much of the evidence
pointed to case assignment only in a [Spec-Head] syntactic configuration. The
specifier of T was available only when the subject was an NP that did not require
case. Furthermore, the human subjects in the examples do not bear case morphemes.
In Turkish, expressions that are scrambled or sluiced keep their case morphology. I
assume that case is, if anything, a PF phenomenon, i.e. absence of case on DPs leads
to a PF crash. Again, it would be odd to say that that for this circumstance only, case
is assigned via Agree and that PF does not require the case morpheme to have
phonetic content.
The second option is also unappealing because we are coming up with
exceptions to what has been a very consistent state of affairs. We have seen no
evidence to presuppose the existence of a default case. All case-marking has been
quite predictable and falls within standard theoretical assumptions. Allowing human
subjects to have no case would completely negate what we have had to assume based

184

on the behavior specific to human nominals. We have just seen evidence that, in fact,
whereas non-human nominals can be bare and remain in-situ, human nominals must
always be case-marked and can never remain in-situ.
What do we know about human nominals and DPs in general? We know that
they must raise in the structure and be overtly case-marked. If we assume that case-
marking is a PF phenomenon required for visibility/interpretability, then we can argue
that contrastive focus also gives us a PF reflex. That is, contrastively focused
elements must receive intonational stress. This marked pronunciation is obviously
visible at PF. In addition, contrastive focus is a feature that must be checked in the C
domain. Turkish is a Wh-in-situ language. I assume that, just as with Wh-movement
for interrogatives, movement for contrastive focus occurs in covert syntax, perhaps at
LF. By assuming this, we can account for the acceptability of the SR form for
contrastively focused human subjects. The intonation provides the visibility for the
phrase at PF and the focus features drive the raising of the expression at LF. In this
way, both interface requirements of a DP are met. The implication is that other than
the EPP which is a requirement in narrow syntax, raising and case are LF and PF
interface requirement respectively.
There is also a slightly different way to look at the phenomena here. The
issue seemed to be that the grammar requires specifics, in general, and +human
subjects, in particular, to have case and to raise. I had assumed that specifics “had to”
raise to give us the Diesing-style mapping. We know that in these examples, there is
no raising of the focused expression in the syntax because [Spec, TP] had to be vacant
to license the SR form. But it is not clear that the focused expression must raise at

185

LF. It is also possible that the interpretation of specificity is not because of heirarchy
per se but simply a coincidental by-product of DPs—which are interpreted as specific
at LF—having to raise. Under this story, all other requirements being met, it could
also be possible to interpret something as specific, even at LF, without raising. A
crucial requirement is satisfaction of the Case Filter. Now if contrastive focus were
also a case, or to be more precise, if it did at the PF interface whatever case does, then
the “Case” requirement would be met. And the specific interpretation would follow
from the features on D°.
Now as to the PF Case requirement, there is also a view from Korean that
focus and structural case may be doing the same work. Schutze (2001) and Hong
(2002) present evidence from Korean case-stacking phenomena that the structural
case marking on top of inherent case-marked expressions is actually a focus marker.
More interestingly, their data shows that focus and structural case marking are
incompatible. Let’s assume that this prohibition holds for Turkish. That would mean
that in the SR examples, the contrastive focus intonation on the subject does the work
of structural case-marking at PF, this being the flip-side of the second structural case
marking in Korean which is presumably interpreted as focus marking at PF (and LF).
At the same time, we would have to assume that the focus marking in the NSR
examples with genitive subject is not on the subject (structural case and focus being
incompatible) but rather on the T projection. This would follow nicely with our
expectation that there would be a difference in the scope of a contrastively focused
SR vs. NSR clause.

186

I bring up these issues both to unpack what is fundamental to the spirit of this
research project and what is peripheral, and also to point out topics of interest, but
really they are orthogonal to this thesis, and it does not matter here what explanation
one may wish to adopt.
One further issue remains. How do we explain the contrastively focused NSR
examples (69) through (71)? Herburger (2000) notes that (contrastive) focus can
target as little as a quantifier or as much as an entire phrase. In my account, I
assumed that in the SR examples, the entire DP is focused and raises at LF. This
contrasts with the NSR examples where only the Quantifier/Demonstrative of the
subject raises at LF. I have avoided a detailed discussion of the different
interpretation we expect in these two constructions. Again, I have made predictions,
but the question is tangential to this work. I do want to add, though, that in addition
to the scopal differences, the prediction is that there will be a difference in intonation
patterns between these two types of clauses. In the SR version, the entire subject
phrase will be stressed whereas in the NSR version only the demonstrative will
receive stress. It is difficult to tease apart these two or indeed to develop diagnostics
to tease apart the different interpretations between these structures, but it may be
worthwhile, at least, to make a prediction.


187

Chapter 6: Relativization in Psych Verb Constructions



1 Background


The central claim in this thesis is that the properties observed in Turkish relative
constructions can be explained by A-movement and minimality. We saw for example
that relativization of a non-subject, which canonically requires the Non-subject
Relative (NSR) form, is possible with the Subject Relative (SR) form in
unaccusatives, as in (1) but not in unergatives, as in (2).
161
We concluded this was
because the NP subject intervened between the Wh-expression and Tº in unergatives,
but not in unaccusatives, as shown in the tree in (3) for (1), and in (4) for (2).

1) a. [Ø
i
su ak-an] dam
i

water pour-SR roof
‘the roof water pours/drips from’

b. Dam-dan su ak-ıyor.
roof-ABL watership pour-pres.prog.-3s
‘Water is pouring/dripping from the roof’

c. [su-yun ak-tıg-ı] dam
water-GEN pour-NSR-3s roof
‘the roof that the water is dripping/pouring from’

2) a. *at-lar kos-an saha
horse-pl run-SR field
Intended: ‘the field where horses run’



161
As noted in Chapters 2 and 3, the subject must be non-specific, and thus an NP in this account.

188

b. at-lar*(-ın) kos-tug-u saha
horse-pl-GEN run-NSR-3s field
‘the field where the horses run’

3) [Ø
i
su ak-an] dam
i

water pour-SR roof
‘the roof water pours/drips from’

CP

roof
+Wh TP C°

t-roof
+Wh VP T° [+EPP]

roof-ABL
+Wh NP V°
water pours


4) *[at(-lar) kos-an] saha
horse-pl run-SR field

CP


TP C°

field
+Wh vP T° [+EPP]

NP-horses
VP v° [-EPP, -Case]

field-LOC V°
+Wh run



However, we have seen the SR form is licensed in another derivation with
scrambling. When the Wh-expression is embedded in a larger DP which scrambles
around the subject, thereby circumventing the intervention effects of the subject, the
SR RC is acceptable. In the tree in (5), the complex locative DP roughly equivalent
to ‘the field’s inside’ first scrambles (and adjoins) to a projection below TP in move

189

, after which the +Wh genitive expression ‘field’ in its Spec raises to [Spec, TP]
and checks T’s EPP feature. This derivation licenses the SR form.

5) [[ Ø
2
iç-in-de]
1
at t
1
kos-an] saha
2

inside-AGR-LOC horse run-SR field

CP


TP C°

field
+Wh vP T° [+EPP]

DP-LOC vP

NP-horses
VP v° [-EPP, -Case]

DP-LOC V°
run
[[field+Wh]-GEN inside]



There is a class of verbs that does not seem to be as well-behaved in this respect: the
class of predicates that denote psychological states, so called psych verbs. Before we
begin, let’s refresh our memory as to potential confounding factors. First, the SR
form is licensed for non-subjects when the RC subject is an NP (non-specific).
Second, human nominals cannot be NPs. Thus, for non-subject relatives, any time
the clausal subject is +human, the SR form will be barred because the DP subject
must move to [Spec, TP] for case. It is only when a +Wh expression moves to [Spec,
TP] that the SR form is triggered. Thus, to even get the analysis off the ground, we
must control for the human features of the subject. This means that for non-subject
extraction from psych verbs using the SR form, the subject can never be +human.

190

With this much introduction, let’s now look at the behavior of these verbs with
respect to relativization.

1.1 Classes of Turkish psych verbs
Turkish psych verbs fall into several types.
162
These are shown below. In Group 1
and 2, the subject is the Experiencer with an inherently case-marked Theme. In
Group 3, we have an Experiencer subject and an accusative, or structurally case-
marked, Theme. In Groups 4 and 5, the Experiencer is the Accusative object or
Dative object respectively.
163,164


162
Turkish allows a Causative morpheme in many of these psych verbs. Compare (i) and (ii) with (6)
and (7), respectively. The discussion in this Chapter is limited to psych verbs in what I assume is the
base form, that is, without the Causative morpheme.
(i) Bu ben-i sasır-t-tı.
this me-ACC surprise-CAUS-PST
‘This surprised me’
(ii) O ben-i kız-dır-dı
S/He/that me-ACC anger-CAUS-PST
‘S/He/that made me mad/angered me’
Many of the intransitive psych verbs (i.e. those that take PP complements) become transitivized (assign
accusative case to the direct object) with the addition of a Causative morpheme. This suggests that the
Causative is introduced in the vP, a la Pesetsky (1995).
163
It may be more than coincidental that Levin (1993) identifies four classes of psych verbs in English:
Subject Experiencer that takes an object Theme (admire), Subject Experiencer that takes a PP Theme
(marvel), Object Experiencer (amuse), and Object of preposition Experiencer (appeal). These can be
viewed as analogous to the kinds listed for Turkish. Thus, it seems that languages allow psych verbs to
range over all verb classes.
164
Pesetsky (1995) notes that the Theme of psych verbs can be divided into two semantic types: the
Target of emotion and the Subject Matter of the Emotion. The Target of emotion object is evaluated
by the Experiencer, as in (i), where the impression is that Mary gave the play a bad evaluation. The
Subject Matter Theme is shown in (ii) where “play” only participates in a linking to the Experiencer,
that is the play that didn’t please Mary might be an excellent play, but it was written by her rival.
(i) The play didn’t appeal to Mary.
(ii) The play didn’t please Mary.
A preliminary evaluation of Turkish psych verbs does not yield such a tidy division, as the semantic
denotation of the Theme seems to be rather arbitrary. Thus, whereas the Group 1 Subject Experiencer
verb kız- ‘to get angry (at)’ permits both Target and Subject Matter as Themes, the verb bayıl-
‘love/get a kick out of’ seems to require a Target Theme. Similarly, in Group 2, the verb kork- ‘fear’
permits both Target and Subject Matter Themes, while nefretet- ‘despise’ imposes a Target
interpretation on the Theme. Likewise with the Object Experiencer verbs which differ as to the
interpretations they permit for the Causer, either Target or Subject Matter. The Group 4 verb rezil et-
‘disgrace’ allows both semantic roles for the Causer, but sık- ‘bore/frustrate’ requires the Causer to be
the Subject Matter. I mention these to point out that there seems to be no regularity in interpretation
that can be deduced from the various types of verbs or the Case that the arguments bear.

191

Group 1: Subject Experiencer Dative Theme

6) Ben bun-a sasır-dım.
I this-DAT surprised
‘This surprised me’ (Literally: ‘I felt.surprise to this’)

7) Ben on-a kız-dım
I him/that-DAT got.angry
‘I got mad at him’

8) Ben san-a gıpta ed-iyorum.
I you-DAT envy do-PRES
‘I envy/emulate you’

9) Ben san-a bayıl-ıyorum
I you-DAT faint-PRES
‘I love you’ (Literally: ‘I swoon over you’)

10) Ben o-na güven-irim
I that-DAT trust-PRES
‘I trust him’

Group 2: Subject Experiencer Ablative Theme

11) Ben on-dan kork-tum
I that-ABL fear-PST
‘I got scared by that’

12) Ben bun-dan zevk al-dım
I this-ABL got-pleasure
‘I enjoyed this’

13) Ben sen-den nefret ed-iyorum
I you-ABL hatred do-PRES
‘I despise you’


192

14) Ben sen-den bık-tım
I you-ABL fed.up
‘I am fed up with you’

Group 3: Subject Experiencer Accusative Theme

15) Ben sen-i sev-iyorum
I you-ACC love-PRES
‘I love you’

16) Ben sen-i özlü-yorum
I you-ACC miss-PRES
‘I miss you’

17) Ben o-nu / sen-i arzulu-yorum
I s/he/it-ACC/you-ACC desire-PRES
‘I desire s/he/it/you’

18) çocuklar-ı-nı köpek özle-yen adam
I you-ACC miss-PRES
‘I miss you’

Group 4: Accusative Object Experiencer

19) O ben-i etkile-di
It me-ACC affected
‘It affected me’

20) O ben-i rezil et-ti
He/It me-ACC disgrace do-PST
‘He/It disgraced me.’

21) O ben-i sıkıyor
He/It me-ACC bores
‘He/It bores/frustrates me’



193

22) O ben-i boz-du
He me-ACC humiliated
‘He humiliated me’

Group 5: Dative Object Experiencer

23) O ban-a dokun-uyor
s/he/that me-DAT upset-PRES
‘S/he/That upsets me’

24) O ban-a tuhaf gel-di
that me-DAT odd come-PST
‘That seemed strange/odd to me’

25) O ban-a eziyet et-ti
s/he/that me-DAT bother/disturb do-PST
‘S/He/That bothered/disturbed me’

26) O ban-a malum ol-du
that me-DAT obvious be-PST
‘It/That became known to me’ (as in ‘I found out’)



2 Turkish Psych Verbs and Relativization


All psych verbs behave as expected with respect to canonical relativization. As the
examples in (27) demonstrate, the SR is acceptable only for subject extraction and the
NSR is required when relativizing non-subjects.

27) a. [haber-e sasır-an] adam
news-DAT surprise-SR man
‘the man who felt surprised by the news’


194

b. *[haber-e sasır-dıg-ı] adam
news-DAT surprise-NSR man

c. [adam-ın sasır-dıg-ı] haber
man-GEN surprise-NSR-3s news
‘the news that the man felt surprised by’

d. *[adam(-ın) sasır-an] haber
man(-GEN) surprise-SR news

2.1 Experiencer subjects

The complication with these Subject Experiencer verbs is that, at first glance, they
seem to require human subjects. Thus, the example in (28) is marginal. And the
example in (29) can be viewed as an instance of coercion of sorts, that is, that the
subject ‘dog’ is being endowed with human characteristics. Note that example (31)
with human Experiencer as the subject is perfect. If the Experiencer subject is being
encoded as “human-like”, it would account for the RC in (30)a, where the SR form
with an NP subject is unacceptable. Contrast this with the corresponding but good
NSR clause in (30)b, where the DP subject is case-marked. Is this because there is
something peculiar about the structure of psych verbs, or is (30)a bad because non-
subject extraction using the SR form is not possible with human subjects, and here
‘dog’ is behaving as if it were syntactically +human.

28) ??at tren sesi-ne sasır-dı
horse train noise-DAT felt-suprised
Intended: ‘The horse was surprised by the noise of the train’

29) köpek/??at egetimci-nin yeni düdük çal-ma-sı-na sasır-dı
dog/horse trainer-GEN new whistle blow-INF-3POSS-DAT felt-surprised
‘The dog/??horse was surprised by the trainer(’s) blowing a new whistle’

195

30) a *[[ Ø
1
yeni düdük çal-ma-sı-na] köpek sasır-an] egetimci
1

new whistle blow-INF-3POSS-DAT dog feel.surprised-SR trainer
Intended: the trainer who dogs felt-surprised by (his) blowing a new whistle’

b. [[Ø
1
yeni düdük çal-ma-sı-na] köpeg-in sasır-dıg-ı] egetimci
1

new whistle blow-INF-3POSS-DAT dog-GEN feel.surprised-NSR-3S trainer
‘the trainer who the dog felt-surprised by (his) blowing a new whistle’

31) ögrenci müdür-ün yeni düdük çal-ma-sı-na sasır-dı
student principal-GEN new whistle blow-INF-3POSS-DAT felt-surprised
‘The student was surprised by the principal(’s) blowing a new whistle’

Surely, it cannot be possible for a language to encode dogs as having “human-like”
qualities and horses not. Furthermore, we are looking at syntactic reflexes here, prior
to evaluation or interpretation by the semantic component. Let’s therefore keep our
analysis at the syntactic level and just assume at this point that Subject Experiencer
Psych verbs seem to require DP subjects. We had determined that ç-features rest on
D. In Turkish, ç-features are required for “the quality of being human”. But recall
that we had determined that this was a selectional requirement imposed by the
lexicon. We can assume that this quality can be extended idiosyncratically to animals
depending on the individual user. This would be analogous to the alternation we see
in English between the +human pronouns him/her and the –human it when referring
to animals. People who love dogs refer to their pets as “he” or “she” while a non-
dog-lover may use “it” to refer to the same animal. It is no surprise, then, that horse
lovers would find the marginal (28) and (29) with ‘horse’ acceptable. The proposal,
then, is that psych verbs require the Experiencer subject to have a D-feature (which is
the same as ç-features), and that there is some leeway in the acceptability of this
feature on animals which is posited idiosyncratically. This will account for the facts
so far.

196

Let’s look at another example, the verb gıpta ‘to envy/covet/emulate’.
Relativizing a locative using the NSR using gıpta is fine, (32)a. As expected,
relativizing a non-subject using the SR form is bad in (32)b-c because the clausal
subject ‘villagers’ is +human and must obligatorily raise to [Spec, TP] for case.

32) a. [köylüler-in egitimli kadınlar-a gıpta et-tig-i] ülke
villagers-GEN educated women-DAT envy.do-NSR-3s country
‘the country where the peasants envy educated women’

b. *[köylü(-ler) egitimli kadınlar-a gıpta ed-en] ülke
villager(s) educated women-DAT envy.do-SR country
Intended: ‘the country where peasants envy educated women’

c. *[egitimli kadınlar-a köylü(-ler) gıpta ed-en] ülke
educated women-DAT villager(s) envy.do-SR country

Look what happens when the +Wh-expression is embedded. In the NSR example in
(33)a, the relative head ‘villagers’ has been extracted from the complex DP-subject
‘villagers’ daughters’. Not surprisingly from what we saw in previous chapters
regarding extraction from complex subjects, the parallel SR form in (33)b is also
acceptable. When extracting from within the Dative object though, note that the NSR
is acceptable (34)a, whereas the parallel SR form is unacceptable. As demonstrated
in (34)b-c, the subject ‘villagers’ daughters’ may not remain in situ or without case.
165

Again, the +human subject must raise to [Spec, TP] for case, making [Spec, TP]
unavailable for the +Wh-expression. Notice now the acceptable SR example in (35)
where the verb has been passivized. Here the Experiencer has been demoted; it is
now the complement of an adjunct ‘by-phrase’ with ablative case. With the
Experiencer former-subject out of the way in an adjunct clause, [Spec, TP] is now

165
Example (34)c is an attempt to avoid intervention effects by scrambling the dative object around the
subject. .

197

available for the +Wh specifier of the Dative DP, a configuration which licenses the
SR form.

33) a. [[Ø
1
kızlar-ın-ın] egitimli kadınlar-a gıpta et-tig-i] köylüler
1

girls-AGR-GEN educated women-DAT envy.do-NSR-3s villagers
‘the villagers
1
whose [such that (their
1
)] daughters envy educated women’

b. [[Ø
1
kızlar-ı] egitimli kadınlar-a gıpta ed-en] köylüler
1

girls-AGR educated women-DAT envy.do-SR villagers
‘the villagers
1
whose [such that (their
1
)] daughters envy educated women’

34) a. [egitimli kadınlar-ın [Ø
1
kızlar-ı-na] gıpta et-tig-i] köylüler
1

educated women-GEN daughters-AGR-DAT envy.do-NSR-3s villagers
‘the villagers
1
who [such that] educated women envy (their
1
) daughters’

b. *[egitimli kadın(-lar) [Ø
1
kızlar-ın-a] gıpta ed-en] köylüler
1

educated woman(-pl) daughters-AGR-DAT envy.do-SR villagers
Intended: ‘the villagers
1
who [such that] educated women envy (their
1
) daughters’

c. *[[ Ø
1
kızlar-ın-a] egitimli kadın(-lar) gıpta ed-en] köylüler
1

daughters-AGR-DAT educated woman(-pl) envy.do-SR villagers

35) [[Ø
1
babalar-ı-na] genç adamlar taraf-ın-dan gıpta ed-il-en] askerler
1

fathers-AGR-DAT young men viewpoint-ABL envy.do-PASS-SR soldiers
‘the soldiers
1
whose [such that (their
1
)] fathers are envied by young men’

Let’s say we accept the lexico-semantic fact that it doesn’t make sense for gıpta
‘envy/covet’ to have a non-human subject. After all, a dog “envying” the food or the
collar of another dog is a little weird. Let’s take a look at another verb that in
principle should permit a non-human subject. It seems though that the same facts
hold for the Subject Experiencer verb güven ‘trust’. First, let’s look at the behavior of
this verb with a human subject as in (36). As expected, extraction of a non-subject is
fine using the NSR form (36)a and bad using the SR form (36)b. Passivization
renders the human subject out of the way, and the SR form is again licensed (36)d.

198

36) a. çocuklar [okulun müdür-ü]-ne güvenirler
children school-GEN principal-AGR-DAT trust
‘(The) children trust the school’s principal’

b. [[Ø
1
müdür-ü-ne] çocuklar-ın güven-dig-i] okul
1

principal-AGR-DAT children-GEN trust-NSR-3s school
‘the school whose principal the children trust’

c. *[[ Ø
1
müdür-ü-ne] çocuk(-lar) güven-en] okul
1

principal-AGR-DAT child(ren) trust-SR school
Intended: ‘the school whose principal children trust’

d. [[ Ø
1
müdür-ü-ne] güven-il-en] okul
1

principal-AGR-DAT trust-PASS-SR school
‘the school whose principal is trusted’

Now let’s look at this Experiencer subject verb with a non-human subject. As
expected, the SR form is barred when relativizing a non-subject because the subject
intervenes. This is demonstrated in the tree in (37)b for the unacceptable SR in (37)a.
In (37)a, the EPP of T fails to be satisfied. The non-specific subject hayvan ‘animal’
cannot satisfy it, and, whereas the Dative Wh-expression can satisfy the EPP, it is
blocked from doing so by the intervening subject. The only way to relativize the
Dative Theme of the verb ‘trust’ is with the NSR form with a specific subject, as in
(38)a.
166
As demonstrated in the tree in (38)b, the subject raises to [Spec, TP]
satisfying T’s EPP and is assigned genitive case while the +Wh-Dative long-distance
A-bar moves to [Spec, CP].

166
In Chapter 5, I showed that the denotation of non-specificity on a nominal expression where a D-
feature is imposed on it (as in the case of humans, and as we will see, psych verb Experiencers) is
achieved in Turkish by the use of a partitive construction, (i). I assume that the underlying structure of
the subject in (i) is as in (ii) where the non-specific (existential) ‘animals’ is in the restrictor of the null
DP-animals. Thus, the meaning in (i) is technically ‘the person who [of the animals] (some) animals
trust’.
(i) [hayvanlar-ı-nın güven-dig-i] insan
animals-AGR-GEN strust-NSR person
‘the person who animals trust’
(ii) [
DP
[
DP
animals-GEN] [
NP
animals-POSS.AGR] Dº ]-GEN

199

37) a. *[hayvan güven-en] insan
animal trust-SR person
Intended: ‘the person who animals trust’

b.
CP


TP C°


vP T° [+EPP]

animal
VP v°

person-DAT V°
+Wh trust



38) a. [hayvan-ın güven-dig-i] insan
animal-GEN trust-NSR person
‘the person who the animal trusts’

b.
CP


TP C°

DP-animal-GEN
vP T° [+EPP]

DP-animal
VP v°

person-DAT V°
+Wh trust


We saw in (5), an example of an alternative derivation using the SR form: the +Wh-
expression is a “free-rider” in a larger DP and literally gets carried around the
blocking element. Recall how this worked: the relativized expression is embedded in
a DP (as the specifier), the DP scrambles around the subject, and adjoins to a position
lower than [Spec, TP]. The relativized expression is now free to raise to [Spec, TP]
because there are no interveners, and the SR form is triggered. This alternative

200

derivation does not seem to be possible with a Subject Experiencer psych verb. As
shown in the examples in (39), the SR form is still unacceptable.

39) a. *[[Ø
1
çocuklar-ı-na] hayvan(-lar) güven-en/kız-an] insanlar
1

children-AGR-DAT animal(s) trust-SR/get.angry-SR people
Intended: ‘the people whose children animals trust/get angry at’

b. [[Ø
1
çocuklar-ı-na] hayvanlar-ın güven-dig-i/kız-dıg-ı] insanlar
1

children-AGR-DAT animals-GEN trust-NSR/get.angry-NSR people
‘the people whose children the animals trust’

The conclusion that Experiencer psych verbs “select” for human subjects is too strong
as evidenced by the acceptability of (39)b. We must therefore assume that Turkish
psych verbs require a subject with a D feature. Note that we are making a claim that
psych verb selection is evaluated in overt syntax, prior to LF.
167

Although we are looking at Turkish facts here, this requirement about
Experiencers may be universal. Experiencers need not be definite but they cannot be
existential. In English, for example, in sentence (40) we get the reading there were
some children who played in this park. It certainly doesn’t mean all children in
general, played in the park. Contrast this with sentence (41) where the reading seems
to be that all (contextually relevant) children, in general, trusted the policeman.

40) Children played in this park

41) Children trusted this policeman.


167
In essence, this is the flip-side of the requirement that +human nominals have a D feature. That is,
we have two instances of selection, one nominal, that human nominals must merge with a D, and the
other verbal, that psych verbs select for a DP as Experiencer, that produce identical intervention effects
in the syntax.

201

The implication is that psych verbs do not allow Experiencers to be existential, that
this is a lexico-semantic selectional requirement, and that this requirement is satisfied
in unique ways in varying grammars. In Turkish, absence of a D-feature is
interpreted as existential.
168
As shown in the table in (42), only an indefinite non-
specific nominal does not require a D°.
169
And, as we have seen repeatedly, the
presence of a D-feature on a nominal has syntactic consequences.

42) Encoding Definiteness and Specificity in Turkish

Definite Indefinite Indefinite
Specific Specific Non-Specific
Requires D° Requires D° No D°


We are at a point where we can predict the behavior of an expression in a RC based
on whether it is a DP or NP. Let’s look at Subject Experiencer psych verbs that take
an Ablative Theme, for example. Not surprisingly, as shown in (43)a-b, extraction of
a locative is allowed with the NSR, but not the SR. Extracting from the subject
permits both forms, (44)a-b. Extracting from the Ablative Theme is permitted only
with the NSR form, regardless of the word order, (44)c-d-e. This is typical behavior
of a clause with a +human subject. In (44), the relative head is the inanimate okul

168
See Enç (1991) and Chapter 3 of this dissertation. Karimi (2003) offers a revised version of Enç
(1991) according to which a nonspecific nominal (an NP, according to our account here), either lacks a
referent (=kind-level) or is existential. The implication is twofold: that neither of these interpretations
are permitted by the selectional requirement of the Experiencer of a psych verb, and that semantic
notions such these and their counterparts are encoded in the lexicon.
169
Chomsky (1999: fn.10) associates D with referentiality; nonreferential nominals such as non-
specifics, quantified and predicate nominals need not be assigned D.

202

‘school’ which moved out of the larger Theme DP [school’s naughty students]. I
include the example in (45) to demonstrate that making the relative head +human, as
in ‘parents’ from the Theme DP [parents-GEN naughty children-AGR] does not effect
the outcome: non-subject extraction using the SR form is impossible with a +human
clausal subject.

43) a. [ögretmenler-in yaramaz ögrenciler-den bık-tıg-ı] okul
teachers-GEN naughty students-ABL fed.up-NSR school
‘the school where the teachers are fed up with naughty students’

b. *[ögretmenler yaramaz ögrenciler-den bık-an] okul
teachers naughty students-ABL fed.up-SR school
‘the school where teachers are fed up with naughty students’

44) a. [[Ø
1
ögretmenler-i-nin] yaramaz ögrenciler-den bık-tıg-ı] okul
1

teachers-AGR-GEN naughty students-ABL fed.up-NSR school
‘the school whose teachers are fed up with naughty students’

b. [[Ø
1
ögretmenler-i] yaramaz ögrenciler-den bık-an] okul
1

teachers-AGR naughty students-ABL fed.up-SR school
‘the school whose teachers are fed up with naughty students’

c. [ögretmenler-in [Ø
1
yaramaz ögrenciler-in]-den bık-tıg-ı ] okul
1

teachers-GEN naughty students-AGR-ABL fed.up-NSR school
‘the school where the teachers are fed up with its naughty students’

d. *[ögretmenler [Ø
1
yaramaz ögrenciler-in]-den bık-an] okul
1

teachers naughty students-AGR-ABL fed.up-SR school
Intended: ‘the school where teachers are fed up with (its) naughty students’

e. *[[Ø
1
yaramaz ögrenciler-in]-den ögretmen(-ler) bık-an] okul
1

naughty students-AGR-ABL teacher(s) fed.up-SR school

45) a. [ögretmenler-in [Ø
1
yaramaz çocuklar-ın]-dan bık-tıg-ı] anne-babalar
1

teachers-GEN naughty children-AGR-ABL fed.up-NSR parents
‘the parents who the teachers are fed up with their naughty children’

b. *[ ögretmenler [Ø
1
yaramaz çocuklar-ın]-dan bık-an] anne-babalar
1

teachers naughty children-AGR-ABL fed.up-SR parents
‘the parents who teachers are fed up with their naughty children

203

Furthermore, just as with Dative Theme psych verbs, the Experiencer subject with
Ablative Theme behaves the same as RCs with +human subjects. This is
demonstrated in (46)b with the inanimate ‘publishing house’ as the Experiencer
subject; extraction of a non-subject using the SR form is unacceptable even when the
Theme has been scrambled around the subject. This is evidence that the psych verb
bık ‘fed.up’ must require its Experiencer subject to be a DP. Just as we saw with
subjects that were +human (which we determined in the Chapter 5 had to be DPs), the
Experiencer subject cannot remain in situ without case.

46) a. [ Ø
1
[yazar-ın geç kalmasın]-dan bık-an] yayınevi
1

author-GEN tardiness-AGR-ABL fed.up-SR publishing.house
‘the publishing house which is fed up with the author’s tardiness’

b. *[ [Ø
1
geç kalmalar-ın]-dan yayınevi bık-an] yazarlar
1

tardiness-AGR-ABL publishing.house fed.up-SR author
Intended: ‘the authors whose tardiness publishing houses are fed up with’

Note that there is no requirement on the lexical item ‘publishing house’ itself, that it
be a DP. We saw in Chapter 5, an example, where the nominal ‘publisher’ led to a
derivational crash but not the expression ‘publishing house’. This example is
repeated as (47). The RC in (47)a is unacceptable with ‘publisher’ as the subject
because ‘publisher’ being +human must enter the derivation as a DP. No such
requirement exists on ‘publishing house’ which as an NP can remain in situ.

47) a. [[Ø
1
yazar-ı]-nı yayınevi /*yayıncı aray-an] makale
1

author-POSS-ACC publishing.house/publisher search-SR article
‘the article whose author publishing.houses/*publishers are looking for’



204

yayınevi-nin
b. [[Ø
1
yazar-ı]-nı /yayıncı-nın ara-dı-ı] makale
1

author-POSS-ACC publishing.house-GEN search-NSR-3S article
/publisher-GEN
‘the article whose author the publishing.house/the publisher is/was looking for’


We are forced to conclude that it is the psych verb that is requiring the same lexical
item, ‘publishing house’, to behave as a DP.
2.1.1 Subject Experiencers with accusative Theme
We already know that relativization of an accusative DP is not possible using the SR
form. Accusatives cannot move to [Spec, TP]. We saw that the possessor of the
accusative can be relativized using the SR form as long as the subject is an NP. The
SR clause in (48)a is one such example, with the derivation in (48)b. The subject,
dog, is an NP and remains in situ. The object with the +Wh specifier man-GEN,
moves around the subject to a higher Spec of vP, as in , and is assigned accusative
case. The genitive possessor then raises from the Spec of the accusative object to
[Spec, TP], in , and then again to [Spec, CP], in .

48) a. [ [Ø
1
çocuklar-ı]-nı köpek yala-yan] adam
1

children-AGR-ACC dog lick-SR man
‘the man
1
who [such that] a dog/dogs licked (his
1
) children’
b.
CP


TP C°


vP T° [+EPP]

DP+ACC
man dog
VP v°

DP V°
trust
[ [ man’s ] children]
+Wh+GEN


205

But, psych verbs with Experiencer subjects do not permit such a derivation. The
example in (49)a is unacceptable. The derivation in (49)b should proceed in an
identical manner to the one in (48)b. Why does this derivation crash? Because the
Experiencer subject cannot be an NP; it must raise for case to [Spec, TP]. The
derivation in (49)b is a Case Filter violation.

49) a. *[ [Ø
1
çocuklar-ı]-nı köpek sev-en] adam
1

children-AGR-ACC dog love-SR man
Intended: ‘the man
1
who [such that] dogs love (his
1
) children’

b. CP


TP C°


vP T° [+EPP]

DP+ACC
dog
VP v°

DP V°
trust
[man’s children]


In conclusion, none of the subject Experiencer psych verbs will allow non-subject
relativization using the SR form. The subject Experiencer has to be a DP and must
obligatorily raise to [Spec, TP] for case. This will bleed the SR form.
2.2 Experiencer objects

Let’s now turn to psych verbs with Object Experiencers. From what we observed in
Subject Experiencers, will we find that Object Experiencers are also required to be
DPs? The evidence certainly points in this direction, as shown by the unacceptability
of the sentence in (50) without accusative case on the Experiencer object ‘dog(s)’.

206

Notice from the English equivalent that there is no sense of ‘dogs’ as being definite;
but it seems the syntax requires that the Experiencer object be case-marked, a
requirement on DPs. This is support for the view that psych verbs do not permit an
existential Experiencer.

50) Bu çocuk köpek(-ler)*(-i) çok sık-ar.
this child dog(-pl)*(-ACC) much bother-AOR
‘This child bothers dogs a lot.’ (In the sense that ‘dogs feel bothered’)

While the Experiencer object is required to be a DP, there is no such categorial
requirement on the subject. As demonstrated in (51)b, extraction from the
Experiencer object is possible using the SR form. This is because the subject,
‘disastrous news’, is a non-specific NP which can (indeed must) remain without case
in situ. In this phrase, [Spec, TP] is vacant for the +Wh-expression, ‘parents’ to move
to, and the SR form is required.

51) a. [felaket haberleri-nin [Ø
1
çocuklar-ı-nı] etkile-dig-i] anne-babalar
1

disastrous news-GEN children-AGR-ACC affect-NSR ] parents
‘the parents whose children the disastrous news affected’

b. [ [Ø
1
çocuklar-ı-nı] felaket haberleri etkiley-en] anne-babalar
1

children-AGR-ACC disastrous news affect-SR ] parents
‘the parents whose children disastrous news affected’

And, as we had hypothesized, the Experiencer object cannot be non-specific, at least
not syntactically; it must be a DP and bear accusative case, (52).

52) a. *Felaket haberleri çocuk etkil-er
disastrous news child affect-AOR

207

Intended: ‘Disastrous news affects children’

b. Felaket haberleri çocug-u / çocuk-lar-ı etkil-er
disastrous news child-ACC/children-ACC affect-AOR
‘Disastrous news affects the child/the children/children’

Another example of an Accusative Object Experiencer psych verb is büyüle ‘to
fascinate’. Recall that an accusative DP cannot A-move to [Spec, TP]. This is why
(53)a is bad even though the RC subject, ‘magic tricks’ is an NP, and cannot raise to
Tº. When the subject is a DP, as in (53)b, the NSR form is required; the +Wh-
accusative object, ‘children’ A-bar moves to [Spec, CP]. In the SR in (53)c, the
relative head is embedded in the accusative object. The NP subject remains in situ,
while the possessor of the Accusative Experiencer raises to [Spec, TP] triggering the
SR form.

53) a. *[sihirbaz hileleri büyüley-en] çocuklar
magician tricks fascinate-SR children
Intended: ‘the children who are fascinated by magic tricks’

b. [sihirbaz hileleri-nin büyüle-dig-i] çocuklar
magician tricks-GEN fascinate-NSR children
‘the children who are/were fascinated by the magic tricks’

c. [[ Ø
1
çocuklar-ı]-nı sihirbaz hileleri büyüley-en] anneler
1

children-AGR-ACC magician tricks fascinate-SR mothers
‘the mothers
1
who [such that] magic tricks fascinated (their
1
) children’

Note that relativization of a locative is not possible using the SR form (54)a. The tree
in (54)c demonstrates that both the subject and the accusative object intervene

208

between the locative and Tº.
170
A-movement, move , of a locative to [Spec, TP] is
a minimality violation, but A-bar movement, as in , is possible, hence the
acceptability of the NSR in (54)b.

54) a. *[çiçekler çocuklar-ı büyüle-yen] bahçe
flowers children-ACC fascinate-SR garden
Intended: ‘the garden where flowers fascinate children’

b. [çiçekleri-nin] çocuklar-ı büyüle-dig-i] bahçe
flowers-GEN children-ACC fascinate-NSR garden
‘the garden where the flowers fascinate children’

c. CP


TP C°


vP T° [+EPP]

children+ACC
flowers
VP v°

garden-LOC
+Wh children V°
fascinate





This confirms what we saw in previous chapters, extraction from the accusative
object using the SR form is possible when the subject is an NP. In the SR example in
(55), the subject, ‘Iranian films’ is an NP which remains in situ. The accusative
object is in [Spec, vP], and the +Wh-teachers has raised from its Spec to [Spec, TP].


170
Except in impersonal passive constructions which have no external argument. We saw in Chapters 1
and 2 that these require the SR form as the +Wh-locative is often the only DP in the clause that can
satisfy T’s EPP feature.

209

55) [[Ø
1
ögrenciler-i]-ni Iran filmleri büyüle-yen] ögretmenler
1

students-AGR-ACC Iranian films fascinate-SR teachers
‘the teachers whose students Iranian films fascinate’

Recall that the SR form is impossible when the subject is +human. Compare the
good (56)a with the unacceptable (56)b with a +human subject. The unacceptability
of (56)b is predictable because we have already concluded that +human subjects
cannot be NPs that remain in situ; they must raise to [Spec, TP] for case.

56) a. [[Ø
1
ogular-ı]-nı konusmalar/seks filmi tahrik ed-en] anneler
1

sons-AGR-ACC speeches/sex film arouse.do-SR mothers
‘the mothers whose sons speeches sex films provoke/arouse’

b. *[[Ø
1
ogular-ı]-nı konusmaci(-lar)/seksi kadın(-lar) tahrik ed-en] anneler
1

sons-AGR-ACC speaker(s)/sexy women arouse.do-SR mothers
‘the mothers whose sons speakers/sexy women provoke/arouse’



3 Conclusion


Let’s take stock. Relative clauses with Experiencer subject psych verbs exhibit
behavior similar to RCs with human subjects: non-subject extraction using the SR
form is not possible. Psych verbs with Experiencer objects allow non-subject
extraction using the SR form as long as the subject is an NP and the movement to
[Spec, TP] of the relative head obeys minimality. There is the added complication of
the accusative object which is frozen for further A-movement. Just as in other
transitive (i.e. accusative) constructions, non-subject relativization of the expression
in the Spec of the object licenses the SR form when the clausal subject is an NP. The

210

conclusion of the evidence here is that we have not had to assume any new
technology to explain relativization with respect to psych verbs. We saw identical
behavior in extraction from clauses with human subjects, and we determined this was
because +human features required a DP layer. The only “innovation” in this chapter
is that psych verbs seem to require that the Experiencer be non-existential, that is,
they display a selectional restriction, prior to overt syntax, that the Experiencer be a
DP, regardless of its theta-role or initial merge position. Although we looked mainly
at Subject Experiencers, we saw evidence that suggested this was true for Accusative
objects as well. This requirement is harder to test on Dative Experiencers. Recall the
two diagnostics used to test DP-hood: overt case and required raising to a functional
case-assigning projection. Neither test can be used to determine the category of a
Dative expression.
171
First, datives enter the derivation already case-marked, and
second, the subject is base-generated in a position higher than the Dative, and thus
intervenes in the movement of the Dative to [Spec, TP] (regardless of whether the
subject is an NP or DP).
172
For Datives, there is no way of assessing whether it must
A-move, or whether it blocks A-movement of lower nominals. These tests require
the availability of the SR form, which can never be licensed when relativizing the
Dative Experiencer or a lower expression. Having said this, however, it is reasonable
to think that the requirement that Experiencers be DPs probably holds of Dative
Experiencers as well.

171
These are the two syntactic diagnostics used throughout this work. Other diagnostics may be
available, WCO, Binding, and semantic interpretation.
172
In Chapter 4, we saw that a c-commanding NP induced intervention effects even though the NP
itself could not satisfy the EPP on T°.

211

Obviously, this is not the complete story about psych verbs in Turkish, but the
fact that we were able to account for diverse behavior within a limited theoretical
account is interesting. The facts here are reminiscent of what Belletti and Rizzi
(1988) (B&R) conclude in their study of Italian Psych verbs.
Substantive distinctions between -roles are irrelevant within formal
grammar but play a crucial role at the interface between formal
grammar and other cognitive systems. In fact, they contribute to
determining the initial syntactic representations (D-structures)
through a system of mapping principles projecting -structures onto
syntactic structures ... [e]verything is mediated through structure, and
grammatical processes only refer to structural information which
indirectly reflects information ... -hierarchies and the like
intervene only once, in the formation of D-structures. From there
on, reference to such entities is excluded in formal grammar.
(Belletti and Rizzi 1988:294-295 [bold font mine])

I have highlighted the last lines in bold because this is exactly what the evidence from
Turkish psych verbs seem to be indicating. We have seen that Turkish psych verbs
require that the Experiencer enter the derivation as DPs. They do not permit NP
Experiencers. NPs do not need to raise for case, and we saw that Experiencer
subjects and direct objects behave in the syntax as if they were DPs. We were unable
to coerce them to behave as NPs. It must be that these expressions enter into the
derivation (at D-structure, to use B&R’s terminology) with these features and that the

212

derivation proceeds, failing or crashing, based on the requirement imposed on the
nominal elements prior to Merge.
Although Belletti and Rizzi’s work was under a different framework and for a
different language, the evidence in this chapter seems to support their proposal.
Indeed, we were able to explain Turkish relatives without resorting to additional
projections for psych verbs a la Pesetsky (1994). If we extend our theory merely to
include the lexical-selectional requirement that psych verbs denote states that only
humans can exhibit, we need not posit any other rules to explain the behavior of
relativization from psych-verb constructions. To be more precise, the evidence
suggests that the constraint is that psych verbs require the Experiencer to be perhaps
sentient (as encoded by -features) and non-existential. The syntactic consequence is
that the Experiencer of psych verbs is always a DP. Thus, subject Experiencers
always need case, and [Spec, TP] will always be occupied by the Experiencer subject.
This is not to say that the discussion in this chapter is the complete story on
the structure of psych verbs. Obviously other syntactic (and semantic) properties of
Turkish psych verbs need to be studied. But, on first pass, psych-verbs, otherwise
interesting because of their diversity in assigning Experiencer-Theme roles, offer
nothing exciting in terms of relativization. Just like other examples throughout this
thesis, the non-subject SR can only be licensed when the move of the +Wh-DP to
[Spec, TP] obeys minimality.


213

Chapter 7: Relativization from Infinitivals in Turkish


1 Background


Sezer (1986) notes that like Japanese, Turkish does not obey Ross’ (1967) Sentential
Subject Constraint. Examples (1), (2) and (3) show extraction from English,
Japanese
173
and Turkish
174
, respectively.

1) a. The teacher
1
[who
1
the reporters expected [that the principal would fire t
1
]] is
a crusty old battleax.

b. *The teacher
1
[ who
1
[ that the principal would fire t
1
was expected by the
reporters]] is a crusty old battleax.

c. The teacher
1
[who
1
it was expected by the reporters [that the principal would
fire t
1
]] is a crusty old battleax.

2) a. [watakusi ga t
1
au koto/no] ga muzukasii hito
1

I meet that difficult person
‘(Lit.) the person whom that I see (him) is difficult’

b. [kimi ga t
1
au koto/no] ga atarimae no hito
you meet that matter.of.fact person
1

‘(Lit.) the person whom that you see (him) is matter of fact’

c. [kare ga t
1
kaita koto] ga yoku sirarete-iru bun
1

he wrote that well known-is article
‘(Lit.) the article which that he has written (it) is well known’




173
Japanese examples from Kuno (1973:241). I have added t (for trace) in 2) and (3) for ease of
comparison.
174
The Turkish examples (3) through (12) are from Sezer (1986). .

214

3) a. [[t
1
iyilei-ce-i] son derece üpheli ol-an] hasta
1

recover-FUT-3s last degree doubtful be-SR patient

hastane-den yürü-yerek çık-tı.
hospital-ABL walking.by leave-PST

‘The patient
1
[who [[that (he
1
) would recover] was extremely doubtful]]
walked out of the hospital.’

b. [[t
1
daha uzun sür-ece-i] anla-ıl-an] ekonomik kriz
1

yet long last-FUT-3s realize-PASS-SR economic crisis

memurlar-ı bunalt-tı.
civil.servants-ACC depressed

‘The economic crisis
1
[which [it is realized that [(it
1
) will continue longer]]]
depressed the civil servants.’

c. [[t
1
tamir ed-il-me-si milyonlar-a malol-an] stat
1
-ta koyun-lar otlu-yor.
repair do-PASS-INF-3s millions-DAT cost-SR stadium-LOC sheep-pl grazing
‘Sheep are grazing in the stadium
1
[which [[(its
1
) being repaired] cost millions]].’

We saw in (3) that sentential subjects are not islands in Turkish. However, according
to Sezer (among others) infinitival sentential subjects are islands. Relativizing out of
an infinitival subject, as in the (b) examples in (4) and (5), is unacceptable.

4) a. Stad-ı tamir et-mek pahalı-ya maloldu.
stadium-ACC repair do-INF expensive-DAT cost-PST
‘To repair the stadium was costly.’

b. *[[ Ø
1
tamir et-mek] pahaliya malol-an] stad
1

repair do-INF expensive-DAT cost-SR stadium-ACC
Intended: ‘the stadium that to repair it was costly’

5) a. Problem-i çöz-mek zor-dur.
problem-ACC solve-INF difficult-be-AOR
‘To solve the problem is difficult.’

b. * [[Ø
1
çöz-mek] zor ol-an] problem
1

solve-INF difficult be-SR problem
Intended: ‘the problem which to solve is difficult.’

215

Sezer demonstrates that infinitive clauses in Turkish are not of themselves islands. In
examples (6) through (8), the infinitive is a complement or verbal argument. And the
example in (9) shows that relativization out of infinitival adjunct clauses is also
possible.

6) a. Bakan [meclis-te konu-mak] iste-di
minister parliament-LOC speak-INF want-PST
‘The minister wanted to speak in the parliament.’

b. [bakan-ın [Ø
1
konu-mak] iste-di-i] meclis
1

minister-GEN speak-INF want-PST-3S parliament
‘the parliament that the minister wanted to speak (in)’

7) a. yeni idare [kitaplar-ı yasakla-ma]-ya çalı-ıyor
new administration books-ACC ban-INF-DAT try-pres
‘The new administration is trying to ban (certain) books.’

b. [yeni idare-nin [Ø
1
yasakla-ma]-ya çalı-tı-ı ]] kitaplar
1

new administration-GEN ban-INF-DAT try-NSR-3s books
‘the books that the new administration is trying to ban’

8) a. Ali [kitab-ı oku-mak]-tan zevk al-ıyor.
Ali book-ACC read-INF-ABL pleasure take-aor
‘Ali gets pleasure from (i.e. enjoys) reading the book.’

b. [Ali-nin [Ø
1
okumak]-tan zevk al-dı-ı]] kitap
1

Ali-gen read-INF-ABL pleasure take-NSR-3s book
‘the book that Ali enjoys reading’ (lit: ‘the book Ali gets pleasure from (to) read’)

9) a. Ali okul-a gir-mek için on yıl ura-tı
Ali school-DAT enroll-INF for ten year struggle-pst
‘Ali tried for ten years to get into that school’

b. [Ali-nin [[Ø
1
gir-mek] için] on yıl ura-tı-ı] okul
1

Ali-GEN enroll-INF for ten year struggle-NSR-3s school
‘the school that Ali tried for ten years to get into’


216

To explain the unacceptability of extraction from infinitival sentential subjects, Sezer
formulates the constraint in (10).

10) The Unmarked Sentential Subject Constraint

Nothing may relativize out of a clause that is unmarked for agreement
and is dominated by a subject NP node.

This constraint seems to correctly predict the minimal pair in (11)b and (12)b. In
these examples, both infinitival clauses are subjects, but the infinitive verb in (11)b
bears agreement inflection whereas in (12)b, the verb is an uninflected infinitive.
175


11) a. [[Kitab-ın yazıl-ma-sı] Ali-ye bebin liraya otur-du
book-GEN write-PASS-INF-3s Ali-DAT 5000 lira come-PST
‘The writing of (this) book cost Ali 5000 lira.’

b. [[Ø
1
yazıl-ma-sı] Ali-ye bebin liraya otur-an] kitap
1

write-PASS-INF-3s Ali-DAT 5000 lira come-SR book
‘the book that its writing cost Ali 5000 lira’

12) a. [[Kitab-ı yaz-mak] Ali-ye bebin liraya otur-du
book-ACC write-INF Ali-DAT 5000 lira come-PST
‘The writing of (this) book cost Ali 5000 lira.’

b. *[[Ø
1
yaz-mak] Ali-ye bebin liraya otur-an] kitap
1

write- INF Ali-DAT 5000 lira come-SR book
Intended: ‘the book that to write (it) cost Ali 5000 lira’

Unfortunately, the constraint in (10) is stipulative, and it would obviously be better to
identify a principled account of the phenomena, if possible. In fact, the constraint
does not seem to exist at all. Let’s take another look at the bad examples in (4) and

175
We will look at inflected infinitival constructions a little further on in this chapter.

217

(5) repeated as (13) and (15). Notice that in (13)b the RC is the SR form which
means that the relativized expression would have to move through the RC [Spec, TP],
but the relative head inside the infinitival subject has accusative case. We have
already seen that DPs with accusative case are barred from moving to [Spec, TP].
Let’s see what happens when we tweak the infinitival subject a little such that the
relativized head has some other case. As shown in (14)b, the result is a perfectly
well-formed relative clause.

13) a. Stad-ı tamir et-mek pahalı-ya maloldu.
stadium-ACC repair do-INF expensive-DAT cost-PST
‘To repair the stadium was costly.’

b. *[[ Ø
1
tamir et-mek] pahalı-ya malol-an] stad
1

repair do-INF expensive-DAT cost-SR stadium-ACC
Intended: ‘the stadium that to repair it was costly’

14) a. [[ köylü-ler-in evler-i-ni] tamir et-mek] pahalı-ya malol-du.
villager-pl-GEN houses-AGR-ACC repair do-INF expensive-DAT cost- PST
‘To repair the villagers’ houses was costly’

a’. [[[ Ø
1
evler-i-ni] tamir et-mek] pahalı-ya malol-an ] köylü-ler
1

houses-AGR-ACC repair do-INF expensive-DAT cost- SR villager-pl
‘the villagers whose houses that to repair (them) was costly’

b. [[ ehir-in sokaklar-ı-nı] tamir et-mek] pahalı-ya malol-du
city-GEN streets-AGR-ACC repair do-INF expensive-DAT cost- PST
‘To repair the streets of the city [lit: city’s streets] was costly’

b’. [[[ Ø
1
sokaklar-ı-nı] tamir et-mek] pahalı-ya malol-an ] ehir
1

streets-AGR-ACC repair do-INF expensive-DAT cost- SR street
‘the city whose streets that to repair (them) was costly’

The same can be accomplished by tweaking example (15). Again changing the
relative head to an expression that does not have accusative case yields an acceptable
relative clause, as in (16).

218

15) a. Problem-i çöz-mek zor-dur.
problem-ACC solve-INF difficult-be-AOR
‘To solve the problem is difficult.’

b. * [[Ø
1
çöz-mek] zor ol-an] problem
1

solve-INF difficult be-SR problem
Intended: ‘the problem which to solve is difficult.’

16) a. [[makine-nin /araba-nın motor-u-nu] çalıtır-mak] zor dur /ol-du
machine-GEN/car-GEN engine-AGR-ACC start-INF difficult be-AOR/be-PST
‘To start the (this) machine’s/car’s engine is/was difficult’

a. [[ Ø
1
motor-u-nu] çalıtır-mak] zor ol-an] makine
1
/araba
1

engine-AGR-ACC start-INF difficult be-SR machine/car
‘the machine/car which to start (its) engine is/was difficult’

b. [[aratırmacı-lar-ın makaleler-i-ni] bastır-mak] kolay gel-di
researcher-pl-GEN papers-AGR-ACC publish-INF easy come-PST
‘To publish the researchers’ papers was easy (lit: came easily)’

b. [[ Ø
1
makaleler-i-ni] bastır-mak] kolay gel-en] aratırmacılar
1

papers-AGR-ACC publish-INF easy come-SR researchers
‘the researchers which to publish (their) papers was easy (lit: came easily)’

Let’s be clear about exactly how we tweaked the good examples above. In previous
chapters, we saw that non-subject SR clauses were acceptable for unaccusative verbs
but not for unergatives (the subject must always be non-specific, of course). We
determined that the unacceptability in unergatives was due to intervention effects
from the subject whose base position we assumed to be in [Spec, vP]. In the
constructions above, 1) the subject could not move to [Spec, TP] because it was an
NP, 2) the NP subject blocked movement of lower expressions from moving to [Spec,
TP], and 3) the accusative direct object in a vP Spec (higher than the NP subject) was
barred from moving to [Spec, TP] because of its case. We saw that embedding in a
scrambled expression circumvents the intervention from the subject. Similarly, by

219

making the accusative object more complex, we saw that it was possible to extract the
specifier of the object. As demonstrated in the tree in (17)c, with possessor-possessee
direct object as in (17)b, the “possessor”, or genitive expression kitap ‘book’ in the
Spec of the accusative object is able to move to [Spec, TP] without intervention
effects from the subject.
176


17) a. [[Ø sayıfalar-ı]-nı kedi parçala-yan] kitab
pages-AGR-ACC cat tear.up-SR book
‘the book whose pages a cat tore up’

b. accusative object prior to extraction of possessor ‘book-GEN’:
[
DP
[
DP
kitab-ın] sayıfalar-ı]-nı
book-GEN pages-AGR-ACC
‘the book’s pages’

c. CP

+Wh-book+GEN
TP C°

+Wh-book+GEN
vP T°

DP
1
+ACC
NP- cat
+Wh-book VP v°
+GEN pages D°
DP
1

tear.up
[book’s pages]





176
Recall that we were able to get around the intervention effects of an NP subject in unergative
constructions in a similar manner: by scrambling a DP/PP around the subject, we were able to A-move
the specifier of the scrambled expression to [Spec, TP], thus triggering the SR form. See Chapters 2,
4, and 5.

220

2 Uninflected Infinitivals

In this section we will be mainly looking at uninflected infinitivals. Here, I will use
the term ‘infinitival’ to refer to a construction containing an infinitive with no
agreement inflection.
177
Now, note that the well-formed infinitival RC examples
(14)a-b and (16)a-b above have the SR form. It is interesting that the NSR form is
unacceptable for all these cases, as their equivalents in (18) demonstrate.

18) a. *[[[ Ø
1
evler-i-ni] tamir et-mey]-in pahalı-ya malol-du-u ] köylü-ler
1

houses-AGR-ACC repair do-INF-GEN expensive-DAT cost- NSR villager-pl
‘the villagers whose houses that to repair (them) was costly’

b. *[[[ Ø
1
sokaklar-ı-nı] tamir et-mey]-in pahalı-ya malol-du-u ] ehir
1

streets-AGR-ACC repair do-INF-GEN expensive-DAT cost- NSR city
‘the city whose streets that to repair (them) was costly’

c. *[[[ Ø
1
motor-u-nu] çalıtır-may]-ın zor ol-du-u] makine
1
/araba
1

engine-AGR-ACC start-INF-GEN difficult be-NSR machine/car
‘the machine/car which to start (its) engine is/was difficult’

d. * [[ Ø
1
makaleler-i-ni] bastır-may]-ın kolay gel-di-i] aratırmacılar
1

papers-AGR-ACC publish-INF-GEN easy come-NSR researchers
‘the researchers which to publish (their) papers was easy (lit: came easily)’

Let’s return to Sezer’s examples (6)-(8) of infinitivals in positions other than subjects
that did not exhibit “island” effects. In all these examples, the infinitival phrases bear
case, except for the infinitival direct object in (6). The matrix verbs in (6)-(8) are
listed in (19). The verbs in (19)b-c select for inherently case-marked “arguments”
178

while the verb iste ‘want’ in (19)a assigns optional accusative case.

177
We will look more closely at extraction from inflected infinitivals later in this chapter. Although
both inflected and uninflected forms are infinitivals, I use the bare term ‘infinitival’ to denote the latter
to reduce wordiness.
178
I use the term “argument” in its standard usage here. (This is different from other Chapters of this
thesis where I used this term to refer strictly to those expressions merged without inherent case. This

221

19) a. iste: Aye ders(-ı) iste-di
want Aye homework(-ACC) want-PST

b. çalı: Aye ders-e çalı-tı
try/work.on Aye homework-DAT work-PST

c. zevk.al: Aye ders-ten zevk al-dı
enjoy (Lit: gain pleasure) Aye homework-ABL enjoy-PST

Now note that the examples in (6) are unacceptable when the infinitival complement
clause is marked with accusative case, as in (20).

20) a. *Bakan [meclis-te konu-may]-ı iste-di
minister parliament-LOC speak-INF-ACC want-PST
‘The minister wanted to speak in the parliament.’

b. *[bakan-ın [Ø
1
konu-may]-ı iste-di-i] meclis
1

minister-GEN speak-INF-ACC want-PST-3S parliament
‘the parliament that the minister wanted to speak (in)’

So, whereas infinitival clauses can be case-marked, for example, (7)b with dative case
and (8)b with ablative case, the facts in examples (18) and (20) lead to the conclusion
that uninflected infinitival clauses do not permit structural case marking. Because
structural case in Turkish is uniformly assigned in a Spec-Head configuration, after
raising to a functional projection, we will assume the same for uninflected infinitivals.
That is, infinitivals as direct objects in transitives do not raise to [Spec, vP], and
infinitivals as sentential subjects do not raise to [Spec, TP]. If they did, they would
be marked with genitive case as subject of RCs. This prohibition against raising is
what forces the SR form for RCs with infinitival subjects and prohibits the NSR form;
[Spec, TP] is left vacant for the +Wh expression to move through. The EPP of T

denotation was for expediency and did not carry any theoretical import except perhaps to illustrate that
we lack a term to refer to those expressions that merge into theta positions without lexical case.)

222

must be satisfied,
179
and in the absence of the (sentential) subject moving to [Spec,
TP], the Wh-expression must move there.
180

The above assumptions have a further implication: we are, in essence,
assuming that uninflected infinitival clauses are NPs: NPs cannot have structural case
and they must remain in their first merge positions (but they allow extraction from
within them). We see the same behavior for uninflected infinitivals, so we will
consider them syntactic NPs.
Let’s review our conclusions thus far for uninflected infinitivals:
Uninflected infinitival clauses are NPs in terms of their syntactic behavior: they do
not move, nor can they be structurally case-marked.
181
Infinitival clauses are not
islands.
Now let’s look at another example from Kornfilt (1997), (21)a from (21)b.

21) a. *[ Ø
1
yüz-mek] güzel ol-an] deniz
1

swim-INF nice be-SR sea
Intended: ‘the sea which to swim (in) is nice’

b. [ deniz-de yüz-mek] güzel.dir
sea-LOC swim-INF nice.be-AOR
‘It’s nice to swim in the sea’


We said that the reason extraction from the infinitival sentential subject in (4) and (5)
was bad was because the Wh-element has accusative case which was barred from

179
I am assuming that T has an EPP feature when it is selected by C. In a sentence with an infinitival
sentential subject with no relativization, I assume no CP projection; T in this sentence will have no
EPP feature.
180
The reader must already have several questions in mind regarding the nature of these infinitival
phrases and case-marking. I ask for the reader’s patience as I try to present the issues one by one.
181
Note that in example (15)a, the verb is in the aorist tense which must be used for generic subjects. I
assume that the aorist either has no T projection or has a defective T, one that neither assigns case nor
has an EPP feature. This is consistent with the subject in this example being an NP.

223

moving to [Spec, TP]. But in (21), the relative head has inherent locative case prior
to extraction. We have seen many examples where locative extraction using the SR
form was fine. The examples in (22)a and (23)a are also bad, even though the
relativized expression is marked with inherent case in each.

22) a. *[[ Ø
1
holan-mak] zor ol-an] kız
1

like-INF difficult be-SR girl
Intended: ‘the girl who to like is difficult’

b. [(bu) kız-dan holan-mak] zor-dur
(this) girl-ABL like-INF difficult be-AOR
‘It’s difficult to like this girl’ ‘(Lit.) It’s difficult to feel good from this girl’

23) a. *[[ Ø
1
git-mek] kolay ol-ma-yan] ehir
1

go-INF easy be-neg-SR town
Intended: ‘the town that to go to is not easy’

b. [(o) ehir-e git-mek] kolay deil
(that) town-DAT go-INF easy be-neg
‘It’s not easy to go to that town’

Again, the problem seems to be an intervention effect because if we embed the Wh-
expression in a PP and scramble the PP, we derive an acceptable RC, as in (24) which
is minimally different from (21).

24) [[ Ø
1
için-de] yüz-mek] güzel ol-an] deniz
1

in-LOC swim-INF nice be-SR sea
‘the sea which to swim (in) is nice’

Let’s look at this construction a little closer. The structure of the PP deniz-in için-de
‘in the sea’ is shown in (25)a with tree in (25)b.


224

25) a. [
DP
[
DP
deniz]-in [
NP
iç-in] D° ]-de
sea-GEN inside-AGR-LOC
‘(Lit.) at the sea’s inside’

b. DP+LOC

sea+GEN
NP D°

DP N°
sea inside+AGR



Now let’s take a look at the structure of the infinitival clause, as shown in (26).
182


26) [PRO deniz-de yüz-mek] = [PRO [in the sea] to swim]

CP/NP


TP C°/N°

PRO
VP T°

DP/PP V°
swim
[in the sea]


Again, recall that in our story about the SR form, we saw intervention effects from
the subject in unergatives but not in unaccusatives.
183
We were able to circumvent
the intervention effects in unergative constructions by embedding the Wh-expression
in a PP and scrambling the PP above the in-situ subject in [Spec, vP]. This is shown
in (27). In the bad (27)a, the subject blocks the raising of the locative Wh-expression

182
I have nothing to say just yet as to whether infinitivals have a CP layer or are only TPs. The issue is
orthogonal to what is being discussed and will make no difference to the discussion at this point.
183
By subject, I mean the argument merged into the outermost theta position (not assigned lexical
case).

225

‘couch’ to [Spec, TP]. However, when ‘couch’ is embedded in a complex DP/PP,
this DP/PP can scramble around the subject and ‘couch’ can move to [Spec, TP], and
then to [Spec, CP] triggering the SR form.

27) a. *[bayan Ø
1
otur-an] kanepe
1

woman sit-SR couch
Intended: ‘the couch that a woman is sitting on’

b. [[Ø
1
üst-ün]-de bayan otur-an] kanepe
1

top-AGR-LOC woman sit-SR couch
‘the couch that a woman is sitting on (top of)’
[Lit: ‘the couch whose top a woman is sitting on’]

I have argued elsewhere that scrambled elements may only scramble once after which
they are frozen for further movement.
184
Let us assume this is correct. The result is
that in order to get around an intervening subject, a Wh-expression may not scramble
because it will become frozen in the scrambled position. However, a Wh-expression
embedded in the specifier of a DP, can move with the DP as it scrambles around the
subject, after which the Wh-expression can move to a functional projection.
Importantly, the Wh-expression must be in the specifier of the DP so that the DP and
the Wh-element will be equidistant from the target.
185
Returning to the example in

184
Although scrambling is beyond the scope of this work, I have speculated that the reason for this
“freeze” after one scrambling is due to Recoverability. It seems to me Feature checking leaves a
“trail” of sorts, but scrambling, as far as I can see, does not check any features; certainly, it does not
check case or EPP features (see Chapters 2 and 4). So, you’re allowed “one free move”. If a move is
possible (i.e. does not violate minimality, PF and LF conditions), a DP can take it, and the (LF)
interface component can reconstruct back one possible move, but that’s all. Reasoning through this
possibility is not feasible within the confines of this work, but I mention it for future research. But see
fn. 26 in Chapter 4 where I offer other possible explanations.
185
Equidistance is computed in terms of Chomksy’s (2000) Defective Intervention Constraint, (i).
(i) Defective Intervention Constraint (DIC):
In the structure, > > , where > is c-command, and and match
probe , but is inactive, the effects of matching are blocked.
I have adopted this for all movement in the sense that if there is nothing that c-commands but not ,
then and are equidistant from a higher probe. This definition resolves two problems: not only does

226

(24), let’s revisit what has to happen inside the infinitival phrase, as shown in (28).
Here the complex DP/PP scrambles above PRO (which, for the moment, I have
located in [Spec, TP]) and, I assume, adjoins to TP.
186


28) [ [[[
DP
[
DP
deniz-in] iç-in] D° ]-de] PRO [
DP/PP
deniz-in iç-in-de] yüz-mek]
sea-GEN inside-AGR-LOC swim-INF
‘to swim in the sea’ ‘(lit) in the sea’s inside PRO to swim’

CP/NP


TP C°/N°

DP/PP+LOC
PRO
VP T°

DP/PP V°
swim
[sea’s inside]-LOC


2.1 Review of assumptions

Let’s review what we have determined so far. First my assumptions:
1- An uninflected infinitival clause is an NP and remains in situ.

it define what is and is not an intervener, it also provides us with a measure of Economy. When two
expressions are “equidistant” in terms of the above description, the movement of one expression is not
more “economical” than movement of the other. Richards (2005) also uses such a measure of
equidistance in explaining movement in Tagalog.
186
I make the assumption that this is an adjunction site for three reasons. First, nothing in the data
throughout this work suggests that Turkish has multiple specifier positions. Quite the contrary, unique
specifier positions are often the culprit in moving prohibitions. Second, this movement freezes the
expression. There is evidence that adjuncts, though porous for movement from within them, are frozen
for A-movement. Finally, accusative objects can A-scramble above the subject in [Spec, TP].
Whereas movement to the specifier of T is clearly prohibited for accusatives, one can make the case
that adjunction to T is not a case assigning position, and this is the reason accusative objects can
scramble/adjoin to T.

227

2- (Because of 1) [Spec, TP] of the RC of which the infinitival is the subject
remains vacant.
3- (Because of 2) the relativized expression must move from within the infinitive
clause to [Spec, TP] of the RC, to check T’s EPP feature.

With these assumptions in mind, the solution to the infinitival sentential subject
puzzle then, is as follows:
1- An accusative direct object of the infinitival sentential subject may not be the
relative head of the RC because an accusative expression is barred from
moving to [Spec, TP].
2- Inherently case-marked expressions can be the relative head, except that they
are lower than PRO in the infinitival clause.
3- An expression may A-scramble around PRO, in which case, the expression in
its Spec can be the relative head. This expression will move out of the
scrambled DP in the infinitival phrase to the RC [Spec, TP], triggering the SR.

Let’s revisit other bad examples to see that we can indeed get around the intervention
effects. Repeating (22)a as (29)a, note that by embedding the relative head kız ‘girl’
in a larger DP, I can now relativize out of the infinitival subject, as in (29)b.

29) a. *[[ Ø
1
holan-mak] zor ol-an] kız
1

like-INF difficult be-SR girl
Intended: ‘the girl who to like is difficult’

b. [[ [
DP
Ø
1
akalar-ın]-dan holan-mak] zor ol-an] kız
1

jokes-AGR-ABL like-INF difficult be-SR girl
‘the girl whose jokes which to like is difficult’


228

Likewise for (23)a repeated as (30)a. By making the Wh-expression ehir ‘city’ a
possessor of the DP [city’s neighborhoods], it is possible to scramble this complex
DP around PRO, and then extract ‘city’ thereby avoiding the intervention from PRO.

30) a. *[[ Ø
1
git-mek] kolay ol-ma-yan] ehir
1

go-INF easy be-neg-SR town
Intended: ‘the town that to go to is not easy’

b [[ [
DP
Ø
1
mahaleler-i]-ne git-mek] kolay ol-ma-yan] ehir
1

neighborhoods-AGR-DAT go-INF easy be-neg-SR town
‘the town whose neighborhoods that to go to is not easy’

If this analysis is on the right track, we must conclude that, like the unergative
subject, PRO is also an intervener. Extraction of the locative Wh-DP was not
possible unless we embedded it, and left behind a remnant. We must conclude that in
such a derivation, we are circumventing intervention effects for A-movement within
the infinitival. Recall that such a strategy is not needed for A-bar movement which
can be long-distance. PRO blocks the A-movement of a lower expression. When that
expression is a constituent of a larger element that scrambles around PRO, it evades
intervention by PRO and is free to A-move.
2.2 Non-subject infinitivals

Let’s now turn to non-subject infinitivals beginning with item (6) repeated as (31).

31) a. Bakan [meclis-te konu-mak] iste-di
minister parliament-LOC speak-INF want-PST
‘The minister wanted to speak in the parliament.’

b. [bakan-ın [Ø
1
konu-mak] iste-di-i] meclis
1

minister-GEN speak-INF want-NSR-3S parliament
‘the parliament that the minister wanted to speak (in)’

229

Note in the RC in (31)b that the relative head has inherent locative case prior to
movement, and yet no embedding and scrambling was necessary. Extraction of a
locative from a complement infinitival clause is perfectly acceptable. There seem to
be no intervention effects from PRO in this construction. Why would that be? Notice
that in this example, we have obligatory control (OC) PRO. Could it be that arbitrary
PRO creates intervention effects and OC PRO does not? Let’s look at another
example.
In (9) repeated as (32), we see extraction of a dative from an adjunct
infinitival clause. Again, this is a case of OC PRO. I assume that the dative is
generated lower than PRO, and yet it can move past the PRO without the intervention
effects we saw with non-OC PRO.

32) a. Ali [[ PRO
OC
okul-a gir-mek] için] on yıl ura-tı
Ali school-DAT enroll-INF in.order.to ten year struggle-pst
‘Ali tried for ten years in order to get into that school’

b. [Ali-nin [[PRO
OC
Ø
1
gir-mek] için] on yıl ura-tı-ı] okul
1

Ali-GEN enroll-INF in.order.to ten year struggle-NSR-3s school
‘the school that Ali tried for ten years in order to get into’

This suggests that arbitrary PRO and OC PRO are different animals. In fact, it lends
support to the Hornstein (1999) proposal that OC PRO is derived via movement. If
this is indeed the case, then the consequences are such that OC PRO is a residue of
movement leaving behind either nothing (except relevant features it has checked on

230

functional heads)
187
or a trace which does not serve as an intervener for A-movement
from its c-commanding domain.
188

However, things are not so simple. In example (33)a, note that we again have
an instance of OC PRO except that it is object control. In stark contrast to its English
counterpart in (33)b, the Turkish example is unacceptable.

33) a. *Ahmet Aye-ye [ Ø
1
oku-mak] bir kitap
1
] al-dı.
Ahmet Aye-DAT read-INF one book buy-PST
Intended: ‘Ahmet bought Aye a book to read’

b. John bought Mary [a book [ to read Ø ] ].


Let’s look closer at the structure of (33)a and perhaps we can account for its
unacceptability. First, note that the infinitival in this sentence is not an argument.
The matrix direct object is kitap ‘book’. Furthermore, (in spite of the English gloss)
there is no relative clause the structure of (33)a.
189
Wh-movement in Turkish, or
more specifically, movement to the CP domain, is possible only in relative clauses,
sluicing structures and for Topic movement. So the question is how did ‘book’
become the relative head? If it’s just a matter of ‘book’ raising out of the infinitival,
then it is on the wrong side of ‘to read’, as specifiers are leftward.

187
By this I mean that the history of the derivation is “accessible”, i.e. features have been checked or
deleted so that there is no crash at the interface levels.
188
We are at this point still looking at intervention effects with the infinitival itself. As demonstrated
by example (34)/(35) the controller, i.e. the antecedent, does not seem to be an intervener.
189
First, the infinitival lacks the verbal morphology of an RC. Second, recall that all RCs require that
the EPP of T be checked. In the infinitival in (33)a neither the PRO subject nor the accusative direct
object can check T’s EPP.

231

Extraction out of Object Control PRO infinitives seems to be possible, as
shown by (34), where the infinitive is an argument (the direct object) of the
embedded RC verb ‘want’.

34) Ahmet Aye-ye [[Ø
1
oku-mak] iste-di-i] kitab
1
]-ı al-dı.
Ahmet Aye-DAT read-INF want-NSR-3s book-ACC buy-PST
‘Ahmet bought Aye the book he/she wanted to read’

Notice though that in (34) the subject of the RC can be either Ahmet or Aye. The
structure of (34) then, is really as in (35) where the RC subject is a null pronoun
which can take Ahmet or Aye as an antecedent. So, again, we are back to a case of
Subject Control for the PRO of the infinitival, which picks up its referent depending
on the referent of the subject of the RC ‘[the book that he/she wants to read]’.

35) Ahmet
1
Aye
2
-ye [pro
1/2
[PRO
1/2
Ø
3
oku-mak] iste-di-i] kitab
3
]-ı] al-dı.
Ahmet Aye-DAT read-INF want-NSR-3s book-ACC buy-PST
‘Ahmet bought Aye the book he/she wanted to read’

In (35), ‘book’ moves out of the infinitival phrase without intervention from the OC
PRO subject. The question is what position does the Wh-book land in in the
embedded relative clause? I assume that ‘book’ is marked with accusative case inside
the infinitive, i.e. it receives accusative case from the v° of ‘read’ in the infinitival
phrase. Let us assume for the moment (although we will see arguments for it later)
that elements moving out of an infinitival do not A-bar move.
190
I come to this
conclusion because we saw intervention effects from PRO which would not have
been possible if the expression could A-bar move. The accusative ‘book’ must move

190
This could mean that either there is no CP projection or that the CP layer has no specifiers, i.e. no
features, EPP or uWh, to check via phrasal movement.

232

directly to [Spec, vP] of the RC verb ‘want’, which is unoccupied because the
infinitival clause, being an NP, cannot move there. Although this is an instance of an
accusative expression moving to an accusative assigning position, this move is
presumably possible because structural case is evaluated at PF; there is no issue of
case mismatch. I suggest that this is what happens with possessor DPs moving to
[Spec, TP]. Because a possessor DP must raise to the [Spec, DP] of the possessee D°,
I assume that the genitive assigned by D° is also a structural case. We had
determined that only inherently case-marked elements may move to a structural case
assigning position. But, an embedded possessor genitive, assigned structural case by
D°, can move to [Spec, TP] of a subordinate clause because the case in that position
is morphologically identical; embedded T assigns “genitive” case. Returning to the
movement of ‘book’ in (35), it seems clear that [Spec, vP] is the only position
available, as [Spec, TP] is taken up by the RC subject. From the RC [Spec, vP], the
relative head ‘book’ A-bar moves to the RC [Spec, CP], after which it is promoted to
the external head position.
Let’s look at another example, the unacceptable (36). What is the difference
between (34) and (36) that makes one bad and the other good? Note what happens
when we change the RC verb to ‘want’, as in (37). The relative clause becomes
acceptable. The answer here lies in the verb. The clausal complement of the verb
‘tell’ must obligatorily be a DP. As shown in (38), the verb ‘tell’ takes a case-marked
infinitival complement, but, crucially, one that must have agreement inflection.
191

These are the inflected infinitivals that we will look at in more detail in Part 2.

191
Perhaps this is because the “thing” you utter is referential; it cannot be non-specific. It seems that
even in English, the only way to denote a non-specific, existential utterance as the complement of said

233

36) *Ahmed-in Aye-ye [ Ø
1
oku-mak] söyle-di-i kitap
1

Ahmet-GEN Aye-DAT read-INF tell-NSR-3s book
Intended: ‘the book John told Mary [ PRO to read]’

37) [Ahmed-in [ Ø
1
oku-mak] iste-di-i] kitap
1

Ahmet-GEN read-INF want-NSR-3s book
‘the book Ahmet wanted [ PRO to read]’

38) Ahmet Aye-ye [ kitab-ı oku-ma-*(-sı)]*(-nı) söyle-di
Ahmet Aye-DAT book-ACC read-INF -3s -ACC tell-PST
Ahmet told Aye to read the/this book’

Kornfilt (1997) points out that the choice of complement type, i.e. a bare infinitive, an
inflected infinitive or a “Factive”
192
NSR clause, is determined by the verb. For
example, as shown in (39), the verb kork ‘fear’ allows for all three types of
complements.

39) a. (ben) [PRO öl-mek]-ten kork-uyor-du-m
I die-INF ABL fear-PROG-PST-1s
‘I was afraid of dying’ ‘(Lit.) I was afraid to die’

b. (ben) [Ahmed-in öl-me-sin]-den kork-uyor-du-m
I Ahmet-GEN die-INF-3s-ABL fear-PROG-PST-1s
‘I was afraid that Ahmet had died’ ‘(Lit.) I was afraid Ahmet to have died’

c. (ben) [Ahmed-in öl-dü-ün]-den kork-uyor-du-m
I Ahmet-GEN die-INF-NSR-3s-ABL fear-PROG-PST-1s
‘I was afraid that Ahmet had died’
‘(Lit.) I was afraid Ahmet to have died’ (Kornfilt: p.51)


is through the use of passive voice. Compare (i) where something is indefinite but specific, with (ii)
where something is indefinite and non-specific.
(i) John said something.
(ii) There was something said.
192
The term “Factive” is used in the literature to refer to NSR –DIK complement clauses as opposed to
“Active” infinitival complement clauses. These labels carry no theoretical implications for the purpose
of this work.

234

Note, however, that the verb kork ‘fear’ takes an inherently case-marked argument, in
this case an ablative. We saw elsewhere that inherently case-marked elements are
ambiguously specific or non-specific. For us this means, that they may be either DPs
or NPs. If we are right in assuming that the bare infinitive is an NP, and the inflected
infinitive is a DP, then the ablative structures of (39)a and (39)b would be as in (40)a
and (40)b, respectively. That is, the bare infinitive in (39)a is an NP.

40) a. [
NP
PRO öl-mek]-ten
die-INF ABL

b. [
DP
Ahmed-in öl-me-sin]-den
Ahmet-GEN die-INF-3s-ABL

While we are looking at these structures, let’s take a moment to remark that the
subject of the inflected infinitival in (40)b bears genitive case. This is expected
because we are assuming that D° assigns genitive case to the expression in its
specifier. The approximate structure of 40)b is shown in (41) where what is
important is that the subject Ahmet must have raised to a structural case-assigning
position if it bears overt genitive case.
193


41) [
DP
Ahmed-in öl-me-sin]-den
Ahmet-GEN die-INF-3s-ABL

DP/TP?-ABL

Ahmet-GEN
VP D°/T°?

Ahmet V°
die



193
See fn. 196.

235

The point of these examples is to demonstrate that just as with nominals, one can use
case-marking to determine the category, but it must be structural case, either
accusative or the embedded genitive. Structural case can only appear on DPs and
direct object DPs of accusative transitives must bear overt accusative case.
Interestingly, OC PRO direct object clauses are always bare whereas the
inflected clauses must have accusative case, as in (42). This also complies with the
movement account of OC PRO because N° does not assign case, but D° does. If we
do not assume a special case for PRO, we can conclude that the subject of die in (42)a
raises for case reasons, whereas no such raising is necessary for the subject of the DP
infinitival in (42)b.

42) a. (ben) [PRO öl-mek](*-i) iste-mi-yor-um
I die-INF want-NEG-PROG-is
‘I don’t want to die’

b. (ben) [Ahmed-in öl-me-sin]*(-i) iste-mi-yor-um
I Ahmet-GEN die-INF-3s-ACC want-NEG-PROG-is
‘I don’t want Ahmet to die’

In example (43), the infinitive bears inherent dative case, and because it is
uninflected, we are assuming it is an NP. In this example, we have an instance of
object control. Relativization of the ablative from within the infinitival is not
possible, (44). Is this due to intervention effects from PRO, as we saw earlier? That
is, the ablative is presumably lower in the structure than PRO. Let’s change the
infinitival verb to a transitive, like “to read” for example. By providing the infinitive
a direct object that must raise to get accusative case, we can skirt PRO (which I am

236

assuming remains in its base position in [Spec, vP]) using the “free-rider” strategy we
saw in previous chapters.

43) (o) Ahmed-i [sınıf-tan kaç-ma]-a zorla-dı
he Ahmet-ACC class-ABL run.away-INF-DAT compel-PST
‘He compelled Ahmet to run away from class’

44) *[ pro Ahmed-i [ Ø
1
kaç-ma]-a zorla-dı-ı] sınıf
1

Ahmet-ACC run.away-INF-DAT compel-NSR-3s class
‘the class he compelled Ahmet to run away from’

With this change in the infinitival, we see that an accusative object can indeed be
relativized, as in (45).
194


45) [ pro Ahmed-i [ Ø
1
okumay]-a zorla-dı-ı] kitap
1

Ahmet-ACC read-INF-DAT compel-NSR-3s book
‘the book he compelled Ahmet to read’

Let’s try another structure, this time trying to relativize an instrumental. Again, the
result in (46)b is unacceptable. We must assume this is because, unlike the accusative
direct object in a specifier of vP higher than PRO, an instrumental is generated in a
position lower than PRO.
195
PRO is an intervener.


194
I am for the moment ignoring the issue of T’s EPP feature in the infinitival; we saw that an
accusative DP cannot move to [Spec, TP], and therefore T’s EPP would not be satisfied in 45). I ask
for the reader’s patience as this issue will be resolved in Section 3.
195
In the sentence in (i), the antecedent of his in the instrumental phrase [his ball] cannot be Ahmet.
Note that the non-specific subject, dolphin, is in situ in [Spec, vP]. The locative has raised to [Spec,
TP] to satisfy T’s EPP feature. The instrumental must be lower than [Spec, vP] in order to allow the
binding of the pronoun by the subject.
(i) [Ahmed-in
1
havuz-un
1
]-da bir yunus.balıı
2
[top-u
*1/2
]-ile oynu-yor.
Ahmed-GEN pool-3sAGR-LOC one porpoise ball-3sAGR-INST play-PRES
‘In Ahmet’s
1
pool, a dolphin
2
is playing with his
*1/2
ball’

237

46) a. (o) Ahmed-i [ PRO bu kaık-la ilac al-may]-a zorla-dı
he Ahmet-ACC this spoon-INST medicine take-INF-DAT compel-PST
‘He compelled Ahmet to take medicine with this spoon’

b. *[(o-nun) Ahmed-i [PRO Ø
1
ilac al-may]-a zorla-dı-ı] kaık
1

he-GEN Ahmet-ACC medicine take-INF-DAT compel-NSR-3s spoon
Intended : ‘the spoon that he compelled Ahmet to take medicine (with)’

Object control PRO seems to pattern with arbitrary PRO in that it induces
intervention effects. Extraction from these infinitival clauses is possible, but only
when the expression is structurally higher than PRO. This is demonstrated in the tree
in (47) for the ‘[PRO (to) read book-ACC]’ we saw in (45). In (47) (where we are, for
the moment, only concerned with the derivation up to the vP projection), the direct
object merges in V and gets its theta-role, PRO picks up its theta-role when it merges
in [Spec, vP], and the direct object raises to a higher Spec of vP to be assigned
accusative case. PRO is no longer an intervener for movement of the direct object.

47) [ pro Ahmed-i [ Ø
1
okumay]-a zorla-dı-ı] kitap
1

Ahmet-ACC read-INF-DAT compel-NSR-3s book
‘the book he compelled Ahmet to read’

XP?/NP


vP X?°/N°

DP-book+ACC
(+Wh) PRO
VP v°

DP-book V°
(+Wh) read


Here’s an even better way to prove the point about the intervention from PRO: by
“tweaking” the unacceptable (46)b. Recall that we can scramble over PRO, but then

238

the scrambled expression is “frozen”, but its specifier is free to move. Let’s make the
instrumental kaık-la “spoon-INST” in (46)b a larger DP by giving it a +Wh possessor.
The new sentence appears as (48)a, with the corresponding relative clause in (48)b. I
have changed the matrix (RC) subject and direct object to 1
st
and 2
nd
person pronouns
to avoid any ambiguity as to the possessor of ‘spoon’ which has 3
rd
person singular
agreement. Notice that the relative clause is now good! Why? Because the +Wh-
hospital was able to ride parasitically as a constituent of a scrambled DP as it moved
around (and higher than) PRO, and then move again into the matrix RC clause.

48) a. Ben sen-i [[
DP
hastane-nin kaı-ı]-la
1
PRO t
1
ilac al-may]-a
I you-ACC hospital-GEN spoon-3s-INST medicine take-INF-DAT
zorla-dı-m
compel-PST1S
‘I compelled you to take medicine with the hospital’s spoon’

b. [ben-im sen-i [[
DP
Ø
2
kaı-ı]-la
1
PRO t
1
ilac al-may]-a
I-GEN you-ACC spoon-3sAGR-INST medicine take-INF-DAT
zorla-dı-ım] hastane
2

compel-NSR-1s hospital
‘the hospital whose spoon I compelled you to take medicine (with)’


Another non-subject control verb is tavsiye etmek ‘to recommend’ where the
controller is the indirect object. Extraction of an accusative object is not possible, as
in (49)b. Neither is extraction of the dative expression, ada ‘island’ in (50)b. But, as
expected, the possessor of a direct object of the infinitival can be relativized, as in
(51)b. Again, the possessor was a “free rider” as the direct object raised over PRO
for case.


239

49) a. Biz-e [her gün erbet-i iç-me]-ye tavsiye et-ti
us-DAT every day syrup-ACC drink-INF-DAT recommend do-PST
‘He recommended to us to drink the syrup every day’

b. *[biz-e [her gün Ø
1
iç-me]-ye tavsiye et-ti-i] erbet
1

us-DAT every day drink-INF-DAT recommend do-NSR-3s syrup
Intended: ‘ the syrup that he recommended to us to drink every day’

50) a. Biz-e [o ada-ya git-me]-ye tavsiye et-ti
us-DAT that island-DAT go-INF-DAT recommend do-PST
‘He recommended to us to visit (go to) that island’

b. *[bize [Ø
1
git-me]-ye tavsiye et-ti-i] ada
1

us-DAT go-INF-DAT recommend do-NSR-3s island
Intended: ‘the island that he recommended to us to visit (go to)’

51) a. pro Ahmed-e [[
DP
aratırmacı-nın makale-si]-ni oku-may]-a
(he) Ahmet-DAT researcher-GEN paper-AGR-ACC read-INF-ACC
tavsiye et-ti
recommend do-PST
‘He recommended to Ahmet to read the researcher’s paper’

b. [ pro Ahmed-e [[
DP
Ø
1
makale-si]-ni oku-may]-a tavsiye et-ti-i]
(he) Ahmet-DAT paper-AGR-ACC read-INF-ACC recommend do-NSR
aratırmacı
1

researcher
‘The researcher whose paper he recommended to Ahmet to read’


So what can we conclude? The examples in (48) and (51) demonstrate that
movement from within the infinitive is possible, but only when the expression is
above PRO. This is evidence that PRO that is controlled by an indirect object is an
intervener. Extraction is possible only when the RC head is a “free rider” on an
expression that moves around the PRO subject. As for (47), I will tentatively suggest
that the accusative direct object of the infinitive is too far away for A-bar movement
and has no available intervening A-position to move to. Remember that we are

240

assuming that movement (excluding scrambling) is driven by the EPP. The verb
recommend takes a dative argument which means that its v° does not assign case and
does not have an EPP feature. Whereas the T° of recommend does have an EPP
feature, the subject, he, must move to [Spec, TP] to be assigned case. The example in
(47)b is an argument for phases. By the time matrix (RC) C° merges with the
structure, the most deeply embedded vP, that of the infinitival has been spelt-out and
is no longer accessible for even long distance A-bar movement. What is remarkable
though is that successive cyclic A-bar movement does not seem to be available to
save this derivation; otherwise, we would not have seen the contrast between (46)b
and (48)b. Movement to a CP projection of the infinitival clause is not an option.
Let’s return to the examples of subject OC PRO in (7) and (8) repeated as (52)
and (53). In these examples, movement out of the infinitival is good. In the RCs in
(52)b and (53)b, the subject is overt and has genitive case. This means that [Spec,
TP] is occupied. Again, the accusative object of the infinitive must have moved first
to [Spec, vP] of the RC, and then to [Spec, CP]. As noted above, A-bar movement
through a [Spec, CP] of the infinitive clause doesn’t seem to be an option, and it is
not possible for the direct object to move in one fell swoop from the infinitive [Spec,
vP] to the matrix (RC) [Spec, CP].

52) a. yeni idare [kitaplar-ı yasakla-ma]-ya çalı-ıyor
new administration books-ACC ban-INF-DAT try-pres
‘The new administration is trying to ban (certain) books.’

b. [yeni idare-nin [Ø
1
yasakla-ma]-ya çalı-tı-ı ]] kitaplar
1

new administration-GEN ban-INF-DAT try-NSR-3s books
‘the books that the new administration is trying to ban’


241

53) a. Ali [kitab-ı oku-mak]-tan zevk al-ıyor.
Ali book-ACC read-INF-ABL pleasure take-aor
‘Ali gets pleasure (i.e. enjoys) reading the book.’

b. [Ali-nin [Ø
1
okumak]-tan zevk al-dı-ı]] kitap
1

Ali-gen read-INF-ABL pleasure take-NSR-3s book
‘the book that Ali enjoys reading’ (lit: ‘the book Ali gets pleasure from (to) read’)



3 Inflected infinitivals


It was argued above that the bare uninflected infinitival –mak behaves like an NP.
Kornfilt (1997) also points this out. Though Turkish complement clauses can
generally be separated from the matrix verb by a number of elements, as in (54)b, the
bare infinitival complement does not permit any element between itself and the verb,
as in (55)b. This prohibition, however, does not hold for inflected infinitival
complements, which are case-marked and permit elements between them and the
verb, (56)b. Kornfilt notes that “the infinitive not marked for case behaves like
incorporated nouns which are non-specific and are not marked for case” (p. 407).

54) a. Hasan Ali-ye [Aye-nin yarı-ı kazan-dı-ın]-ı söyle-di
Hasan Ali-DAT Aye-GEN race-ACC win-NSR-3s-ACC tell-PST
‘Hasan told Ali that Aye won the race’

b. [Aye-nin yarı-ı kazan-dı-ın]-ı Ali-ye HASAN söyle-di
Aye-GEN race-ACC win-NSR-3s-ACC Ali-DAT HASAN tell-PST
‘HASAN (focus) told Ali that Aye won the race’

55) a. Hasan
1

1
yarı-ı kazan-mak] isti-yor
Hasan race-ACC win-INF want-PRES.PROG.
‘Hasan wants to win the race’

242


b. *[Ø
1
yarı-ı kazan-mak] HASAN
1
isti-yor
race-acc win-INF HASAN want-PRES.PROG.
‘HASAN (focus) wants to win the race’

56) a. Hasan Ali-ye [Aye-nin proje-yi bitir-me-sin]-i söyle-di
Hasan Ali-DAT Aye-GEN project-ACC finish-INF-3s-ACC tell-PST
‘Hasan told Ali that Aye should finish the project’

b. [Aye-nin proje-yi bitir-me-sin]-ı Ali-ye HASAN söyle-di
Aye-GEN project-ACC finish-INF-3s-ACC Ali-DAT HASAN tell-PST
‘HASAN (focus) told Ali that Aye should finish the project’

In our framework, and completely in line with what we have seen for non-clausal
arguments, the non-case-marked infinitival is an NP and must remain in-situ, whereas
the inflected infinitival is a DP (i.e. behaves syntactically as a DP), and must raise for
case.
It is this difference which explains the contrast in extraction from sentential
subjects. Let’s look at Sezer’s examples again repeating the minimal pairs in (11)
and (12), as (57) and (58). We have concluded that the structure of these infinitival
subjects is different, with the uninflected infinitival being an NP and the inflected
infinitival being a DP. Actually, these examples aren’t minimal pairs at all, as (57) is
a passivized infinitive. Let’s look at a more evenly matched pair, such as those in
(59) and (60).

57) a. [[
DP
Kitab-ın yazıl-ma-sı] Ali-ye bebin liraya otur-du
book-GEN write-PASS-INF-3s Ali-DAT 5000 lira come-PST
‘The writing of (this) book cost Ali 5000 lira.’

b. [[
DP
Ø
1
yazıl-ma-sı] Ali-ye bebin liraya otur-an] kitap
1

write-PASS-INF-3s Ali-DAT 5000 lira come-SR book
‘the book that its writing cost Ali 5000 lira’


243

58) a. [[
NP
Kitab-ı yaz-mak] Ali-ye bebin liraya otur-du
book-ACC write-INF Ali-DAT 5000 lira come-PST
‘The writing of (this) book cost Ali 5000 lira.’

b. *[[
NP
Ø
1
yaz-mak] Ali-ye bebin liraya otur-an] kitap
1

write- INF Ali-DAT 5000 lira come-SR book

In our new examples, (59) and (60), the verbs are comparable except that one is an
inflected infinitive with a genitive subject (59), and the other is a bare infinitive, (60).
Since we have committed ourselves to treating these infinitival phrases as nominals,
let’s make comparable assumptions about the internal structure of these phrases. That
is, just as we had assumed that a bare NP does not have a case-assigning DP
projection above it (although there may be other projections that would normally be
between DP and NP), let’s assume that the bare infinitival also lacks a structural case-
assigning projection, i.e. a TP projection.
196
Now, let’s assume that PRO in (60) does
not need case. This is not so far a stretch seeing as how the generic PRO in these
sentential subject environments is non-specific, and non-referential anyway. Thus,
PRO in (60) remains in its base-generated theta position. With these assumptions, the
structure of the infinitival in (60) is as in (61).

59) a. [[
DP
Ali-nin sokaklar-ı temizle-me-si] belediye-ye bebin dolar-a ol-du
A.-GEN streets-ACC clean-INF-3s municipality-DAT 5000 dollar-DAT come-PST
‘[(For) Ali to clean the streets] cost the municipality 5000 dollars.’

b. [[
DP
Ali-nin Ø
1
temizle-me-si] belediye-ye
Ali-GEN clean-INF-3s municipality-DAT
bebin dolar-a ol-an ] sokaklar
1

5000 dollar-DAT come-SR streets
‘the streets which (for) Ali to clean cost the municipality 5000 dollars.’


196
To be more precise, the label on these phrases is for ease of exposition only. My point is that the
infinitive in (60)/(61) lacks a case-assigning functional projection above the vP. I have nothing to say
about the category of the head of such a projection and have referred to it as D°/T°.

244

60) a. [[
NP
Sokaklar-ı temizle-mek] belediye-ye bebin dolar-a otur-du
streets-ACC clean-INF municipality-DAT 5000 dollar-DAT come-PST
‘[To clean the streets] cost the municipality 5000 dollars.’

b. *[[
NP
Ø
1
temizle-mek] belediye-ye bebin dolar-a otur-an] sokaklar
1

clean-INF municipality-DAT 5000 dollar-DAT come-SR streets

61) [
NP
sokaklar-ı temizle-mek]
streets-ACC clean-INF


CP(?)

vP C°

streets+ACC
(NP)-PRO
VP v°

streets V°
clean


Again, I have nothing to say at this point, about the CP projection, except to say that
it does not seem to have an available Spec position. I do assume that, as with all
transitive subjects, PRO first-merges in [Spec, vP], after which the direct object
streets raises to a higher vP Spec to receive accusative case. As we saw, extraction of
a genitive of the direct object was possible because there was no intervener between
the relative head and the matrix clause.
In contrast to (60), the infinitive in (59) is DP-like, and based on our new
assumptions, this means that there is an external structural case assigning head in this
clause. The structure of the infinitive subject in (59) is shown in (62). In this
structure, the subject raises to [Spec, TP] and is assigned genitive case.



245

62) [
DP
Ali-nin sokaklar-ı temizle-me-si]
Ali-GEN streets-ACC clean-INF-3s

CP(?)

DP?/TP C°

Ali+GEN
vP D°?/T°

streets+ACC
Ali
VP v°

streets V°
clean





4 Taking stock

So now the question is how is it that whereas both structures are porous for
extraction, only the NP-like bare infinitive exhibits A-movement intervention effects
from PRO. We saw no such effects from the DP-like inflected infinitive.
Let’s readdress our assumptions about the structure of these two clauses. It
would be preferable, of course, if the two could be as similar as possible. One option
would be to postulate that the bare infinitive, like its inflected counterpart, also has a
TP projection. We are assuming throughout that T° does not need to assign case. To
account for the difference in raising of nominals within the infinitivals, we would
have to stipulate that the TP projection is such that no DP may raise above it unless
its EPP feature has been checked. Thus, when there is a non-PRO subject in [Spec,
TP], other expressions may A-bar move to the infinitival [Spec, CP]. In the absence

246

of [Spec, TP] being occupied, a Wh-expression must first move to [Spec, TP] before
it can move to the CP layer.
This scenario would account for the A-movement intervention from PRO in
the vP, and the lack of intervention from inflected infinitivals with a non-PRO subject
in [Spec, TP]. This assumption also implies that the NP/DP-like contrast of the
infinitival is a reflex of the infinitive having assigned structural case to an external
argument.
Of course, there is the evidence from OC PRO. If we assume that the subject
of an OC infinitival is an expression that moves for case reasons, then positing a TP
layer that would be capable of assigning case would not work. I am suggesting that
because it is NP-like in its overt case and displacement restrictions, a bare infinitival
cannot assign case to its external argument. If it were the contrary, and structural
case was assigned to the subject of an OC infinitival, then certainly, that subject
would be barred from moving to the Spec of a functional head in the matrix sentence
whose case did not morphologically match. And here’s the problem: The subject in
the infinitival would be marked with genitive case prior to raising to the matrix
clause. Once in the matrix, the subject would be assigned nominative case.
197
This
would result in a structural case mismatch which, it was argued, results in a (PF)
crash. Meeting the case-matching requirement in these circumstances is untenable as
far as I can see.
198


197
This discussion makes sense only if one assumes the Inverse Case Filter. If T° did not have to
discharge a Case feature, the problem of being assigned a second structural case in the matrix clause
would be moot.
198
In fact, this may be the reason there is no possessor-raising construction in Turkish. The structural
case mismatch prohibition restricts possessor raising only to those constructions we have seen
consistently throughout this work, that of a genitive possessor to a genitive subject position of an
embedded clause.

247

In the absence of any other compelling factors, then, let’s assume that the
structure of a Turkish bare infinitival is at most a vP (although there may be other
aspectual phrases, able to, for example, host a time adverbial, that are irrelevant for
case and the EPP), and the inflected infinitival is at most a TP.
199

With this in mind, let’s look at the following examples from Kornfilt (1997):

63) *[Çal-mak] bir sonat
play-INF one sonata
Intended: ‘a sonata to play’

There are a number of reasons why (63) is bad. If this is an OC infinitival, where is
the antecedent of PRO, and how does that antecedent get case? If it is a case of
PRO
ARB
(as in an infinitival sentential subject), then a direct object can’t be extracted.
Furthermore, there are no RC verbal morphemes here; there is no CP layer in this
structure.

64) *[Cem-in çal-ma-sı] bir sonat
Cem-GEN play-INF-3s one sonata
Intended: ‘A sonata for Cem to play’

Example (64) is unacceptable because the infinitive, being inflected, is a DP without
case. In addition, as with the previous example, there is no RC morphology here, so
sonata could not have been relativized. Example (64) is simply two unrelated
phrases: one being [a sonata], and the other, [Cem’s playing].

65) [[[Çal-ma]-a bala-mak] iste-di-im] bir sonat
play-INF-DAT start-INF want-NSR-1s one sonata
‘a sonata I want to start to play’

199
Pointing out that infinitivals do not allow relativization (for example, (61)), Kornfilt (2005, and
elsewhere) argues that Turkish infinitivals are not CPs but DPs.

248

In (65), we have two NP infinitivals with OC PRO one embedded in the other, with
this structure embedded still in a relative clause. The derivation for (65) proceeds as
shown in (66), as follows. Because I am assuming that OC infinitivals are derived via
movement, I will refer to the subject of the infinitivals as ‘PRO-1s’ (PRO-
1stPersonSingular). This is to say that the subject ben ‘I’ picks up its theta-role as
PRO-1s in the infinitival, although technically there is no “PRO” at all. I merely use
the term PRO out of convention.
First, note that the +Wh-sonata is the complement of Verb1 çal ‘play’.
PRO-1s merges as the subject, and sonata raises to [Spec, vP]. This is NP1, (66)a.

66) a. [
VP
sonata + play ]

[
vP
PRO-1s
1
[
VP
sonata + play ] v°]

[
vP
sonata+ACC [
vP
PRO-1s
1
[
VP
sonata + play ] v°]

[NP1 [
vP
sonata+ACC [
vP
PRO-1s
1
[
VP
sonata + play ] v°]]


vP = NP1

sonata+ACC
PRO-1s
VP v°

sonata V°
play


NP1 merges as the complement of bala ‘start’. PRO-1s raises from [Spec, vP] to the
agentive theta position in [Spec, vP] of start. sonata raises from [Spec, vP] of NP1 to
[Spec, vP] of start, which is now NP2, as in (66)b.


249

65) b. [
VP
[NP1 [
vP
sonata+ACC [
vP
PRO [
VP
sonata + play ]]]-DAT] + start]

[
vP
PRO-1s
1
[
VP
[
vP
sonata+ACC [
vP
PRO-1s
1
[
VP
sonata + play ]]-DAT +
start] v° ]

[
vP
sonata+ACC [
vP
PRO-1s
1
[
VP
...
... [NP1 [
vP
sonata+ACC [
vP
PRO-1s
1
[
VP
sonata + play ]]]-DAT]+ start]
v° ]

[NP2 [
vP
sonata+ACC [
vP
PRO-1s
1
[
VP
...
... [NP1 [
vP
sonata+ACC [
vP
PRO-1s
1
[
VP
sonata + play]]]-DAT] +
start] v° ]]


vP = NP2

sonata+ACC
PRO-1s
+2,+1 VP v°

vP = NP1 V°
start
sonata+ACC
PRO-1s
+1 VP v°

sonata V°
play



NP2 merges as the complement to iste ‘want’. PRO-1s raises from [Spec, vP] of NP2
to another theta position in [Spec, vP] of want. sonata raises from [Spec, vP] of NP2
to [Spec, vP] of want.
T° merges with this vP, and ‘I’ raises to [Spec, TP] to be assigned genitive
case. sonata raises to [Spec, CP] and the NSR relative clause is triggered when T°
moves to C°. Finally, sonata is promoted to the external head of the RC.



250

65) c. [
VP
[NP2 [
vP
sonata+ACC [
vP
PRO-1s
1
...
...[
VP
[NP1 [
vP
PRO-1s
1
[
VP
sonata + play ]]-DAT + start] v° ]] ] + want ]

[
vP
I
1
[
VP
[NP2 [
vP
PRO-1s
1
[
VP
...
... [NP1 [
vP
PRO-1s
1
[
VP
sonata + play ]]-DAT] + start] v° ]] + want ] v° ]

[
vP
sonata+ACC [
vP
I [
VP
[NP2 [
vP
PRO-1s
1
[
VP
NP1[
vP
PRO-1s
1
[
VP
sonata
+ play ]]-DAT] + start] v° ]] + want ] v° ]

[
TP
I+GEN [
vP
sonata+ACC [
vP
I
1
[
VP
[NP [
vP
PRO-1s
1
[
VP
[
vP
PRO-1s
1
[
VP

sonata + play ]]-DAT + start] v° ]] + want ] v° ] T° ]

[
CP
sonata+ACC [
TP
I+GEN [
vP
sonata+ACC [
vP
I [
VP
[NP [
vP
PRO [
VP
[
vP

PRO [
VP
sonata + play ]]-DAT + start] v° ]] + want ] v° ] T° ] C°]

[NP [
CP
sonata [
TP
I+GEN [
vP
I [
VP
[NP [
vP
sonata PRO [
VP
[
vP
PRO [
VP

sonata + play ]]-DAT + start] v° ]] + want ] v° ] T° ] C°] one sonata]


CP

sonata
+Wh+ACC TP C°

I+GEN
vP T°

sonata+ACC
“I”+3
+1,+2 VP v°

vP = NP2 V°
want
sonata+ACC
PRO-1s
+2,+1 VP v°

vP = NP1 V°
start
sonata+ACC
PRO-1s
+1 VP v°

sonata V°
play




251

This derivation is possible because the point where the object and the subject raise
from one infinitival to the next, there are no interveners, and (if one adopts a theory of
phases) at each strong phase, they are at the edge of the phase. There are no strong
phases between the vPs, and thus there is no reason movement should be barred from
the most deeply embedded infinitival clause to outermost relative clause.
Note also that the movement of the structurally case-marked accusative to the
[Spec, vP] positions of the two outer clauses is possible because there is no case
mismatch. We are assuming that movement of a structural case-marked expression to
a structural case-marking position is barred for PF reasons. What drives this
movement? The EPP feature of v°. On the other hand, if the accusative object is
satisfying the EPP of v°, something else must be motivating the movement of PRO-
1s-ben. I assume PRO moves to the intermediate positions to pick up a theta role and
for enlightened self-interest. It is moving toward a case-marking position—although
I do not assume look-ahead. Movement of PRO-ben to a theta position in v° does not
satisfy the EPP of v°. The +Wh-sonata still has Wh-features to be checked but
movement through the intermediate [Spec, vP] positions is driven by other factors.
One is that these intermediate v°s have a theta role which must be discharged; other
factors may be Economy (shorter moves), and phases.
200




200
A discussion as to the viability of either of these would take us too far afield here. I mention them
merely as possible motivators, depending on one’s particular theory.

252

Chapter 8: Conclusion
The goal of this dissertation has been to show that the variation in Turkish relative
clauses can be accounted for with a minimum of technology. Within the Minimalist
Program, concepts such as the EPP, Minimality, Case, and the structure of nominals
are recurring themes. An attempt has been made to remain in the spirit of
Minimalism, and refrain from constructing new and unnecessary or redundant
grammatical rules. The evidence in this work seems to confirm that Minimality is a
vital design feature in the grammar as we saw its effects throughout the Turkish
examples. We concluded that it was indeed Minimality or intervention effects we
were observing because for almost every bad example, we were able to construct an
alternative good derivation that included scrambling around the blocking expression
and leaving behind a remnant. We assumed that the +Wh-relative head evacuated the
scrambled expression because it bore the theta role of the gap site and left behind
possessive agreement on the head noun in the remnant. My use of the EPP was more
pragmatic. By showing that specificity is encoded in Turkish by D, and that DPs
must raise, it was both convenient and expeditious that I would rely on the EPP, i.e. a
discrete feature that triggers movement of a DP into the Spec of a functional head. I
have admitted that other accounts may be possible but all the anti-EPP accounts (such
as movement for case, movement for Topic, base-generation and lowering at LF) I
have seen thus far, lack the elegance of the account presented here, at best, or fail to
explain the facts.


253

In addition, my aim was to keep the analysis completely within overt syntax. Little or
no appeals were made to lexical or semantic processes. In fact, the lexicon was
appealed to for selection only, prior to Merge into a derivation, and the semantic
component was referred to for interpretation only, after Spell-Out. The semantic
interpretations of the data were meant to be reflexes of the syntactic operations, as
there was no LF operation or constraint affecting the syntax. The same goes for PF.
No PF constraint was appealed to until after Spell-Out when a derivation would either
converge or crash at PF.
This research project was an exercise in uncovering the clausal structure of
Turkish by using the SR RC form as a diagnostic. We found affirmation that the
structure we had posited for unaccusatives and unergatives was essentially correct.
Or at least, the intervention phenomena we saw in Turkish seemed to indicate that
subjects of unergatives “were in the way” while those of unaccusatives were not. We
also saw that infinitivals seemed to have a smaller structure than was supposed. The
evidence indicated that they may not even contain a TP projection. Furthermore, it
was interesting that all PRO subjects created intervention effects (which could be
skirted by scrambling and remnant movement) except subject control PRO. How is it
that only subject control PRO is a different animal from the others?
The chapters on infinitivals and psych verbs offered analytic tools with which
to probe underlying syntactic structure. Consider the following examples from
Kornfilt (2004). The psych verb üz- ‘sadden’ permits an inflected infinitival as a
subject, but not a factive clause, (1). On the other hand, we see the opposite in the
minimally different sentence with söylenti ‘rumor’ as the subject. Now the factive is

254

acceptable as the complement of rumor, but not the infinitival, (2). It may be that
referentiality, i.e. a categorial difference in the nominal, imposed by selectional
restrictions is the culprit here, but these kinds of minimal pairs need to be examined
more closely in light of the work here.

1) a. [Ali-nin ev-den kaç-ma-sı] ben-i üz-dü
Ali-GEN home-ABL run.away-INF-AGR me-ACC saddened
‘Ali’s running away from home saddened me’

b. *[Ali-nin ev-den kaç-tı-ı] ben-i üzdü
Ali-GEN home-ABL run.away-NSR-AGR me-ACC saddened
Intended: ‘Ali’s running away from home saddened me’

2) a. *[Ali-nin ev-den kaç-ma(-sı)] söylenti-si] ben-i üz-dü
Ali-GEN home-ABL run.away-INF-AGR rumor-AGR me-ACC saddened
Intended: ‘the rumor that Ali ran away from home saddened me’

b. [[Ali-nin ev-den kaç-tı-ı] söylenti-si] ben-i üzdü
Ali-GEN home-ABL run.away-NSR-AGR rumor-AGR me-ACC saddened
‘the rumor that Ali ran away from home saddened me’

On another note, we concluded that human nominals required -features which are
encoded in D. Because they are DPs, +human subjects bar the use of the SR form for
relativization of any other nominal in the clause. Similar effects were observed with
subject experiencers of psych verbs. They too, it seems, must be DPs. We conclude
that the D feature does a lot of work in Turkish. It is the locus of specificity, it
encodes -features, it can satisfy the EPP, it can enter into agree relations with a verb,
it requires case, and its projection can scramble.
If the analysis presented here is on the right track, it raises a host of questions,
some of which I have alluded to and speculated answers to above. There are other

255

questions. For example, one wonders why the SR form is licensed when relativizing
a subject, i.e. an expression from the syntactic position of the subject which is [Spec,
TP]. Along these lines, it is interesting that the Turkish facts resemble Wh-extraction
facts in Tagalog, where the generalization seems to be that Wh-extraction is allowed
only if the expression has a trace in [Spec, IP]. It was though that only subjects could
be extracted in Tagalog, but Nakamura (1993) shows that, in fact, it is the “structural”
subject or topic that can be extracted. Certainly, movement in Tagalog and in
Malagasy, a language with similar properties, could be reexamined in view of the
evidence presented here.
We also saw that accusative DPs could not move to [Spec, TP] whereas other
case-marked expressions could. Is it really because the accusative object is “frozen”
for further A-movement? I have discussed the Inverse Case Filter throughout this
work at times presenting arguments in its favor and at times rejecting it. If T° has a
Case feature it must discharge, then why is it that only structurally case-marked
expressions are barred from moving to Tº. I offered one possible reason: a crash at
the PF interface due to morphological mismatch.
Another puzzling phenomenon we encountered was that scrambled
expressions can neither scramble again nor move further. I presented some ideas as
to why this should be, but these need to be fully worked out or disproved.
Furthermore, what kind of movement is scrambling? Although J. Kornfilt, G. Aygen,
and M. Kural, among others, have worked on scrambling in Turkish, the facts here
shed new light on this phenomenon. Certainly it does not seem to be the case that
scrambling is EPP driven. I mildly concurred with Kural that scrambling seemed to

256

be A-bar movement, but if this is so, a range of questions emerge: where is the
projection, what features are checked, why not to the higher CP, why not successive
cyclic?
Other questions that I have barely addressed concern Case: How is genitive
case licensed in embedded contexts? Is genitive case structural or inherent? Can it
be either, depending on whether it is fulfilling a thematic or functional purpose? How
is it that DPs need case but not NPs? Could it be that in Turkish only -features must
satisfy the Case Filter? Is this something that can be parametric? Could it be correct
that structural case mis-match causes a PF crash. How is it that +human nominals
must have case and raise unless they are contrastively focused? I offered an account
in Chapter 5, but it was tenuous at best and certainly deserves more investigation.
What is it about contrastive focus that makes the expression bearing it immune to
certain overt requirements?
I expect that the proposals made in this dissertation will need to be altered as a
result of future research. On the other hand, the analysis provided machinery with
which to conduct an investigation which has led to the unmasking of other
phenomena that were not so obvious. Thus, if the solutions presented do not
completely hold, the hope is the questions they raised and the logic of the
argumentation will prove valuable.






257



258

Bibliography

Abney, Steven. 1987. The English noun phrase in its sentential aspect. Ph.D.
dissertation, MIT.
Aoun, J. and A. Li. 1989. Constituency and Scope. Linguistic Inquiry 20:141-172.
Aygen-Tosun, Gülat. 1999. Specificity and Subject-Object Positions/Scope
Interactions in Turkish. Ms. Harvard University, Cambridge, MA.
Aygen, Gülat. 2001. Extractability and the nominative case feature on tense. In
Harvard Working Papers in Linguistics.
Aygen, Gülat. 2001. Are there “Non-Restrictive Prerelatives” in Turkish? In
Harvard Working Papers in Linguistics.
Baker, Mark. 1988. Incorporation. Chicago: University of Chicago Press.
Baker, Mark and Osamuyimen T. Stewart. 2002. A Serial Verb Construction Without
Constructions. Ms, Rutgers University.
Barker, Chris, Jorge Hankamer, and John Moore. 1990. Wa and Ga in Turkish.
Syntax Research Center, Linguistics, Cowell College, UCSC, Santa Cruz.
Belletti & Rizzi. 1988. Psych Verbs and -Theory. Natural Language and Linguistic
Theory 6:291-352. Dordrecht: Kluwer Academic Publishers.
Benincà, Paula. 1980. Nomi senza articulo. Rivista di grammatica generativa
5:51-63.
Bianchi, Valentina. 1999. Consequences of antisymmentry: Headed relative clauses.
Berlin: Mouton de Gruyter.
Bianchi, Valentina. 2000. The Raising Analysis of Relative Clauses: A Reply to
Borsley. Linguistic Inquiry 31:123-140.
Borer, Hagit. 1984. Restrictive relative clauses in Modern Hebrew. Natural Language
and Linguistic Theory 2:219-260.
Borsley, Robert D. 1997. Relative clauses and the theory of phrase structure.
Linguistic Inquiry 28:629-647.
Boškovi Željko. 2002. A-movement and the EPP. Syntax 5:167-218.

259

Boškovi Željko. 2005. On the locality and driving force of wh-movement. Talk
presented at the 2005 Mayfest, University of Maryland, College Park.
Brame, Michael K. 1968. A new analysis of the relative clause: Evidence for an
interpretive theory. Unpublished Manuscript, MIT, Cambridge, MA.
Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel.
Burzio, Luigi. 2000. Anatomy of a generalization. In Eric Reuland (ed.), Arguments
and case: Explaining Burzio’s generalization. Amsterdam: John Benjamins. pp.
195-240.
Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.
Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris.
Chomsky, Noam. 1982. Some concepts and consequences of the theory of
government and binding. Cambridge, MA: MIT Press.
Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press.
Chomsky, Noam. 1995. Categories and transformations. In The Minimalist Program
(Chapter 4), 219-394. Cambridge, MA: MIT Press.
Chomsky, Noam. 1999. Derivation by Phase. MIT Occasional Papers in Linguistics,
No. 18. MITWPL, Cambridge, MA. (Reprinted in Michael Kenstowicz (ed.) Ken
Hale: A Life in Language, 1-52. Cambridge, MA: MIT Press, 2001.)
Chomsky, Noam. 2000. Minimalist Inquiries: The Framework. In R. Martin, D.
Michaels, and J. Uriagereka (eds.) Step by Step Essays on Minimalist Syntax in
Honor of Howard Lasnik. Cambridge, MA: MIT Press.
Chomsky, Noam. 2001. Beyond Explanatory Adequacy. MIT Occasional Paper in
Linguistics 20. MITWPL, Cambridge, MA.
Chomsky, Noam. 2005. Lecture presented at LSA Summer Institute, MIT.
Colline, Chris, and Höskuldur Thráinsson. 1996. VP-Internal Structure and Object
Shift in Icelandic. Linguistics Inquiry 27:391-444.
Csató, Eva A. 1985. A syntactic analysis of participle constructions in Modern
Turkish. In Beinci Milletler Arası Türkoloji Kongresi, 23-28 Eylül 1985, Tebliler
1, Türk dili, stanbul: stanbul Üniversitesi, 1985. 39-56.
Del Gobbo, Francesca. 2003. Appositives at the Interface. Ph.D. dissertation,
University of California, Irvine.
Deising, Molly. 1992. Indefinites. Cambridge, MA: MIT Press.

260

Diesing, Molly. 1996. Semantic Variables and Object Shift. In H. Thrainsson, S.D.
Epstein and S. Peter (eds.) Studies in Comparative Germanic Syntax, Vol. II.
66-84. Dordrecht: Kluwer Academic Publishers.
Enç, Mürvet. 1991. The Semantics of Specificity. Linguistic Inquiry 22:1-25.
Erguvanlı, Eser E. 1984. The Function of Word Order in Turkish Grammar.
Berkeley: University of California Press.
Frank, Robert. 2002. Phrase Structure Composition and Syntactic Dependencies.
Cambridge, MA: MIT Press.
Freeze, R. 1992. Existentials and Other Locatives. Language 68(3):553-595.
Freidin, Robert and Rex Sprouse. 1991. Lexical Case Phenomena. In R. Freidin (ed.)
Principles and Parameters in Comparative Grammar. Cambridge, MA: MIT
Press. pp. 392-416.
George, Leland M. and Jaklin Kornfilt. 1981. Finiteness and Boundedness in Turkish.
In Frank Heny (ed.) Binding and Filtering, Cambridge, MA: MIT Press. pp.
105-127.
Hankamer J. and L. Knecht. 1976. The Role of the Subject/Non-Subject Distinction
in Determining the Choice of Relative Clause Participle in Turkish. North East
Linguistic Society, McGill University
Heim, Irene. 1982. The Semantics of Definite and Indefinite Noun Phrases, Doctoral
dissertation, University of Massachussetts, Amherst.
Higginbotham, James. (1987) Indefiniteness and Predication, in E. J. Reuland and A.
G. B. ter Meulen (eds.) The Representation of (In)definiteness. Cambridge, MA:
MIT Press, 43-70.
Hiraiwa, Ken. 2001. On Nominative-Genitive Conversion. In O. Matushansky and E.
Guerzoni (eds.) A Few From Building E-39, MIT Working Paper in Linguistics 39.
MITWPL. Cambridge, MA, pp.66-123.
Hoji, Hajime. 1985. Logical form constraints and configurational structures in
Japanese. University of Washington Ph.D. dissertation. Seattle, WA.
Holmberg, Anders, and Christer Platzack. 1995. Object Shift and the Role of Case. In
The Role of Inflection in Scandinavian Syntax. Oxford: Oxford University Press.
Hong, Soo-min. 2002. Multiple Case Stacking in Korean. Ms. Univ. of Maryland,
College Park.
Hornstein Norbert. 1999. Movement and Control. Linguistic Inquiry 30(1), pp. 69-96.

261

Hornstein, Norbert and Jacek Witkos. 2001. Yet Another Approach to Existential
Constructions. Ms. University of MD, College Park.
Isever, Selçuk. 2003. Information structure in Turkish: the word order-prosody
interface. Lingua 113: 1025-1053.
Jonas, Dianne, and Jonathan D. Bobaljik. 1993. Specs for Subjects: The role of TP in
Icelandic. In MIT working papers in linguistics 18: Papers on Case and agreement
I, 59-98. MITWPL, MIT, Cambridge, MA.
Kamp, H. 1981. A Theory of Truth and Semantic Interpretation. In J. Groenendijk,
T. Janssen, and M. Stockhof (eds.) Formal Methods in the Study of Language,
Mathematisch Centrum, Amsterdam.
Karimi, Simin. 2003. On object positions, specificity, and scrambling in Persian. In S.
Karimi (ed.) Word Order and Scrambling. Oxford: Blackwell Publishers. pp.
91-124.
Karimi, Simin. 2005. A Minimalist Approach to Scrambling Evidence from Persian.
Berlin: Mouton de Gruyter.
Kayne, Richard S. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press.
Kelepir, Meltem. 2001. Topics in Turkish Syntax: Clausal Structure and Scope. MIT
Ph.D. dissertation. Cambridge, MA.
Kennelly, Sarah. 2003. The implications of quantification for the role of focus in
discourse structure. Lingua 113:1055-1088.
Kiparsky, P. and Kiparsky C. 1971. Fact. In L. Jakobovits and D. Steinberg (eds.),
Semantics: An Interdisciplinary Reader. Cambridge: Cambridge University Press.
Knecht, Laura. 1985. Subject and Object in Turkish. Ph.D. Dissertation. MIT,
Cambridge, MA.
Kornfilt, Jaklin, 1984. Case Marking, Agreement and Empty Catagories in Turkish.
Harvard University Ph. D. dissertation. Cambridge, MA.
Kornfilt, Jaklin, 1984. Case Marking, Agreement and Empty Catagories in Turkish.
Harvard University Ph. D. dissertation. Cambridge, MA.
Kornfilt, Jaklin, 1991. Some current issues in Turkish syntax. In Hendrik Boeschoten
and Ludo Verhoeven, E. J. Brill (eds.), Turkish Linguistics Today. Leiden: The
Netherlands.
Kornfilt, Jaklin. 1988. A typology of morphological agreement and its synactic
consequences. In Proceedings of the Chicago Linguistics Society, 24: 117-134.

262

Kornfilt, Jaklin. 1997a. Turkish. Routledge: New York.
Kornfilt, Jaklin. 1997b. On the Syntax and Morphology of Relative Clauses in
Turkish. In Dilbilim Aratırmaları. Kebiçek Yayıları: Ankara.
Kornfilt, Jaklin. 2003. Scrambling, Subscrambling, and Case in Turkish. In S. Karimi
(ed.) Word Order and Scrambling. Oxford: Blackwell Publishers. pp. 125-155.
Kornfilt, Jaklin. 2005. Agr in Turkish as an expression of categorial features. MIT
Working Papers in Linguistics: Proceedings of WAFL 2 Conference 2004 in
Istanbul, Turkey.
Kural, Murat. 1992. Properties of Scrambling in Turkish. Ms., UCLA.
Kuno, 1973. The structure of Japanese language. Cambridge, MA: MIT Press.
Kural, Murat. 1993. V-TO(-I-TO)-C in Turkish. In Filippo Beghelli & Murat Kural
(eds.) UCLA Occasional Papers in Linguistics, v.11.
Kural, Murat. 1997. Postverbal Constituents in Turkish and the Linear
Correspondence Axiom. Linguistics Inquiry 28:3 pp. 498-519.
Lasnik, Howard. 1998. Exceptional Case marking: Perspectives old and new. In Z.
Boskovic, S. Franks, and W. Snyder (eds.) Formal Approaches to Slavic
Linguistics: The Connecticut Meeting 1997. Michigan Slavic Publications, pp.
187-211.
Lasnik, Howard. 2001. A note on the EPP. Linguistic Inquiry 32:356-362.
Lasnik, Howard. 2002. On exceptional Case marking constructions. In Korean
Linguistics Today and Tomorrow: Proceedings of the 2002 Association for Korean
Linguistics, International Conference on Korean Linguistics, pp. 39-54.
Lavine, James, and Freidin, Robert. 2002. The Subject of Defective T(ense) in Slavic.
Journal of Slavic Linguistics 10: 253-89.
Levin, Beth. 1993. English Verb Classes and Alternations A Preliminary
Investigation. Chicago: University of Chicago Press.
Longobardi, Guiseppe. 1994. Reference and Proper Names: A Theory of
N-Movement in Syntax and Logical Form. Linguistics Inquiry 25:609-665.
Mandelbaum, D. 1994. Syntactic Conditions on Saturation. Ph.D. dissertation,
CUNY.
McCloskey, James. 1990. Resumptive pronouns, A'-binding and levels of
representation in Irish. In Randall Hendrick (ed.). Syntax & Semantics, vol. 32.
The Syntax of the Modern Celtic Languages. New York; Academic Press.

263

Meral, Hasan Mesud. 2004. Resumptive Pronouns in Turkish. Master’s Thesis,
Boaziçi University, Istanbul.
Miyagawa, S. 2001. EPP, scrambling and wh-in-situ. In Michael Kenstowicz (ed.)
Ken Hale: A Life in Language. Cambridge, MA: MIT Press.
Moore, John. 1998. Turkish Copy-Raising and A-Chain Locality. Natural Language
and Linguistic Theory 16: 149-189. .
Murasugi, Keiko. 2000. Japanese Complex NP’s and Antisymmetry. In R. Marton,
D. Michaels, and J. Uriagereka (eds.) Step by Step: Essays in Minimalist Syntax in
Honor of Howard Lasnik. Cambridge, MA: MIT Press.
Nakamura, Masanori. 1993. An economy account of Wh-extraction in Tagalog. In
Proceedings of WCCFL 12. Stanford: CSLI.
Ortega-Santos, Iván. 2005. On Locative Inversion and the EPP in Spanish. Paper
presented at the VIII Encuentro Internacional de Lingüística del Noroeste,
University of Sonora, Mexico.
Ouhalla, Jamal. 1993. Negation, Focus and Tense: The Arabic Maa and Laa. Rivista
di Linguistica 5.275-300.
Özsoy, Sümrü. 1994. Türkçe’de Ortaç Yapısı. Dilbilim Aratırmaları. Ankara: Hitit.
Perlmutter, D. 1978. Impersonal Passives and the Unaccusative Hypothesis. In
Berkeley Linguistic Society IV, 157-189. University of California.
Perlmutter, D. 1969. Deep and Surface Constraints in Syntax. New York: Holt,
Rinehart & Winston.
Pesetsky, David and Esther Torrego. 2001. Tense to C:Causes and Consequences. In
Michael Kenstowicz (ed.) Ken Hale: A Life in Language. Cambridge, MA: MIT
Press.
Pesetsky, David. 1994. Zero Syntax Experiencers and Cascades. Cambridge, MA:
MIT Press.
Postal, Paul. 1966. "On So-Called Pronouns in English", in F. Dineen S.J. (ed.),
Report of the 17th Annual Round Table Meeting on Languages and Linguistics,
Georgetown University Press, Washington, D.C. Reprinted in D. A. Reibel and S.
A. Schane (eds.), Modern Studies in English: Readings in Transformational
Grammar. NJ: Prentice-Hall (1969).
Rackowski, Andrea. 2002. Subject and Specificity: The Case of Tagalog, NELS 32.
Richards, Norvin. 2005. The Agreement sequence for successive-cyclic movement in
Tagalog and English. Talk presented at Mayfest, Univ. of Maryland, College Park.

264

Rizzi, Luigi. 1990. Relativized Minimality. Cambridge: MIT Press.
Rizzi, Luigi. 1996. Residual Verb Second and the Wh-Criterion. In Adriana Belletti
and Luigi Rizzi (eds.) Parameters and Functional Heads Essays in Comparative
Syntax. New York: Oxford University Press.
Rizzi, Luigi. 1997. The Fine Structure of the Left Periphery. In L. Haegeman (ed.),
Elements of Grammar. Handbook in Generative Syntax. Dortrecht: Kluwer
Academic Publishers, 281-337.
Ross, John R. 1967. Constraints on Variables in Syntax. Ph.D. dissertation. MIT.
Sauerland, Uli. 2002. “Unpronounced Heads in Relative Clauses”. In K. Schwabe and
S. Winkler (eds.) The Interfaces. Amsterdam: John Bejamins.
Schutze, Carson. 2001. On Korean “Case Stacking”: The varied functions of the
particles ka and lul. The Linguistics Review 18(3), 193-232.
Schachter, Paul. 1973. Focus and relativization. Language 49(1):19–46.
Sezer, Engin. 1986. The Unmarked Sentential Subject Constraint in Turkish. In D.
Slobin and K. Zimmer (eds.) Studies in Turkish Linguistics Vol. 8, pp. 123-135.
Philadelphia: John Benjamins Publishing Co.
Sigursson, Halldór. 1992. The Case of Quirky Subjects. In Working Papers in
Scandinavian Syntax 49:1-26.
Siloni, Tai. 1996. Hebrew Noun Phrases: Generalized Noun Raising. In A. Belleti
and L. Rizzi (eds.) Parameters and Functional Heads, 239-267.
Stowell, Tim. 1989. Subjects, Specifiers, and X'-Theory. In M.R. Baltin and A.S.
Kroch (eds.) Alternative Concepts of Phrase Structure, pp.232-262. Chicago:
University of Chicago Press.
Stowell, Tim. 1991. Small Clause Restructuring. In Robert Freidin (ed.) Principles
and Parameters in Comparative Grammar. Cambridge, MA: MIT Press. pp.
182-218.
Svenonius, Peter. 2002. Subjects, Expletives, and the EPP. NY: Oxford University
Press. 3-27.
Szabolcsi, Anna. 1987. "On Combinatory Categorial Grammar". In Proceedings of
the Symposium on Logic and Language, Debrecen, Budapest: Akademiai Kiado,
pp. 151-162.
Takahashi, Daiko. 1994. Minimality of Movement. Ph.D. Dissertation, University of
Connecticut

265

Travis, Lisa 1984. Parameters and Effects of Word Order Variation. Ph. D.
Dissertation, MIT.
Underhill, R. 1972. Turkish Participles. Linguistic Inquiry 3:87-99.
Vergnaud, Jean-Roger. 1974. French Relative Clauses. Ph.D. Dissertation, MIT.
Yükseker, Hitay. 2003. Bir ‘One’ in A. Özsoy, D. Akar, M. Nakipolu-Demiralp, E.
Erguvanlı-Taylan, A. Aksu-Koç (eds.) Studies in Turkish Linguistics, Proceedings
of the Tenth International Conference in Turkish Linguistics, 2000.
Istanbul:Boaziçi Univ. Press
Zimmer, Karl. 1987. “Turkish Relativization Revisited”. In H. E. Boeschoten and L.
Th. Verhoeven (eds.) Studies on Modern Turkish: Proceedings of the third
Conference on Turkish Linguistics, 57-61. Den Haag: Tilbury University Press.










ABSTRACT

Title of Dissertation:

MINIMALITY AND TURKISH RELATIVE CLAUSES Ilhan Merih Cagri, Doctor of Philosophy, 2005

Directed By:

Professor Norbert Hornstein Department of Linguistics

Turkish relative clauses display a subject/non-subject asymmetry. The subject relative (SR) is licensed for relativization from [Spec, TP]. Whereas the non-subject relative (NSR) is never acceptable for subject relativization, the SR is licensed in clauses where there is no external argument, and when relativizing a non-subject in clauses where the subject is non-specific. Within the framework of the Minimalist Program, Turkish RCs are explained in terms of satisfaction of the EPP of T by a D feature and Minimality effects. As long as no nominal expression intervenes between the relative head and [Spec, TP], the SR is licensed. The SR, then, can be used as a diagnostic for movement through TP. Minimality effects are incurred when there is an intervening nominal between T° and the RC head, and the SR becomes unacceptable. The proposal is that in Turkish, specific nominals, +human nominals, and Experiencers of psych verbs all contain a DP projection. Non-specifics are NPs which cannot satisfy the EPP. NP subjects cannot move to [Spec, TP], and thus

permit the SR form for relativization of non-subjects. NPs create intervention effects, as does PRO, with the exception of subject control PRO which is perhaps a trace of movement. Scrambling ameliorates intervention effects. Once scrambled, expressions are frozen but remain porous for movement of a subconstituent. Differences between inherent and structural Case are suggested with structural case assignment limited to DPs and in a Spec-Head configuration. Structurally casemarked DPs are barred from moving to case-assigning positions unless there is a morphological match. Further proposals include structures for verb classes, including Psych verbs, and structures for infinitivals and +human DPs. Contrastive focus is briefly addressed. Though superficially complex, relativization in Turkish can be accounted for with a minimum of technology. The suggestions here have implications for the theory of the EPP, Case, its assignment and interface conditions, feature satisfaction, and movement.

MINIMALITY AND TURKISH RELATIVE CLAUSES

By Ilhan Merih Cagri

Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park, in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2005

Advisory Committee: Professor Norbert Hornstein, Chair Professor Howard Lasnik Professor Paul Pietroski Professor Amy Weinberg Professor Christopher Cherniak

Ilhan Merih All rights reserved. Box 1346 Ann Arbor. All rights reserved. ProQuest Information and Learning Company 300 North Zeeb Road P. United States Code.O. MI 48106-1346 . UMI Microform 3202436 Copyright 2006 by ProQuest Information and Learning Company. This microform edition is protected against unauthorized copying under Title 17.UMI Number: 3202436 Copyright 2005 by Cagri.

© Copyright by Ilhan Merih Cagri 2005 .

.

Dedication In memory of my parents. Dedicated to Amina and Nadia. ii .

I must first thank God. independent. for blessing me with so many good things in life and for allowing me to experience and succeed in this wonderfully fulfilling endeavor. all cousins. I find myself at a loss for words. However. “So. iii . friends. but not a day went by when she wouldn’t ask. I consider the noble character of my father and the adventurous spirit of my mother gifts that I hope will be passed on to their grandchildren and great-grandchildren. now that I am actually at this juncture. In two years of schooling. I run into people who knew about my mother’s children and grandchildren. or the movie theater manager in Bethesda. My mother was born to Jewish parents in Ottoman Turkey. All aunts and uncles. when are we going to see your doctorate?” My mother was one to shamelessly brag about her children.Acknowledgements I have thought about what I would write in the acknowledgment section of my dissertation long before I wrote the dissertation--sort of like planning where to hang the pictures before you construct the wall. People like taxi drivers in Washington. My parents were immigrants of the old school. neighbors. Colleagues. it would be remiss of me if I did not brag about my mother here. This work is a testament to the strength and talent of my parents. I exaggerate not. So. courageous and generous. knew we were the special ones in the family. MD. my siblings and our kids. and store clerks were all told how exceptional and how successful we were. or the hair dresser in Columbia. They were industrious. To this day. humble. my daughter. Before I begin acknowledging a few of the people I feel grateful to. DC. the Creator. My mother did not live to see the completion of my studies. taxi drivers. have asked me in detail about me.

Turkish. and my mother was banished from the family. Needless to say. and it was from the shop doorway that my mother watched her brother’s Bar Mitzvah procession on its way to the reception from which she was excluded. scandal ensued. and entertained friends of different backgrounds (smattering of Greek and Italian). This didn’t stop my grandfather from using her labor at his shop. My mother lived every day of her life. Not one moment was wasted. The fiance was immediately married off to my mother’s young aunt. classic and current. loved poetry and art. and appreciated all kinds of music. European. worked as a nanny to an American diplomatic family (she had learned English).. My mother embroidered and knit beautifully.S. worked at the Mexican Embassy (she spoke Spanish). I know only that she enrolled in a school of French cuisine (she spoke French). of course. my father thought she was nuts. however. Working in various positions in many different embassies in Washington DC. And in her iv . my mother was as active and as creative. In the U. Of my mother’s exploits there. my mother made our house the scene of glittering soirees. I remember diplomats gushing over the greasy jack my mother had found on the side of a road and placed on our credenza.she was already in the fourth grade. On one occasion. No one in the family was permitted to speak to my mother. at which point her father withdrew her from school so that she could work in his laundry. She hardly slept. My mother was engaged to a nice Jewish boy by the time she was 17 but decided instead to run off with a very handsome (“just like Errol Flynn!”) Muslim Turk 15 years her senior. My mother and father moved to Ankara shortly thereafter. Well. my mother’s sister (believing the lament that the family was disgraced and the other girls would never marry) ran off to Israel.

No matter how unpredictable the disaster. He taught me how to examine things carefully and to reason things through. My father was a fixer. Many will understand when I say that my father loved his family so much that his eyes teared as he gazed upon us. My aunts. and for requiring so little of me being during this period. Ida Dana. I owe my confidence. I thank my brother-in-laws. Idris was there and willing to fix the problem. and being loved by them. and Mati Revah. Herb Conger and Bruce Trock for their encouragement. Nesim Revah and Leon Dana. and this skill I learned from my father. I cannot express how proud I am of her abilities and accomplishments. and stress. and Jews wearing Yarmulkas stood side-by-side as they said the Muslim Jenazeh prayer. one or more of my v . and uncles. Christians. At her funeral. a Jewish cantor sang. To my family. was my standard bearer and shield. I am grateful to Zeki and Yahya for their forbearance. My children watched in awe and suffered with tolerance as I made this long journey. Meri Baruh. He hated to inconvenience anyone. At different times. Homeyra. My father was a peaceful man of dignity and integrity.characteristic way. I thank my father for teaching me about strength and courage. It was by loving them. she worked full-time until three days before her death at age eighty. gave me healthy doses of love and encouragement. Syntax is about problem solving. My daughter. and Rodrigo’s Concierto de Aranjuez blared loudly as she was interred by her family. Çaya and Beyhan and my brother Kemal for their love and support. chaos. Muslims. Idris is owed a special note of gratitude for maintaining his good-natured spirit in spite of quotidian bouts of poverty. that I was able to keep my sanity. Their mantra was: “You can do anything!” I thank my sisters.

This dissertation is in honor of the mothers and wives among my friends. My two daughters-in-law. not really. The professors here are truly dedicated to “teaching”. I have a firm grasp of the issues if not the details. Stephen Crain and Rozz Thornton would give me a solid background in language acquisition.. Mariam and Fatimeh. therefore.. I.) I have countless friends who carried me through this ordeal. I was inspired enough to produce a whole chapter of this thesis.. My degree belongs to the sisterhood of these friends who were happy for me. On one such occasion. David Lightfoot and Howard Lasnik. I consider it an act of God’s good Grace that I ended up in the Linguistics Department at UMCP. You must think I’m crazy! FATIMEH: (weakly) noooo. Juan Uriagereka.children would engage in discussions related to an area of interest to me.. they never waivered in their support of me nor stinted in their kindness and generosity.. Rozz. dedicate Chapter 7 on Infinitivals to Haroon.. I was fortunate to have been able to study syntax with Norbert Hornstein. Laura Benua taught phonology in such a way that even today years after I took her classes. These scholars are tireless in their efforts to help students develop an understanding of the field and adopt sound research methods. encouraged me and uplifted me. it is not enough for an instructor to have made a name for himself in the field. prayed for me. to their collective wisdom and their often selfless aspirations. As every student knows. especially helped me develop sound research vi . but to their credit. spent many a day wondering just how mad their mother-in-law was. (Conversation with Fatimeh: ME: I’m really sorry about the appalling way I acted yesterday.

praised. The grad students during my tenure as a student were a gifted lot. Howard Lasnik taught me how to think and write like a linguist. and for always seeing the best in me. vii . and know-how and invaluable assistance in professional matters. Elixabete Murgia. I was encouraged by the special talents of people like Sachiko Aoshima. Itziar San Martin. Acrisio Pires. It is still amazing to me how accessible these people were. Norbert inspired. David Poeppel. Ana Gouvea. yelled. Especially helpful and generous with his time was Tomohiro Fujii. Cilene Rodrigues. critiqued and criticized. I must also express heartfelt gratitude to Jacek Witkos who was the angel that pulled me out of a particularly barren trough. and for sincerely wanting success for me. and carried me through. Amy Weinberg provided cheer and counseling for the private troubles. open-mindedness. and professionalism. Hirohisa Kiguchi. I am obliged to these people for their assistance and their example during the early years of my studies. Max Guimarães. I owe thanks to Scott Fults. Philip Resnik. Ivan Ortega-Santos and Masaya Yoshida for providing inspiration in the latter years. prodded. They provided the department an atmosphere of talent and intellectual stimulation. I cannot relay how deeply grateful I am to Norbert for his wisdom. and Amy Weinberg. The guidance of these individuals during the final period of my dissertation writing was invaluable. My studies in linguistics would have never gotten off the ground if it were not for Norbert Hornstein. Mitsue Motomura.methodology. From my first syntax course. Paul Pietroski had faith in me when even I doubted myself. I am indebted to them for their patience. Other professors to whom I am indebted for their teaching talent and erudition are Paul Pietroski. Early on. through years of painful highs and lows.

I will name one here: Martha Price. Lydia Grebenyova. Mona Diab. I feel honored to know these linguists and am thankful for the privilege of having studied with them. Nina Kazanina. Usama Soltan and Heather Taylor gave of their friendship as well as their expertise. Without Atakan’s help. viii . I believe God was instrumental in my success. I owe a huge debt of gratitude to a few people in particular who made me believe that a dream could become a reality and who sustained me through the early years. Pritha Chandra. I formed special friendships during this period. Finally. I am indebted to Aylin Bener. and Murat Aytekin for this invaluable assistance. Every linguist needs good informants. I would not have been able to finish much of the work I accomplished in this thesis. I thank you. and He provided Atakan Ince for me just when I needed him. People like Cilene. Leticia Pablos.Thanks also to the generosity of Utako Minai. Soomin Hong. people who can reliably provide judgments and who remain patient for hours. I wish to thank all those people who I cannot name for lack of space. Mehmet Ergene. Youngmi Jeong.

.............................................................. 33 2............................................................................................... 30 2................... 43 2..........1 Clauses that lack external arguments: the SR form........................ 26 2................................... 48 3 A minimalist account: Pestesky and Torrego (2001) ........... 22 2.................................................................................................... 101 7 Explaining the choice in RC forms ........................................................................................................7 The option of the NSR form..............5 Clauses that lack external arguments: the SR form................2 The NSR -DIK form................................................................... 14 1................................................. 14 1 A little Turkish grammar.....ii Acknowledgements . 16 1......................3 The SR form ..................................1 Background ...............................................................2 Overview ................................................................ 39 2...................................... 100 6 Repeat of Pesetsky and Torrego (2001) ................8 Genitive case ..................................................................................... 22 2 Returning to relative clauses: Generalizations ..... 53 3................ix Chapter 1: Introduction ............................................................................................................................... 78 2 Toward an analysis of DP/NP structure and Case in Turkish ....................... 71 3............................................................ 27 2..........................................................2 More examples with the SR form: the possessor of a direct object .............. 78 1 Introduction ................................................................................................................................................................................................5 Optionality of RC forms.................... 4 Chapter 2: Explaining Turkish Relative Clauses.........6 Relative clauses with complex arguments ................................ 76 Chapter 3: Specificity............................................................................................ Case and the EPP .....................................................4 Recap: our new story................................................... 74 4 Conclusion...................................................................11 Semantic reflex of syntactic structure ....................................... 105 8 Summary ................................ 33 2................................4 Diagnostics for non-specific subjects............5 NPs................... 1 1 Background .......................... 63 3.10 Relative clauses with complex arguments ............................................................................................. 59 3................................................................. 29 2.......................................... 82 3 The EPP and case assignment ...................... 94 4 Looking at Turkish ‘Quirky’ relatives .....9 Clearing up the SR option ...................................................................................................................... 35 2...... DPs....................3 Recap..................................iii Table of Contents ........................... 96 5 The subject/non-subject asymmetry is a misnomer ....................................... 14 1..............................................................................4 The EPP............................................ 109 ix ......................... 61 3....... 37 2............................................................................................................1 Pesetsky and Torrego (2001) analysis applied to Turkish relative clauses...................................................... 19 1.......................................Table of Contents Dedication ...............6 The option of the SR form.......................................................................................... 69 3...................... 15 1...............................................................................................3 A look at Turkish nominals: specificity effects ................................................

............................................. 143 1 Background .................................................. 132 4......1 Experiencer subjects........................................................................ 220 2........................................ 110 1..........1 The base position of subjects and intervention by NPs............ 241 4 Taking stock ..................... 252 Bibliography ................................. 140 Chapter 5: Human/Non-human Distinctions in Turkish ......... and displacement.... 180 Chapter 6: Relativization in Psych Verb Constructions ............................................................ 139 6 Conclusion.....................1 Subject Experiencers with accusative Theme ... 124 3 The SR and Movement through [Spec...1..............................................................................................................2 Experiencer objects ...........................................................................................1............................................................ 213 1 Background ................................ case..........................................................................1 The EPP on T in Turkish................................................. 187 1 Background ............... 213 2 Uninflected Infinitivals ..........Chapter 4: The EPP on T and Minimality............................................1 Specificity........................................................................... 193 2................................... 245 Chapter 8: Conclusion ......................................................................... 187 1........................................................ 126 3.......................................................................................... 143 2 The Problem of Human Subjects ....................................................... 127 4 Toward an explanation .............................. 190 2 Turkish Psych Verbs and Relativization ...................................................................... 228 3 Inflected infinitivals . 112 1............... 258 x .................................................................................................................................................................1 Classes of Turkish psych verbs ..................................................................................... 226 2.... 204 2......................................................................................... 119 2 Review of Turkish Relative Clauses .......................... 194 2.........................................................................2 Evidence for the EPP in Icelandic and German ............................................................................................... 161 5 Summary ..............................................................................................1 Relative clauses with non-specific subjects and the SR form.... 109 1 Introduction and Background..................................................... 152 4 Explaining the Behavior of Human Nominals ............................................................................................... 179 6 Contrastive Focus and Human Subjects........................ 205 3 Conclusion.............................................................2 Non-subject infinitivals ............................................................ 147 3 Toward a Solution ............................. 110 1................................................................... 136 5 Versions of the EPP............................... 209 Chapter 7: Relativization from Infinitivals in Turkish .........................................................1 Review of assumptions............................................................................................................... TP] ...................................

where applicable. it includes research into several syntactic properties that have been attested cross-linguistically and theories in generative grammar that have been proposed to explain them. The native strategy is pre-nominal in the sense that the restricting clause which contains a gap precedes the nominal that is the external head of the 1 . I must clarify at the outset that the account presented here is solidly within the Minimalist Program (mainly) of Chomsky (1995. it is assumed outright that the reader agrees with the movement (or displacement) account of syntactic derivations. specifically the EPP on T.and A-bar). however. to point out areas that may be controversial or where an alternative account would work just as well without distracting from the arguments being presented.Chapter 1: Introduction While this work centers on relative clause constructions in Turkish. By way of introduction. Turkish has two relative clause (RC) strategies. thetaassignment at first merge and structural case-assignment/checking by functional heads and other basic hallmarks of generative linguistics). 2000). An attempt has been made throughout to keep theory internal assumptions to a minimum (modulo the initial assumptions regarding movement (both A. Much of this work. The one which is the topic of this dissertation is native to the language. Thus. ends up being about subject-hood and the EPP. X-bar structure and binary branching. and. the other is borrowed from Persian.

the Persian ki. 2 As evaluated against the language relevant diagnostics in Del Gobbo (2003: pp. commonly identified (using a variety of similarly mnemonic labels) as the Subject Relative (SR) clause form and the NonSubject Relative (NSR) clause form based on the grammatical function of the expression that would have appeared in the clause internal gap position. it can still be true that the girl lost her bag. 3 A. Ince (p. whereas the native Turkish RC employs non-finite or subordinate clause inflections. who bought the book. This form also employs the verbal inflections of matrix sentences. The native RC itself has two forms.1 There is no relative pronoun and an overt resumptive pronoun is not permitted in the gap site. as far as I know. Furthermore. This is expected as Turkish is a consistently head-final language. what has not been noted in the literature. 152-162).c. and it requires a resumptive pronoun when relativizing anything other than a subject or direct object. This contrasts with the form borrowed from Persian which is post-nominal (i. it cannot be true that the girl lost her bag if she did not buy the book. whereas in sentence (ii) with the ki RC. In sentence (i) containing the native RC (i).) pointed out the non-restrictive interpretation of the ki RC form in Turkish.3 This work has nothing further to say regarding the borrowed RC form.e. the restricting clause follows the head) there is a relative pronoun. is that whereas the native form is generally a restrictive relative2. (i) kitab-ı al-an kız çanta-sı-nı kaybetti book-ACC buy-SR girl bag-AGR-ACC lost ‘The girl who bought the book lost her bag’ (ii) kız ki kitab-ı al-dı. even if she didn’t buy the book (note the commas around the English equivalent).clause. çanta-sı-nı kaybetti girl COMP book-ACC buy-PST bag-AGR-ACC lost ‘The girl. the borrowed Persian form functions as an appositive in Turkish. It has been noted that this description is not quite accurate in the sense that the SR form is licensed in some circumstances where its function is something other than the subject 1 For arguments against Kayne’s (1994) LCA entailment that specifier-head-complement is the universal order see Kural (1997). lost her bag’ 2 .

3). This gives us another way of viewing the Subject Relative. Svenonius (2002) notes that the notion of subject is “no more than a descriptive label for an epiphenomenal collection of properties” (p. whereas the SR is sometimes licensed for non-subjects. it is licensed only when that expression has passed through the caseassigning position for the canonical subject. except one that is marked with accusative case. The SR examples will include relativization of nominals bearing a variety of grammatical functions and theta roles. any nominal may move to [Spec. the SR form is acceptable for relativization of all nominals except accusative direct objects. in fact. TP]. As will be seen in later chapters. Although the Turkish Subject Relative can be used when relativizing expressions bearing a variety of theta roles. if we think of the relativized expression as the topic within the 3 . This fact has theoretical implications. We will see in Chapter 2. it is licensed when syntactic and discourse prominence converges. It turns out that under appropriate circumstances. and it is the morpho-phonetic mismatch of structural case at PF that disallows movement of an accusative case-marked expression to the structural case-assigning position of [Spec. TP]. Put simply. and discourse-informational (the topic of a proposition) (Svenonius 2002). licensed when the relativized expression moves to (and through) [Spec. that is. the NSR is never acceptable for relativization of the subject. there are three components to “subject-hood”: thematic (the most prominent argument of a predicate). syntactic (identified by case or agreement).of the RC. This means that given the syntactically permissible conditions for the movement. Interestingly. To put it another way. that the SR is. I conclude that Case is checked at PF. TP].

In fact. the reader is urged to consider throughout the larger issues and to test their applicability to his own research interests. 4 GEN Abbreviations used: ABL (ablative). To be more precise. ACC (accusative). emphasis. NOM (nominative). overt pronouns are unacceptable unless they signal a change of topic or contrast. and if we agree that [Spec. AGR (agreement) AOR (aorist). TP] is the most prominent position in the morphosyntactic arena of case and agreement. LOC (locative). as in (1). and indeed subjects and objects appear only for contrast. The only aspect of subject-hood missing in the definition of Subject Relative is that the relativized expression does not have to be the most prominent in the thematic hierarchy of the predicate. occupies the canonical subject case and agreement position of [Spec. POSS (possessive agreement). we will readdress them at the conclusion of this work when the reader has many more facts under his belt. It exhibits both subject and object drop. NSR (non-subject relative). NPI (negative polarity item). It is too early here to discuss these issues in greater detail. then it is when the clausal “topic” is also the clausal “subject” (i. INST (instrumental). TP]) that the SR form is licensed. NoEA (no external argument). 4 .e. PASS (passive). or other marked discourse purposes. I mention them here to point out that although much of what follows is centered on Turkish. PST (past).clause. (genitive). There is no overt Wh-movement. SR (subject relative). nor any complementizers in the trivial sense. DAT (dative). RC (relative clause). INF (infinitive). 1 Background4 A few facts about Turkish: Turkish is a head-final agglutinative SOV language.

As noted above. and is regular in the stacking of its case and inflectional morphemes which are generally suffixival. Turkish relative clauses demonstrate a subject/non-subject asymmetry. ‘-en’ and the -DIK form can appear as ‘-dik’.1) Ahmet oda-ya gir-di. ‘-dık’.5 What makes Turkish useful for research is that. the NSR -DIK form as in (3)a. The -An verbal form bears no agreement morphology. Because the morpho-phonological processes on the verbal morphemes are rather complex. ‘-duk’. The -DIK suffix. I will use Turkish spelling throughout and not convert the Turkish examples to the IPA. the SR -An form as in (2)a. I will denote the Subject Relative /-An/ form as SR and the Non-Subject Relative /-DIK/ form as NSR. ‘-di ’. or a nonsubject. 5 Although it will be argued in Chapter 5 on human DPs that the assumption regarding subject agreement is not straightforward. ‘-tuk’. There is only subject agreement. is followed by possessive morphology which shows agreement with the subject. ‘-tu ’. There are two verbal suffixes which mark relative clauses in Turkish. unlike synthetic languages where several inflectional elements can be fused into one morpheme. which. ‘tık’. ‘-dı ’. ‘-dük’. on the other hand. He sat down in a/the chair. -tı . 6 The capital letters indicate positions that undergo vowel harmony and consonantal assimilation. bears genitive case morphology. (*He) chair-DAT sit-PST ‘Ahmet entered the room. An and -DIK6.’ Turkish exhibits rich morphology. ‘-du ’. the choice of which is generally determined by whether the clause internal gap site is the subject of the relative. ‘-tik’. 5 . when overt. ‘-ti ’. (*O) sandalye-ye otur-du. there is generally a one-to-one mapping between function and morpheme in Turkish which yields more transparency in the syntax. ‘-tük’ and ‘-tü ’ as a result of phonological processes. The –An form can appear as ‘-an’. Ahmet room-DAT enter-PST. ‘-dü ’. no object agreement.

6 . is attributed to Underhill (1972) by Hankamer and Knecht (1976) whose own explanation of the asymmetry is based on grammatical relations such as subject and object. It is used for subordination structures in general. The relative clauses in (4)8 exemplify part (b) of Kornfilt’s generalization. the gap site is a non-subject in a construction where there is no surface subject bearing a thematic role. [Øi divan-da otur-an] bayani Ø sofa-LOC sit-SR lady ‘the lady who is sitting on the sofa’ b. *[Øi divan-da otur-an] bayani Ø sofa-LOC sit-SR-3s lady Kornfilt (1997a) (among others7) describes the asymmetry by postulating that the NSR form is the elsewhere case.2) a. based on Chomsky (1965). 9 I use Kornfilt’s gloss. She refers to one of the nominal morphemes as a compound marker (CM). the gap site is a subject or part of a larger subject b. *[Øi divan-da otur-du -u] bayani Ø sofa-LOC sit-NSR-3s lady 3) a. and the gap site is the (oblique) object of an impersonal passive construction. the SR form is the marked option used in relative clauses when: a. 8 Examples from Kornfilt 1997. [Øi Ankara otobüs-ün-e bin-il-en ] duraki Ø Ankara bus-CM-DAT board-PASS-SR stop9 ‘the stop where the Ankara bus is boarded’ 7 The first modern analysis. [bayan-ın Øi otur-du -u ] divani lady-GEN Ø sit-NSR-3s sofa ‘the sofa that the lady is sitting on’ b. On the other hand. These phrases contain no external argument (NoEA). 4) a. as in impersonal passives and existentials. Csató’s (1985) analysis was along the lines of Chomsky (1981). Note that only the SR form is acceptable.

-3s ‘A ship is sidling up to the harbor’ c) [gemi-nin yana -tı -ı] liman ship-GEN sidle-NSR harbor ‘the harbor that the ship is sidling up to’ Furthermore. *[Øi Ankara otobüs-ün-e bin-il-di -i ] duraki Ø Ankara bus-CM-DAT board-PASS-NSR stop ‘the stop where the Ankara bus is boarded’ c. the relativized expression. 7 . & Moore (1990). As shown in (5)b. the subject. harbor-DAT ship sidle-pres.prog. [gemi yana -an] liman ship sidle-SR harbor ‘the harbor that a ship is sidling up to’ b) Liman-a gemi yana -ıyor.b. ship. The example in (5)c10 demonstrates that. is non-specific. 5) a. [Øi bu durak-tan bin-il-en] otobüsi Ø this stop-ABL board-PASS-SR bus ‘the bus which is boarded from this stop’ d. This sentence contains an overt clause-internal subject. bears dative case clause-internally and can in no way be identified as a subject. there is a difference in interpretation based on whether the SR or the NSR form is used: in 5)a). however. *[Øi bu durak-tan bin-il-di -i] otobüsi Ø this stop-ABL board-PASS-NSR bus ‘the bus which is boarded from this stop’ What is interesting. is that the SR form also appears in phrases such as those in (5)a where the gap site is not the subject. as pointed out by Barker. harbor. the NSR form is also possible. This difference will later be addressed. ship. ship is specific. as expected. and the NSR form is required. whereas in 5)c). Hankamer. in one common dialect there seems to be some “optionality” in the choice of verbal forms particularly 10 As the glosses indicate.

al. (1990). and resumptive pronouns are also banned in simple relatives. [[Ø1 kız-ı] kitab-ı getir-en] adam1 Ø girl-POSS-3s book-ACC bring-SR man ‘the man whose daughter brought the book’ b. we can assume that Turkish RCs operate within the hallmarks of classic RCs: there is an 11 12 Examples from Barker.when relativizing from within a complex argument.12 There are no internally-headed relatives in Turkish. according to the generalization. 8 . The only way to track A-bar movement is with relative clause constructions. The pairs in (6) and (7) are examples of extraction from within a subject and which. [[Ø1 biz-e güven-ece -i]-nin Ø 1p-DAT trust-FUT-COMP-POSS-3s-GEN doubtful be-NSR-POSS-3s man üpheli ol-du -u] adam1 ‘the man who that (he) will trust us is doubtful’ In sum. I exclude Topic and Sluicing here. [[Ø1 biz-e güven-ece -i] üpheli ol-an] adam1 Ø 1p-DAT trust-FUT-COMP-POSS-3s doubtful be-SR man ‘the man who that (he) will trust us is doubtful’ b. permit the NSR form as in (6)b and (7)b. the choice of which is based on whether the gap site is a subject or non-subject. But these phrases do. should not allow the NSR form. et. and which can sometimes be violated in simple relatives. [[Ø1 kız-ı]-nın kitab-ı getir-di -i] adam1 Ø girl-POSS-3s-GEN book-ACC bring-NSR-3s man ‘the man whose daughter brought the book’ 7) a. but which permit a wider (although not completely free) optionality of choice in complex relatives. Turkish is a Wh-in-situ language. what we have in Turkish RC’s is two verbal suffixival elements. in fact. Thus.11 6) a.

only one will be acceptable. may optionally merge with D° giving us the DP shown in (8). TP] for case. . TP].13 For example. 9 . [NP [CP . . If this is correct. [gap]1 . Turkish being a head-final language. then. having moved to (and then out of) [Spec. . I argue throughout this work that the SR relative clause form is indicative of the Wh-expression. it provides us with a diagnostic for A-movement to [Spec. there are three DPs that can potentially be targetted for relativization. It turns out that of the two RC forms. after which it A-bar moves to 13 There is an apparent exception in extraction from sentential subjects as in (6) and (7). the direct object and the locative. and the NSR form will be required otherwise. that the internal gap site (sometimes referred to as the focus) of the RC is a +Wh-expression (or Operator) that undergoes A-bar movement at least to the CP projection. the subject. in the phrase in (9). the relative head. is as in (8) where X1 is the N° head and the CP its complement. one can formulate a RC targeting any one of those nominals as the head of the RC.] N°]. ] X1 ] D°] I assume. the structure of an RC. TP] to [Spec. . 8) [DP [NP [CP . i. . The diagnostic works like this: Given a phrase with several nominals.obligatory gap in the (case) position of the RC head internal to the RC which is coreferenced with the nominal expression (the external head) which the RC is the complement of. Assuming that the subject-DP must first A-move to [Spec.e. and the evidence indicates. TP]. but it will be shown that even here the generalization holds. . The resulting structure. The SR form will be required when the relative head A-bar moves from [Spec. CP].

In (11). DP2-direct object DP3-locative Verb-NSR]] DP11 Extraction of DP2. and again the NSR form is required because the subject must move to [Spec. TP] is occupied by a subject that is non-Wh. 9) [DP1-subject DP2-direct object DP3-locative Verb] 10) a. the NSR form is required and the SR form barred.[Spec. as in (10)a and disallow the NSR form. . the direct object. . the NSR form is obligatory. TP] to be assigned case (to avoid a Case Filter Violation). should require the NSR form and prohibit the SR form because the non-Wh-subject (DP1) must move to [Spec. CP]. 12) [CP Ø1 [TP DP1-Subj+case [vP DP1-Subj [VP Ø1+DAT DP3-Loc Verb-NSR]]] ] DP21 10 . *[CP Ø1 [TP . 11) [CP Ø1 [TP DP1-Subj+case [vP Ø1+ACC [VP Ø1+ DP3-Loc Verb-NSR]]]] DP21 In (12). as in (10)b. on the other hand. TP] for case. [CP Ø1 [TP Ø1(+case) [vP DP2-DO-(ACC) Ø1(+theta) DP3-locative Verb-SR]] ] DP11 b. Any time [Spec. we predict that extraction of DP1-subject will trigger the SR form. the focus is now a dative object.

or to intervention effects created by the presence of a nominal between [Spec. TP]. If one wants to see if an expression can move to [Spec. much of this unacceptability will be due to effects similar to what we saw in (11) and (12). and the NSR form is required. TP]. In this case. is the diagnostic that will be used throughout this work to tease apart movement. The workhorse in this thesis is the latter. TP] one simply targets that expression as the relative head. in a nutshell. we will see that there are occasions where the subject does not raise for case. this can be either because [Spec. As will be seen. we can assume that if the SR form is acceptable. and sees if the SR form will be acceptable or not. what is termed Minimality effects. constraints on (A-)movement of a DP induced by the presence of intervening expressions. as in (12). if we target either the dative DP2 (or the locative DP3) as the relativized expression. and specificity and movement. then the dative must have moved through [Spec. In addition. TP] and the relativized Wh-expression. TP]. contra 11 . As the details in this work are laid out. if the SR form is bad. then the dative did not move through [Spec. 13) [CP Ø1 [TP Ø1+DAT [VP Ø1+DAT DP3-Loc NP1-Subj Verb-SR]]]] DP21 This then.But. On the other hand. we see that there is a correlation between specificity and case. where the subject occupies [Spec. TP] was occupied (by the subject) or an intervening element blocked the movement. as shown in (13).

15 Actually. This restriction does not hold for inherent case-marked expressions. call it what you like. In Chapter 5. TP]. The bulk of the findings in this project leads to one conclusion: that the SR form is an instantiation of movement to [Spec.Chomsky’s assumption even as recently as 2005.15 Because this work is mainly about movement of DPs. in addition to works outlining the Minimalist Program.14 we will see evidence that structurally case-marked expressions are not “frozen” for further movement but are simply barred from moving to another structural case-assigning position. 2005. As we encounter more and more Turkish data. the effects of scrambling and constraints on scrambling are discussed when they become germane. Chapter 3 is a continuation of the argument in the sense that assumptions made in Chapter 2 about specificity and the nature of NPs and DPs are worked out in more detail. 14 12 . we see evidence that Turkish is In lectures delivered at LSA Summer Institute. the controversial issue of the EPP is addressed early. something along the lines of the EPP as a feature that needs to be checked seems to be working in Turkish. Chapter 2 is dedicated to explaining why this conclusion is viable. and hopefully presented in a manner that is more compelling. Here we also examine the structural hierarchy of various verb classes finding support for Perlmutter’s (1978) “Unaccusative Hypothesis” and Burzio’s (1986) similar findings regarding predicate structure in Italian. Initially I assume outright a definition of the EPP as a feature on a functional head which forces movement of a DP to its Spec. we will revisit the EPP and will be led to conclude that. even this is not quite accurate. The evidence seems to suggest that a structurally casemarked expression is barred from moving to a case position where it will be assigned a case with a different morphological form than the one it already bears. MIT. Although scrambling is not a topic per se in this work. Chapter 4 argues for the EPP as a formative feature.

nonhuman features. syntactic referentiality. both inflected and uninflected. I review conclusions that I have reached and point to theoretical questions that this work highlights. not animacy. In Chapter 6. the effects of contrastive focus on movement and case-marking are demonstrated. In Chapter 8. The controversy regarding control PRO (movement or not) emerges because it seems that other than subject control PRO. a D feature. but human vs.e. Rather than being merely a semantic notion. all other control PRO positions serve as interveners for movement. In both these chapters. The facts lead to this conclusion by the end of Chapter 7. The final Chapter of this research project is a compendium of issues that have been visited in this work and that are relevant cross-linguistically. and indeed are remarkable because they seem to reemerge in language after language. 13 . Does this mean that subject control PRO is a trace of movement as proposed by Hornstein (1999). In Chapter 7.sensitive to. is imposed on human subjects and on Experiencers of psych verbs with consequences in terms of movement and intervention effects. we look at relativization out of infinitival clauses. Many of the observations made in the course of this research are useful in presenting a different perspective with which to view phenomena in other languages. we look at similar effects with psych verbs. i. the facts in Chapter 5 make the case that these features play a role in the syntax. Also in Chapter 5.

and indeed. Kornfilt assumes that the NSR -DIK form is not licensed in subject gap RC’s because the strong AGR of this form would. 14 .CP] violating the A -disjointedness Requirement (Aoun and Li 198916). i. McCloskey (1990). Using this fact as an indication of a strong AGR. and Ouhalla (1993). Borer (1984). its Governing Category which in this case is the CP. (1)b. and Moore (BHM) (1990) have provided analyses under a Government and Binding framework. The A’-disjointedness Requirement is a sub-clause of a generalized version of Condition B and was argued for in Aoun & Li (1989). Kornfilt (1984. More recently. The unavailability of pro in the subject position permits a non-pronominal empty category at the subject gap. Kornfilt (1984. Hankamer and Knecht (1976).e. and Knecht (1985). 1988. as in (1)a. This pro would be (A-bar) bound by an Operator in [Spec. the SR form does not bear any agreement. 16 “The A’-disjointedness Requirement: A pronoun must be (A’-)free in the smallest Complete Functional Complex (CFC)”. 1991. Recall that the NSR -DIK form bears agreement with the RC subject. and according to Kornfilt. 1991). must license pro in subject position. 1997b) and Barker. Hankamer. Let’s briefly look at Kornfilt’s (1997b) proposal. Conversely. as well as others.1 Background The subject non-subject asymmetry in Turkish relative clauses has been of interest to many linguists: Underhill (1972).Chapter 2: Explaining Turkish Relative Clauses 1 A little Turkish grammar 1. the weak AGR of the SR -An verbal form cannot license pro. have attempted to provide an account.

[[ ei okul-a gid-en] Opi ] adami school-DAT go-SR man ‘the man who goes/went to school’ b. *[Øi divan-da otur-du -u] bayani Ø sofa-LOC sit-NSR-3s lady 15 . Of the two possible combinations in sentences with “No External Argument” (NoEA).1) a. GAP SITE RC STRATEGY 1Subject -An (SR) 2non-subject -DIK (NSR) 3Subject -DIK (NSR) * 4non-subject -An (SR) 5NoEA -An (SR) 6NoEA -DIK (NSR) * Table 1: Acceptability of possible strategies EXAMPLE (2)a (3)a (2)b (5) (4)a (4)b 2) a. only the SR form (item 5) is licensed despite the fact that these sentences have no canonical subject to extract. [Øi divan-da otur-an] bayani Ø sofa-LOC sit-SR lady ‘the lady who is sitting on the sofa’ b. the NSR form in sentences which contain a subject gap (item 3) is the only form not found in the grammar. and which forms appear in the grammar. *[[proi okul-a git-ti -i] Opi ] adami pro school-DAT go-NSR-poss3s man Intended: ‘the man who goes/went to school’ 1.2 Overview Let us first be clear about the logical possibilities for Turkish relatives. Table 1 demonstrates that of the four possible combinations with external arguments.

18 sentence (6)c shows that a case-marked object cannot remain inside the VP. 19 The interaction between specificity and (especially accusative) Case has been noted by modern (e. and accusative. casemarked expressions are in a different structural position than their bare counterparts. In addition. genitive (of subordinate clause subjects).17 Assuming that Turkish adverbs of manner mark the edge of the VP. Dede (1986). See Kural (1992). that ships sidle) up to’ 1. *[Øi bu durak-tan bin-il-di -i] otobüsi Ø this stop-ABL board-PASS-NSR bus ‘the bus which is boarded from this stop’ 5) [gemi yana -an] liman ship sidle-SR harbor ‘the harbor that a ship is sidling (or.3) a. Erguvanlı-Taylan (1984). which she identifies as nominative. *[Øi divan-da otur-an] bayani Ø sofa-LOC sit-SR-3s lady 4) a. 16 . Erdal (1981). while an object without overt case must remain inside the VP as in (6)a-b. Kornfilt (2004) shows that the correlation between specificity and overt case holds for all structural cases. This is demonstrated for the direct object in (6). Enç (1991). Tura (1986)) and traditional grammarians. Kornfilt (1997). Kornfilt claims that inherently case-marked nominals are ambiguous in their specificity as these expressions enter a derivation already case-marked. [bayan-ın Øi otur-du -u ] divani lady-GEN Ø sit-NSR-3s sofa ‘the sofa that the lady is sitting on’ b. Nilsson (1986).3 A look at Turkish nominals: specificity effects Arguments do not always bear overt case morphology in Turkish. [Øi bu durak-tan bin-il-en] otobüsi Ø this stop-ABL board-PASS-SR bus ‘the bus which is boarded from this stop’ b.g.19 17 18 Examples from Tosun (1999).

Ali bir piyano kirala-mak isti-yor.’ b. Ali one piano rent-INF want-3s-PRES ‘Ali wants to rent a (non-specific) piano. the presence of overt case. ‘specific’ for piano in sentence (9a) means there is a certain piano such that Ali wants to rent it. 20 For example. 7) a. 17 .’ Enç (1991) notes that in Turkish.AOR -1sg c. Ali one piano-ACC rent-INF want-3s-PRES ‘Ali wants to rent a certain piano. This means we have a diagnostic for raising for objects.ACC quickly read-aor-1sg ‘I’ll read the book quickly. indefinite nominals in object position always unambigously receive a specific or non-specific interpretation depending on whether or not they bear overt case morphology.’ So.AOR -1sg ‘I read the book quickly.’ b. Ali bir piyano-yu kirala-mak isti-yor.20 whereas the non-case-marked object in (7)b must receive a non-specific reading. Ben hızlı kitap oku-r-um I quickly book read-AOR-1sg ‘I read books quickly. *Ben hızlı kitab-ı oku-ru-m I quickly book-ACC read. Ben kitab-ı hızlı oku-ru-m I book. a bare object must remain in-situ and be non-specific. The object in (7)a bearing accusative case must be interpreted as a specific piano.6) a. for objects the facts are as follows: a case-marked object must raise from its base position and receive a specific interpretation and conversely.’ d. *Ben kitap hızlı oku-r-um I book quickly read.

the subject of an embedded existential construction in Turkish does not bear overt case (9)a. 18 . If it is correct that temporal adjuncts are generated high in the structure (adjoining perhaps to vP or TP). and as expected. Whereas nominative case is the Ø or null morpheme in Turkish. as in (10). Note the position of the embedded subject in (9)b. one girl-GEN side-his-LOC be-NSR-3s-ACC see-PAST-3p ‘They saw that (of the salient girls) one (of them) was by his side. is the unmarked and non-scrambled cases.The same correlation between specificity. This is consistent with what we have observed so far—that specifics must bear case. The case-marked specific subject must raise. 8) Ali-nin Ankara-ya git-ti -i-ni duy-du-lar Ali-GEN Ankara-DAT go-NSR-3s-ACC hear-PST-3p ‘They heard that Ali went to Ankara’ We know that subjects of existential constructions must be non-specific. as shown by the unacceptable (10)b. as in (8). side-his-LOC one girl be-NSR-3s-ACC see-PAST-3p ‘They saw that there was a girl by his side. overt case and raising can be seen for subjects in embedded environments. as well as its interpretation when the embedded subject is case-marked. Ali-nin bu sabah Ankara-ya git-ti -i-ni duy-du-lar Ali-GEN this morning Ankara-DAT go-NSR-3s-ACC hear-PST-3p ‘They heard that Ali went to Ankara this morning’ 21 What is relevant for us. Bir kız-ın yan-ın-da ol-du -u-nu gör-dü-ler.’ b.21 9) a. then the subject is required to raise above temporal adjuncts. embedded subjects receive genitive case.’ 10) a. Yan-ın-da bir kız ol-du -u-nu gör-dü-ler. of course.

I assume that the EPP is a feature of the functional heads v°. Ay e-nin pasta-yı ka ık-la /bahçe-de /hızlı ye-di -i-ni Ay e-GEN cake-ACC spoon-INST/garden-LOC/fast eat-NSR-3s-ACC saw-PST-3p ‘They saw that Ay e ate the cake quickly/in thegarden/with a spoon’ b. 19 . functional heads have an EPP feature22 and that all phrasal movement is driven by the EPP. the facts show that only case-marked arguments receive a specific interpretation. *bu sabah Ali-nin Ankara-ya git-ti -i-ni duy-du-lar this morning Ali-GEN Ankara-DAT go-NSR-3s-ACC hear-PST-3p Because case morphology and displacement seem to go hand in hand. specifics must be case-marked. Ay e-nin ka ık-la /bahçe-de /hızlı pasta ye-di -i-ni Ay e-GEN spoon-INST/garden-LOC/fast cake eat-NSR-3s-ACC saw-PST-3p ‘They saw that Ay e ate (some) cake quickly/in thegarden/with a spoon’ d. Furthermore. case and specificity holds for objects in embedded environments. gör-dü-ler 11) a. I am not wedded to any particular definition of the EPP. this description of the EPP is different from the notion of the EPP as a structural requirement of an occupied specifier.b. T°. Although the outcome is the same. for me the “EPP” is merely a label for whatever it is that drives overt XP movement. At this point. The examples in 11) demonstrate that the correlation between raising. *Ay e-nin ka ık-la /bahçe-de /hızlı pasta-yı ye-di -i-ni gör-dü-ler A. and non-specifics cannot bear overt case. we can conclude that (at least in embedded environments) case marking on an argument—in the form of genitive case on the subject and accusative case on the object—is evidence of raising. *Ay e-nin pasta ka ık-la /bahçe-de /hızlı ye-di -i-ni gör-dü-ler Ay e-GEN cake spoon-INST/garden-LOC/fast eat-NSR-3s-ACC saw-PST-3p gör-dü-ler c.-GEN spoon-INST/garden-LOC/fast cake-ACC eat-NSR-3s-ACC saw-PST-3p 1. A more detailed discussion of the EPP will follow in Chapter 4.23 Thus T° (and vº) has an EPP feature that must be 22 To be more specific. and C°.4 The EPP I assume that in Turkish.

20 . We saw a similar pair in (9).24 The subject in (12)a. is non-specific and cannot raise. and receives overt case (albeit genitive25). the specific subject raised to T. T’s EPP feature is satisfied by the locative “at/by his side” whereas in (9)b.satisfied.’ b. This is supported by the pair of sentences in 12). Sokak-ta köpek havl-ıyor.’ Another potential argument for an EPP feature on Tº comes from the pair of sentences below from Kural (1992). 24 from Kelepir (2001). to satisfy the EPP. I exclude all scrambling-type movement as it is not germane. dog street-LOC bark-PRES ‘The dog is barking in the street. In (9)a. TP]. Recall that nominative case is the Ø-morpheme. we can assume that the locative expression has raised from VP to [Spec. Köpek sokak-ta havl-ıyor. I assume that the unacceptability of the sentence in (13)a is due to failure to satisfy the EPP of T. It is generally assumed in the literature that Turkish locatives are generated in the VP (Kural 1992). 12) a. so the subject in (12)b is presumably “overtly” case-marked and has raised to [Spec. and scrambling. 25 It is beyond the scope of this thesis to address how embedded subjects receive genitive rather than nominative case. The embedded subject to the right of the locative in (9)a is non-specific and bears no overt case. One approach is Hiraiwa (2001) which suggests that genitive case is assigned by a v-T-C amalgam. which is optional and apparently costless. TP] to satisfy the EPP on Tº. street-LOC dog bark-PRES ‘A dog/dogs are barking in the street. i. By parity of reasoning from the examples we saw above with respect to objects and embedded subjects. Compare (12)a with (12)b where the locative is lower than the specific subject. In later chapters I will distinguish between feature driven movement. Compare with the minimally 23 By movement.e.

I have encoded these facts by assuming that it is the EPP on T that must be satisfied or the derivation will crash. o-ra-dan this-??-ABl that-??-ABL ‘from here’ ‘from there’ (ii) a. Burada [bir tavuk] pi iyor here a chicken is-cooking c.26 13) a. Note that the word for ‘here’ in Turkish is a nominal expression with locative case. bu-ra-ya b. but a locative can raise to T and save the derivation. *[Bir tavuk] pi -iyor a chicken cook-PRES-AGR ‘A chicken is cooking’ b. and in the absence of a specific subject. as demonstrated in (13)c. TP] must be occupied. I do not know what -ra. I include this much here to justify my assumption that in Turkish sentences. (i) a. Although bu and o can function as independent lexical items denoting ‘this’ and ‘that’. as shown in (i) and (ii). o-ra-ya this-??-DAT that-??-DAT ‘to here’ ‘to there’ 21 . another nominal is required to move to that position.different acceptable sentences in (13)b which contains a locative expression.is and assume it means something like ‘place’. 26 I take the expressions ‘here’ and ‘there’ in Turkish to be nominal because they can take a variety of cases. [Spec. bu-ra-dan b. A nonspecific subject cannot satisfy the EPP. bu-ra-da this-“place”-LOC ‘here’ [Literally: ‘at this place’] I will return to a lengthier discussion on the EPP in Chapter 4. respectivelly.

have an EPP feature which must be satisfied by a DP. we will see how much mileage these assumptions buy us in formulating an account of Turkish relatives. The facts about case morphology and displacement fall out if I further assume that only DP’s need satisfy the Case Filter and that the EPP can only be satisfied by DP’s. In order to account for the facts in relative clauses. 2 Returning to relative clauses: Generalizations I have argued that. That is. on the other. functional heads28. overt case. and I will address the arguments behind these assumptions later as well. The issue throughout this thesis is the EPP on T° (and on C°). and the fact that bare arguments in-situ must be non-specific. or a specific nominal. I will discuss this assumption in more detail in the next chapter. DPs. in Turkish. I will not argue whether or not v° has an EPP feature. I mention v° merely for uniformity although the facts do show that accusative objects must raise. I have also assumed that nonspecifics are NPs which do not need to satisfy the Case Filter nor can they satisfy the EPP.27 Neither of these assumptions seems far-fetched. I will adopt an NP/DP distinction for Turkish.1. 22 . Chomsky (1995) defines the EPP as a “strong D feature”. Case and the EPP In order to capture the complementarity between raising. Tº and vº. I assume that nonspecific nominals are NPs and specific nominals are DPs. but let’s take it as reasonable and adopt it for now. and specificity on the one hand. I must further assume that 27 28 In fact.5 NPs. In the next section.

I should mention that rather than Operator movement. Bianchi 1999) for relative clauses. With this much technology in hand. the subject lady.29 14) a. Nothing being argued rests on a specific analysis for relative clauses. First. is a DP. CP] to satisfy the EPP on Cº in movement . In sentence (14)a. let’s return to the facts in Table 1. we are only concerned with operations internal to the relative clause. to [Spec.Cº also has an EPP feature which attracts a +Wh DP to its specifier. although the matching analysis would also work for the issues being presented here. We will not concern ourselves with whether this element further moves to a position external to the clause or whether some matching operation co-indexes it with an external head. I assume that the +Wh element itself moves all the way to [Spec. Vergnaud 1974. As the simplified derivation in (14)b shows. item 2 in Table 1. lady. CP]. being specific. The +Wh DP . the EPP of Tº attracts the subject DP. with modifications. [bayan-ın Øi otur-du -u ] divani lady-GEN Ø sit-NSR-3s sofa ‘the sofa that the lady is sitting on’ b. 29 23 . sofa +Wh lady+GEN VP DP t-lady T° CP TP C° DP t-sofa +uWh V° sit Throughout this paper. I will be assuming the raising analysis (Brame 1968. TP] in movement sofa moves to [Spec. At this point. Kayne 1994. Schachter 1973. . let’s look at the NSR form.

(For simplicity. The subject is also the +Wh relative head.) As demonstrated in (15)b. TP]. movement . bee. T has an EPP feature which must be satisfied. So. the SR form is licensed in the RC we saw in (5) (item 4 in Table 1). Let us begin with subject gap clauses as in sentence (15)a. as noted in No. The 24 . Recall that when the subject is being relativized. I have omitted details in the tree such as the vP projection and accusative case on the object. the SR form must always be used. It is attracted by the EPP of . CP]. the derivation of which appears in (15)b. the subject bee moves from [Spec. is attracted to [Spec. Cº. In contrast to the example in (14) above. bee +Wh t-bee +Wh DP t-bee +Wh CP TP VP C° T° DP girl V° sting Let’s turn now to clauses with non-specific subjects. Recall that the NSR form is obligatory (when relativizing a non-subject) when the subject is specific. let’s shift gears and look at all the RCs that require the SR form to see if we can find any commonality among them. [[Ø1 ] [ [kız-ı sok-an ] ] arı1 Ø girl-ACC sting-SR bee ‘the bee that stung the girl’ b. in movement 15) a. 4 in Table 1.Now. precisely because the subject ship is non-specific.1 in Table 1. The DP subject. TP] to [Spec. repeated as (16)a. The SR form is also licensed when the subject is non-specific and a non-subject is being relativized. No.

the subject ship is specific. [Spec. leaving [Spec. the DP-subject. In this case it attracts the +Wh-DP. [gemi-nin yana -tı -ı] liman ship-GEN sidle-NSR harbor ‘the harbor that the ship is sidling up to’ b. movement . TP] vacant for the +Wh-expression to move to. to satisfy the EPP features of Cº. What is different between the derivations of the RCs in (16)a and (17)a? Notice that in the tree in (17)c. harbor TP t-harbor +Wh NP ship C° CP VP DP t-harbor +Wh T° V° sidle. CP]. *[gemi-nin yana -an] liman ship-GEN sidle-SR harbor 25 . The EPP of Tº must be satisfied by another nominal. 17) a. with derivation (17)c. the NP-subject remained in-situ.subject is an NP and cannot satisfy the EPP on a head. TP] was occupied by a non-Wh element. As demonstrated in the tree in (16)b. and the NSR form is required. This +Wh-element in [Spec.to Notice that in the minimally different example (17)a. whereas in (16)b. ship cannot (and has not) raised from its base-generated position. . gemi yana -an liman ship sidle-SR harbor ‘the harbor that a ship is sidling up to’ b. 16) a.up. harbor. TP] now moves to [Spec.

every time the specifier of T is occupied by a non-Wh expression. as in (14) and (17). No. the NSR form is required. at this stage. I have included the sentence in (18)b to demonstrate that neither of the two nominal expressions in the relative clause. we can now predict that phrases with no external arguments will also require the SR form. the +Wh DP bus moves to [Spec. the SR form is licensed. 2. As the derivation in (19) shows. [Spec. harbor +Wh ship+GEN CP TP VP C° T° DP t-ship DP V° t-harbor+DAT sidle. 5 in Table 1. This means that where the EPP of T is satisfied by an expression that is also +Wh. Because there is no “subject” to occupy it. let’s look at the clause (18)a. TP] will be vacant for a +Wh non-subject to move to it to satisfy T’s EPP feature.up.c. “this stop” and “bus”. TP] to satisfy the EPP on T in . I am merely making an observation. requires structural case: the PP ‘from this stop’ is rendered in Turkish as the nominal this stop with ablative case. To demonstrate this idea.to +Wh In fact. and bus receives inherent dative case. This element then moves to 26 .1 Clauses that lack external arguments: the SR form Having formulated a generalization. I do not mean this to be an explanation.

and the SR form is required. [Øi bu durak-tan bin-il-en] otobüsi Ø this stop-ABL board-PASS-SR bus ‘the bus which is boarded from this stop’ b. CP] to check the EPP on C. In . we have a +Wh element in [Spec. girl’s leg. as in 30 .[Spec. Otobüs-e bu durak-tan bin-il-ir. We assume that there is a vP projection that assigns accusative case (in a [Spec-head] configuration) from sentences 27 . Because the object. In the SR example in (20)a. Again. and the relativized element originates within the accusative object.30 and receives accusative case. the subject. bee. . the subject is non-specific. the EPP of T attracts the I had up to now ignored the vP projection but must include it here to have the object raise above the subject so that we do not incur Minimality violations or intervention from the subject. is a DP. bee remains in its base-generated position.’ 19) +Wh-bus TP t-(+Wh-bus) VP t-(+Wh-bus) VP V° board T° C° CP DP this stop-ABL 2. In this sentence.2 More examples with the SR form: the possessor of a direct object Let us now look at the derivation of a sentence that is more complicated. bus-DAT this stop-ABL board-PASS-AOR ‘The bus is boarded from this stop. it is attracted by the EPP of v. As illustrated in the derivation in (20)b. TP] 18) a. is a non-specific NP which cannot satisfy the EPP.

28 . 20) a. The direct object has had its case checked/assigned. girl raises again from [Spec. Compare (i) with (ii). Of course. I have ignored the issue of features so far and will address this in more detail later. TP]. the SR form is licensed when a +Wh element has moved to [Spec.DP girl-GEN leg+Agr v° V° sting D° T° CP TP C° Notice in (20)b. However. +Wh-girl. (i) Yol-u (bir) araba tıkamı road-ACC one car blocked ‘A car has (or ‘Some cars have) blocked the road’ (ii) Araba yol-u tıkamı car road-ACC blocked ‘The car has blocked the road’ 31 The Wh-expression also deletes C’s +Wh feature. [[Ø1 baca -ın-ı] arı sok-an] kız1 Ø leg-POSS-ACC bee sting-SR girl ‘the girl whose leg a bee/some bees stung’ b.31 Again. where car must now receive a specific interpretation. that movement is of the +Wh element out of the now-accusative object-DP in [Spec. and thus is frozen for movement into another case-checking (A-) position. vP]. girl +Wh +Wh-girl vP DP+ACC NP-bee VP DP +Wh. by some car or other. CP] to satisfy the EPP on C. but rather that the road is blocked. vP].possessor of the object. Perhaps a more or more accurate interpretation for (i) is ‘The road is blocked by a car’ where what is important in the utterance is not that a car has blocked the road. and where it has presumably raised to [Spec. from the specifier of the object in [Spec. it is such as (i). TP] to [Spec. TP] and been assigned nominative case.

to construct an appositive in Turkish one must make use of the borrowed Persian complementizer ki which introduces a clause. [arı-nın [[Ø1 baca -ın-ı] sok-tu -u] kız1 bee-GEN Ø leg-POSS-ACC sting-NSR-POSS. neither RC form permits relativization of the entire DP. the subject must raise for the derivation to converge. and that nominals with specifiers are DPs. TP] to avoid a Case Filter crash. ‘John’s car which I washed’ does not mean ‘(of John’s three cars) the car that I washed’.3S girl ‘the girl whose leg the bee stung’ b. a. Because ‘Anne’s mother’ picks out a unique individual. T has an EPP feature which can only be satisfied by a DP. RC’s in Turkish fail a range of del Gobbo’s (2003) diagnostics for appositives. Whereas relativization of a complex possessor-possessee DP object is not acceptable. i. attraction of any DP other than the subject will violate Minimality. A possessor-possessee structure is a DP and too big to be promoted. If proper names are referential and must therefore be a D. then (i) is bad because a D(P) cannot be promoted. *[ Ø1 arı-nın sok-du -u ] [kız-ın baca -ı]1 Ø bee. even though both arguments are equidistant from the point of view of T’s EPP. nevertheless marked by matrix verbal and case properties. some possessor-possessee relatives sound odd: ‘the woman that bought the fish’ sounds much more natural than ‘Anne’s mother that bought the fish’. extraction of the DP from the specifier of that DP object is acceptable.32 2. When the subject is specific. Thus.33 The SR form is licensed when the expression that satisfies T’s EPP feature is +Wh.3 Recap Let us review our assumptions thus far. I do not assume a left-branch condition exists in Turkish. of the Wh-DP is promoted. [[Ø1 baca -ın-ı] arı sok-an] kız1 Ø leg-POSS-ACC bee sting-SR girl ‘the girl whose leg a bee stung’ Even in English. In a clause where the subject is non-specific. only the head.porous for movement from within it. (ii)a is the NSR form with a specific subject and (ii)b is the SR with a nonspecific subject. In fact. *[ Ø1 arı sok-an ] [kız-ın baca -ı]1 Ø bee sting-SR girl-GEN leg-POSS ‘the girl’s leg that a bee/some bees stung’ b.e. i. it must obligatorily move to [Spec. vP]. is an NP which cannot satisfy the EPP on 32 Interestingly. Although I will not elaborate here. though subordinate. The SR and NSR RCs can only be restrictive (Meral 2004) and this may be one reason possessor-possessee DP’s cannot be relativized. If the subject is a DP. 29 .3S girl-GEN leg-POSS ‘the girl’s leg that the bee stung’ ii. Another way of looking at this might be to remember that we had assumed the raising analysis of RP’s in which it is only the N° of the RC head that is promoted beyond the CP to the matrix clause.GEN sting-NSR. This makes sense if we are assuming that in relativization. a. This would also explain why a RC such as (i) is bad. possessor-possessee DP’s may be more like appositives. or N°.POSS. Both the SR strategy in (i)a and the NSR strategy in (i)b are unacceptable. it must be attracted to the Spec of T. (i) *Anne that bought the fish 33 Except for the accusative case-marked object also in [Spec. girl’s leg. I assume that DP arguments receive structural case in a Spec-Head configuration.

+Wh-fields is attracted by T. This is shown in sentence (21).T. [Ø1 mısır yeti -en] tarla1 Ø corn grow-SR field ‘the field where corn grows’ b. Nominal expressions with bazı are always specific (22)b. Birkaç patterns like the English ‘some’ in that it can receive either a specific on non-specific interpretation (22)a. thus. There are two determiners in Turkish which both mean ‘some’ but differ in terms of their specificity.4 Diagnostics for non-specific subjects34 Enç (1991) points out Turkish has quantificational determiners35 and NPI expressions that have selectional restrictions for specificity. field +Wh t-field +Wh PP/DP t-field +Wh CP TP VP C° T° NP corn V° grows 2. 30 . relativization of any other element will license the SR morpheme. fields is the relative head. At some point in the history of this phrase. 34 I will return to specificity and the nature of DPs and NPs in Chapter 3. but I include a brief discussion here as it is important to the account I am proposing. there was a +Wh element in [Spec. This is because the specifier of T will be free to host the +Wh-non-subject. 35 Although the determiners in question translate as ‘some’ and could be called quantifiers. Because the subject is nonspecific. where the locative DP. the SR form is required. TP]. 21) a. I am interested in their determiner-like properties and so will refer to them as determiners.

i. garden-LOC some children exist ‘There are some children in the garden. Ali hiçbir *kitap/kitab-ı al-ma-dı.’ 31 . Bazı çocuklar bahçe-de. Notice the word orders in (23)b and (23)c: in the existential construction with non-specific subject (23)b. precisely because bazı requires a specific.e. presupposed. 23) a. Ali Zeyneb-DAT some books/books-ACC mail-PAST ‘Ali mailed some of the books to Zeyneb. the locative must raise above the subject whereas in (23)b the specific subject has raised to T.’ b. garden-LOC some children exist ‘There are some children in the garden. Ali Zeyneb-e birkaç kitap/kitab-ı postala-dı. Ali any book/book-ACC buy-neg-PAST ‘Ali didn’t buy any of the books.’ c. *Bahçe-de bazı çocuklar var. some children garden-LOC ‘Some of the children are in the garden. Ali Zeyneb-e bazı *kitap-lar/kitab-lar-ı postala-dı.) This determiner always forms a specific nominal expression in Turkish: it requires accusative case morphology (24)a and is banned from existential constructions (24)b. 24) a.’ The same pattern can be seen in the Turkish negative polarity determiner hiçbir ‘any’ (literally ‘any one’. Bahçe-de birkaç çocuklar var. interpretation.’ b.’ The determiner bazı (but not birkaç) is ungrammatical in existential constructions (23)a but not in the non-existential locative construction (23)c.22) a. Ali Zeyneb-DAT some book/book-ACC mail-PAST ‘Ali mailed some /some-of-the books to Zeyneb.

Likewise. *Bahçe-de hiçbir çocuk yok. if the SR form is licensed when nonrelativized subjects are non-specific NP’s. garden-LOC any child doesn’t-exist ‘There aren’t any of the children in the garden. Both bazı and the determiner birkaç (which allows both the specific and the non-specific readings) are acceptable with the NSR form in (25)a. yields ungrammaticality in (26)b. or the specific NPI. 25) a. If we are on the right track. [Bazı/birkaç arı-nın [[Ø1 baca -ın-ı] sok-tu -u ] kız1 some of the/some bee-GEN Ø leg-POSS-ACC sting-NSR-POSS. [hiçbir arı-nın [[Ø1 baca -ın-ı] sok-ma-dı -ı] kız1 any bee-GEN Ø leg-POSS-ACC sting-NSR-POSS. that is. hiçbir. *[[Ø1 baca -ın-ı] bazı arı sok-an] kız1 Ø leg-POSS-ACC some bee sting-SR girl ‘the girl whose leg some (of the) bees stung’ c. while (25)b demonstrates that bazı is unacceptable in the otherwise grammatical SR clause we saw in (20)a. *[[Ø1 baca -ın-ı] hiçbir arı sok-may-an] kız1 Ø leg-POSS-ACC any bee sting-SR girl 32 .’ Thus certain determiners and NPI items are incompatible with a non-specific interpretations: nominal expressions with bazı ‘some’ and with the negative polarity determiner hiçbir ‘any’ must always be interpreted as specific. we would predict that the SR form would be unacceptable when the subject contains the obligatorily specific determiner. hiçbir. the NPI item. We see that this is indeed the case.3S girl ‘the girl whose leg some of the/some bees stung’ b. bazi. [[Ø1 baca -ın-ı] birkaç arı sok-an] kız1 Ø leg-POSS-ACC some bee sting-SR girl ‘the girl whose leg some bees stung’ 26) a.b.3S girl ‘the girl whose leg no bee stung’ b.

This further supports our argument that the SR form (which must be used when extracting a subject) may be used when a non-subject is being relativized only when the clausal subject is non-specific.3s-GEN book-ACC bring-NSR-3s man ‘the man whose daughter brought the book’ According to our assumptions.6 The option of the SR form In the relative clauses in (27). In the sentences in (27) both the SR and NSR forms are acceptable36.3s book-ACC bring-SR man ‘the man whose daughter brought the book’ b. TP] in the NSR clause in (27)b. it is the possessor of the subject. TP] in the SR clause in (27)a whereas a non-Wh DP must have been attracted to [Spec. i. [[Ø1 kız-ı]-nın kitab-ı getir-di -i] adam1 Ø girl-POSS. [[Ø1 kız-ı ] kitab-ı getir-en] adam1 Ø girl-POSS. We have determined that the SR form is licensed when a +Wh expression has moved to [Spec. my aim is to provide an account for the dialect where both forms are acceptable. 2. 36 33 . Let’s look at how this can be. Although there seems to be some dialectical variation as to the acceptability of both these forms. TP]. a +Wh DP must have moved to [Spec. 27) a. 2. This can happen in RCs where the subject has not been relativized only when the subject is an NP and not a candidate for the EPP or Case.e. the gap site is not the subject.5 Optionality of RC forms Turkish RCs seem to permit a certain degree of optionality in the choice of verbal paradigm. is an NP and cannot satisfy T’s EPP.

book. vP] because it is specific and therefore a DP. under our assumptions. the possessor. the direct object. But. man. Thus. it would be attracted to [Spec.In the SR RC in (27)a. In (27)a. TP]. TP]. 28) the SR form CP TP +Wh-man vP book-ACC NP VP +Wh-DP-man DP V° book bought daughter+Agr N° v° vP T° C° 34 . a +Wh element must have moved to [Spec. we must assume that it was the +Wh possessor of the subject that moved to [Spec. Let’s look at the derivation of (27)a. in the Spec of the subject is a DP with +Wh features. note that in both examples in (27). only the +Wh element from within the subject must have raised to T. The +Wh DP-man. must raise to [Spec. CP]. in . man’s daughter. is attracted by the EPP of T°. because the SR form is acceptable. Since the subject itself is not +Wh—only the possessor man is—the SR form would be barred. If the subject [man’s daughter] in (27)a were a DP. the subject cannot have moved to [Spec. is a non-specific NP. The subject itself must be a non-specific NP that has not raised from its base-generated position. In the tree in (28) for (27)a. The entire subject. after which it moves to [Spec. First. the object was attracted by the EPP of v and has had its case checked/assigned. TP]. in (27)a. TP].

the subject is not a +Wh DP. man. The subject man’s daughter must therefore be a DP in this example. This is because if the entire DP subject does not raise to T. The Wh-element. The EPP of T targets the closest DP. in (29). the element in [Spec.7 The option of the NSR form According to our generalization. TP] then moves in to C. Thus. this is what must have moved to T.) The entire DP subject is attracted to satisfy the EPP of T in . (It also receives overt accusative case. This is shown in derivation (29) for (27)b. in the specifier of the subject in [Spec. Thus the SR form is not licensed.Although this account may explain the acceptability of the SR form. it will remain 35 . this derivation will give us the wrong word order: [t1 book-acc [[t1 daughter]] man1. if the DP-man raises out of the subject to T. Crucially. 2. We will return to this problem shortly. the DPman in its specifier is. TP] must be a non-Wh DP. being a DP. in its Spec. Since the subject of the clause is not a +Wh expression. the derivation won’t converge. raises to [Spec. Let us assume for now that the subject and the element in its Spec are equidistant from T. In move . and thus both candidates for Attract by T’s EPP. the direct object book. T’s EPP can be satisfied by both the in situ subject as well as the expression. man. And it looks like this movement violates Minimality (at least under some definitions). vP]. Let’s look at the derivation in (29) for the NSR example 27)b. This differs from the tree in (28) where the +Wh DP was a specifier of an NP. The DP subject has its case assigned/checked by T. in the NSR clause in (27)b. However.

38 Perhaps a more problematic issue is that in . examples in the literature regarding the so-called Subject Condition effect are not definitively unacceptable. ?*Which booki did you say [[the author of ti ] was very eloquent]? c. the EPP of C must specifically target a +Wh DP for convergence. are not any more degraded than Stowell’s examples. violating the Case Filter.) a. TP] to avoid a Case Filter violation. I am assuming that. ?*Whoi do you consider [[the oldest sister of ti ] to have left]? b. in (29). Stowell (1991) includes sentences such as (i) with the following proviso: “Although there is some variability in the judgements …” (i. Furthermore. which should be worse because we are extracting out of the subject of a finite verb. (ii. TP] is barred for the same reason (the subject would remain without case). ?*Who do you feel [[John’s having visited ti ] was very unwise]? 36 .without case. This assumption and possible variations will be discussed in later chapters. can only attract a +Wh element. ?*Which booki did you find [[the author of ti ] very eloquent]? c. vP].37 The EPP of C°. ?*Whoi do you believe [[the oldest sister of ti ] left]? b. as in .man TP DP vP man-GEN +Wh daughter D° book DP T° C° VP book v° V° bought 37 The evidence from the strict correlation between structural case assignment and displacement leads to the conclusion the structural case in Turkish must be assigned in a Spec-head configuration. on the other hand. the subject DP must raise to [Spec. Whereas there are more options for the EPP of T.38 29) the NSR form CP +Wh. of the two DP’s (the subject and the object) in [Spec. raising of the +Wh DP from the specifier of the subject to [Spec. Even in English. at least under certain conditions.) a. subjects are not islands in Turkish. in his discussion of the SC in extraction from non-finite clauses. the +Wh element was a constituent of the subject. In sum. For example. ?*Who do you judge [[John’s having visited ti ] very unwise]? I find that the sentences in (ii).

note that the EPP of T can be satisfied by the DP subject. accusative. Thus. the NSR form must be used because [Spec. the DP [trace-man daughter] will remain without case. all other cases in Turkish are overtly case-marked. while a structurally case-marked expression must be a DP—the structural cases being nominative (with the phonetically null Ø morpheme). TP]. When the subject is non-specific. Let’s turn our attention now to the specifier position inside DPs and NPs. If the relativized element cannot move into [Spec.8 Genitive case Recall that we assumed that DPs require case and that NPs do not. Before we can resolve the word order problem noted in the derivation of (27)a in (28). Except for nominative case. in its Spec. if the DP-man raises out of the subject to [Spec. the SR form will be barred. it cannot be a target of T’s EPP and [Spec. we need to look at case assignment inside NPs and DPs. To put it another way. and the derivation 37 . the choice of the relative clause form depends on whether [Spec. man. [man’s daughter] as well as the expression. Referring back to the example in (29). the element that checks T’s EPP feature must be the subject which also receives case from T. 2. We will use this overt case marking as a diagnostic for DP’s: a nominal expression without case morphology is an NP. when the subject is a DP. TP] will be available for another DP. and genitive. TP] will not be available for that element to move into. However. TP] is available for the +Wh element or not. no matter what non-subject nominal is being relativized. its in situ remnant. TP].To review what we have determined thus far: whenever the subject is specific.

a case-marked element. 30) a. note that a DP in the specifier of an NP is bad. And finally. that there is agreement between the possessor and the possessee regardless of whether either gets case. these facts are contra Chomsky who assumes that case and agreement go hand-in-hand. This is similar to the asymmetry I assume with nominative case on T. In Chapter 5. we must assume it is a DP. in (31)d. I argue that this is a (possessive) agreement morpheme. In (31)b. A subject DP must be assigned case by T. As we will see throughout this thesis. On another note. Looking now at the examples in (30). we may be able to say that case-marked elements must agree but that the converse need not hold: agreeing elements need not be case-marked. we have a DP with an NP in its specifier. Thus. In (31)a. in RCs. but T doesn’t have to assign case. we have a DP with a DP in its specifier. therefore a DP. walls. Note in (30). we have a nominal. ehir duvar-lar-ı city wall-pl-AGR ‘city walls’ ehir-in duvar-lar-ı city-GEN wall-pl-AGR ‘walls of the city’ [[NP city] walls] b. When city has genitive case. In (31)c. [[DP city-GEN] walls] In (31) I have listed all possible NP/DP combinations. TP] where it satisfied T’s EPP feature and was assigned case by T in a Spec-Head configuration. This is consistent with what we have determined thus far about Turkish nominals. as well as in (31). city is in the specifier position and can be either case-marked or not. Turkish does not permit case assignment via Agree. it must also receive a specific interpretation. in (30)b. In Turkish.won’t converge.39 In (30)a. is in the specifier position. and that it has raised to [Spec. we have an NP with an NP in its specifier. with an NP in its specifier. 38 . when a subject bears overt genitive case. 39 The morpheme I call ‘AGR’ is frequently referred to as a compound marker in the literature.

Notice that derivation (32) still gives us the wrong word order for (27)a repeated below as (33)a.’ ehir duvar-lar-ı-nı gördüm. the surface word order of (32) will be as in (33)b. city wall-pl-AGR-ACC see-PST-1s ‘I saw the [city walls]. it will remain without case and the derivation will crash. [DP [NP city] walls]-ACC d. TP] as in (32).’ [NP [NP city] walls] b. If the DP does not raise. [DP [DP city-GEN] walls]-ACC c. N° does not assign case.31) a.’ ehir-in duvar-lar-ı-nı gördüm.9 Clearing up the SR option Looking back at the derivation in (28). 39 . 2. city-GEN wall-pl-AGR-ACC see-PST-1s ‘I saw the walls of the city. Hence.’ *[NP [DP city-GEN] walls] These facts lead us to conclude that whereas D° assigns genitive case to a DP in its specifier. Failure to raise will result in a Case Filter violation. a DP in its specifier cannot get case and must raise to a case-assigning head in order to receive case. when the subject of a relative clause is an NP. ehir duvar-lar-ı gördüm. * ehir-in duvar-lar-ı gördüm. we can now see why the DP in the specifier of the NP subject had to raise to [Spec. Assuming that man is further promoted to the RC external head position. city-GEN wall-pl-AGR see-PST-1s ‘I saw walls [the city]. city wall-pl-AGR see-PST-1s ‘I saw city walls.

Hornstein and Witkos’ (2001) analysis of transitive expletive constructions (TECs) offers a possible solution. They argue that existential constructions are formed by the merge of the expletive and the associate. [[Øi kız-ı] kitab-ı getir-en] adami Ø girl-POSS. Furthermore. what happens in TECs is that the object and the [expletive-associate] pair are both specifiers of vP at the point when T merges 40 . This move seems to violate Minimality because although the entire subject and the object are now specifiers of the vP. it is not so clear that the element that is a constituent of the subject is in the same minimal domain as the object. and thus equidistant from T.32) the SR form CP TP DP-man +uWh.3S book-ACC bring-SR man ‘the man whose daughter brought the book’ b. the DP-specifier of the NP-subject.3S bring-SR man In (32).+Case vP book-ACC NP VP DP-man +uWh. leaving the NP-subject in its base-generated position inside the vP.-Case DP book daughter+Agr N° v° V° bought vP C° T° 33) a. TP]. was attracted by the EPP of T. and the overt movement of the expletive to [Spec. *[kitab-ı [[Øi kız-ı] getir-en] adami book-ACC Ø girl-POSS.

[NP [DP man-GEN D°] daughter]. the whole NP. In (34)a. one way around violating Shortest Move would be as follows: we could say that although the DP within the NP is attracted by the EPP of T. vP]. This movement is allowed because the NP is equidistant to [Spec. CP]. the [expletive-associate] pair can move to that position. TECs do not exist in English because movement of the expletive from the [expletive-associate] pair will violate Minimality because the expletive (which is a constituent of the [expletive-associate] pair) is not in the same minimal domain as the object. For (32). that is being attracted to [Spec.3S-GEN book-ACC bring-NSR-3s man ‘the man whose daughter brought the book’ 41 .with the vP. 34) a. whose. This movement is similar to that of whose book in sentences such as “Whose book did you borrow?” In this sentence. but pied-piping of the remnant of the category allows for the convergence of a derivation that would otherwise crash with the movement of whose alone. [[Ø1 kız-ı ] kitab-ı getir-en] adam1 Ø girl-POSS. I repeat the relevant examples in (34).3S book-ACC bring-SR man ‘the man whose daughter brought the book’ b. It is pied-piped by the DP-man. T] being in the same minimal domain as the fronted object now in [Spec. it is the element with the +Wh feature. whereas in (34)b. the specifier of the subject receives case from T. moves to T°. in languages where another projection is available above the vP. the entire DP subject receives case. [[Ø1 kız-ı]-nın kitab-ı getir-di -i] adam1 Ø girl-POSS. from which the expletive is now free to move without the issue of minimality. On the other hand.

it is the DP in the Spec of the NPsubject in (34) (35)a that is attracted to T. pied-pipes to [Spec.Adopting the Hornstein and Witkos proposal. TP]. shows that non-specifics in Turkish cannot scramble. the +Wh-DP in the Spec of the subject. NP] is attracted by the EPP of T. it would have received overt genitive case.e.40 In the analysis being proposed here. rather than the entire subject that was attracted to T. that it was only a constituent. One might argue that the entire subject has scrambled to a position higher than the object but lower than [Spec. the SR form is 40 I have not addressed the possibility of scrambling. among others. 35) The SR form with pied-piping of the subject CP +Wh-man TP vP book-ACC NP VP DP-man-GEN +uWh N° v° vP C° T° DP V° book bought daughter+Agr Another reason for proposing this sort of analysis. Karimi (2005) also shows that scrambling is not possible for non-specifics. is because had the subject been attracted by T’s EPP. man’s daughter. TP]. 42 . case and displacement as Turkish. i. Persian shares many of the same phenomena regarding specificity. I reject this idea because Kural (1992) and Kornfilt (2003). as in move . This strategy of avoiding the minimality violation noted for (32) is shown in (35): When the DP-man in [Spec. In her study of scrambling in Persian. and the NP-subject is pied-piped with it. the entire NP.

[[Ø1 biz-e güven-ece -i]-nin Ø 1p-DAT trust-FUT-COMP-POSS.licensed when the expression that is Case-marked by T is also +Wh. [CP]. 41 43 .10 Relative clauses with complex arguments We can now extend our analysis to RCs with even more complex arguments. Although the entire NP-subject is presumably sitting in [Spec. and elsewhere. TP]. where there is a null element “fact” in Nº whose complement is the CP “that [the man] will trust us”.al. to a predicate of states. it is the +Wh-DP in its Spec that is receiving case from T.3S-GEN doubtfu be-NSR-POSS. Relative clauses with sentential subjects permit both RC forms.3S doubtful be-SR man ‘the man who/such that (he) will trust us is doubtful’ b. 37) NP VP V° is. et. [CP – Nº].41 36) a. 42 This is akin to transforming a predicate of individuals. [[Ø1 biz-e güven-ece -i] üpheli ol-an] adam1 Ø 1p-DAT trust-FUT-COMP-POSS.(1990). I propose that the structure of these clauses is as in (37).doubtful N° (fact) CP [the man] will trust us A discussion of these examples first appeared in Csató (1985) and are again discussed in Barker.3S man üpheli ol-du -u] adam1 ‘the man who/such that (he) will trust us is doubtful’ Note that the subject in the RCs in (36) is something akin to “the fact that [the man] trusts us”. as shown in (36).42 It is thus the “factive” NP that receives the theta-role from the predicate “is doubtful”. 2.

38) NP mani +uWh VP V° is. It is possible that the verbs that allow sentential subjects. There is a null resumptive pronoun (RP) bound by man in the subject position of the “factive”-CP. themselves cloud the semantic effects of the old information-new information distinction that the NP-DP subject should entail. is doubtful. is not base-generated inside the “factive”-CP. [DP man [NP t-man [CP …] Nº ] D°] is in [Spec.This structure is similar to the null nominal head selected by factive verbs as analyzed by Kiparsky and Kiparsky (1971). is certain. for the NSR phrase in We would expect there to be a semantic difference based on whether the “factive” subject is an NP or a DP. note that (38) better represents the structure of the RC subject for the RCs in (36). the subject can be an NP as in (38) or the “fact”-clause can be embedded in a DP43 in which case the +Wh DP man will raise from [Spec.doubtful N° (fact) C° T° VP trust CP TP RPi VP us-DAT t-RP Furthermore. but rather is first-merged as the specifier of the NP. the entire DPsubject is attracted by the EPP of T and assigned genitive case. The derivation in (39). man. Once this subject. but see Section 2. man. CP]. raises to [Spec. for now. the +Whelement. I will justify this assumption a little later. I assume that the +Wh head of the RC. When the subject is a DP. Some speakers do feel a slight difference but I have not been able to pin down a definitive diagnostic that will yield consistent results.11 where time adverbial scopal differences are noted. TP] of the relative clause. 43 44 . NP] to the specifier of the DP (and get genitive case).

for example (36)a: to avoid violating the Case Filter. demonstrates a RC with a DP subject. We saw that in (36)b the subject was a DP. 39) the NSR form (36)b CP +Wh-man TP VP DP DP-mani +GEN.(36)b. TP]. Again.+Wh t-mani CP TP RPi VP us-DAT t-RP VP trust T° N° (fact) C° C° T° V° is. prior to the promotion of the relativized element to the external head position. To be clear. the +Wh DP-man must 45 . we are looking at only the internal structure of RC’s. the subject is an NP which cannot satisfy the EPP and consequently does not raise to [Spec.doubtful D° NP The difference between clauses (36)a and (36)b can be attributed to whether the subject is a DP or an NP. on the other hand. the +Wh head of the RC is base-generated in the specifier position of the NP-subject. This is demonstrated in the tree in (40). As we saw in (31). In (36)a. a DP in the specifier of an NP must raise to the spec of a case assigning head. as has been the case throughout.

I adapt this analysis for Turkish which has neither overt determiners nor overt complementizers. it too does not require case. and must be visible for attraction by the EPP. It won’t matter whether the possessor of a DP subject is an NP or a DP because an NP doesn’t require case and a DP possessor will be assigned genitive case by the subject D°. relativized elements must be case-marked prior to A-bar movement. +case VP NP DP-man +Wh. pied-piping of the NP subject is not required because there is no competing DP.doubtful N° (fact) CP One more assumption must be clarified: I assume that all relativized elements are DPs because they must be specific. Let’s take stock and look at the possible options for subjects. However. Borsley (1997). licensing of parasitic gaps. as in [DP which NP]. 46 . and assume that the relativized element in Turkish is a null +Wh-Dº and its NP complement. CP] is a DP headed by a relative Dº. the element that moves to [Spec. See Bianchi (1999.raise to [Spec. the subject itself does not need case. 2000). and weak islands. A subject can either be a DP or an NP. demonstrates that the RC gap acts as a DP-trace with respect to binding. -case C° T° V° is. although disagreeing with Kayne’s raising analysis. a DP 44 Kayne (1994) proposes that in wh-relatives. it must raise to a case-assigning head. If the possessor in its Spec is also an NP. when the subject is an NP. Note that in (40). or have topic-like properties. TP] out of the NP-subject. On the other hand. When the subject is a DP. A possessor in its specifier can be either a DP or an NP. 40) the SR form (36)a CP TP DP-man +Wh.44 As DPs.

and does not need to move to an A-position for case. is not a +Wh element—a constituent is—and as expected the NSR form is required. With this in mind. TP] in to satisfy . The expression in [Spec. the entire subject. 42) the NSR form (36)b [CP [DP Ø1 biz-e güven-ece -i-nin] üpheli ol-du -u ] adam1 Ø 1p-DAT trust-FUT-COMP-POSS. The +Wh DP-man in Spec of the NP-subject must raise to [Spec. T’s EPP and to receive case. the subject of the relative clause in (36)b is a DP. The element in the Spec of the DP subject is assigned case by D°.3S doubtful be-SR man ‘the man who [such that] (he) will trust us is doubtful’ CP +Wh-man TP t-man +case NP DP-man +Wh C° VP T° V° is. let’s look at derivations (43) and (44) for the RCs in (36). Continuing derivation (40) in (41).3S man ‘the man who [such that] (he) will trust us is doubtful’ 47 . TP].3S doubtful be-NSR-POSS.doubtful CP N° By way of contrast. CP] in 41) the SR form (36)a [CP [NP Ø1 biz-e güven-ece -i ] üpheli ol-an ] adam1 Ø 1p-DAT trust-FUT-COMP-POSS. Derivation (39) repeated as (42) demonstrates that the entire DP-subject must raise to receive case. The relative DP-man then moves to [Spec.possessor in the Spec of an NP subject must raise for case or violate the Case Filter. we see why the SR is required in (36)a.

[Pazartesi [[Øi kız-ı-nın] Monday daughter-poss3s-GEN A.11 Semantic reflex of syntactic structure In (41) and (42). We would expect there to be some semantic reflex of the differences in structure. the adverb can only modify the verb of the embedded RC subject. 43) a. The word order of the RCs in (43) is identical.CP +Wh-man TP VP DP DP-man-GEN +Wh CP C° T° V° is.-DAT go-NSR-3s understand-PASS-SR man ‘the man (such that) on Monday it was discovered that his daughter went to Ankara’ Ankara-ya git-ti -i] anla -ıl-an] adami 48 . Monday. we are making a claim about the category of the subject in the matrix RC which has consequences in terms of the syntactic position the subject will occupy. whereas in (43)b Monday modifies the matrix RC verb. [[Pazartesi [Øi kız-ı-nın] Ankara-ya git-ti -i]-nin Monday daughter-poss3s-GEN A.-DAT go-NSR-3s-GEN adami anla -ıl-dı -ı] understand-PASS-NSR-3s man ‘the man (such that) it was discovered that his daughter [went to Ankara on Monday]’ b.doubtful D° 2. but in (43)a. This does seem to be the case as demonstrated in the examples 43) which contain the time adverbial.

Monday merging in T of the sentential subject will give us the right word order but will yield an interpretation where Monday can only be interpreted as modifying the embedded verb. In (44).+uWh t-mani CP N° (fact) C° TP DPj VP t-DPj Ankara RPi NP daughter D° V° go T° T° Voice° (Passive) V° discover D° NP TP Monday 49 . there are two positions in each derivation in which Monday can merge: in the embedded sentential subject or in the RC (which I will call matrix) T. the event of going to Ankara occurred on Monday. we would get the wrong word order because the sentential subject will raise above Monday. Assuming the adverbial expression merges in T. for the NSR in (43)a. 44) the NSR form (43)a +Wh-man TP k CP C° VoiceP VP DPk DP-mani +GEN. On the other hand. if Monday had merged in the matrix T.The derivations in (44) and (45) demonstrate where the difference in interpretation comes from. yielding [[his daughter [ Ankara go]] Monday].

Monday merging in matrix T will give us the correct word order and the interpretation that Monday modifies the RC verb discover. yielding the phrase [[his daughter]1 Monday [vP t1 Ankara go]]. and 2001) suggestion that that in addition to vP and CP. In the SR form in (43)b. but my informants were not able to get a reading where Monday would modify the embedded subject verb “going to Ankara” with the word order of (43)b. the T of the embedded subject and the matrix (RC) T. Another way of looking at the inability of the adverbial to modify the verb.As shown in the tree in (44). it is too far away to modify it. DP is also a Strong Phase. Compare this with the position of the time adverbial in the tree in (45) for the SR example in (43)b where Monday remains below the matrix (RC) T. The complement of C is Spelled-Out when D merges in the structure of this complex subject. one where the subject of the sentential subject raises above Monday. the time adverbial Monday raises with the entire subject to a position higher than the RC verb discover.45 The reading where the event of going happened on Monday is not available in this structure. Monday merging in the subject should conceivably also be possible. As shown in (45). and crucially fails to c-command the matrix (RC) verb. again the adverb can possibly merge in two positions. 45 This inability for the time adverbial to modify the embedded verb is straightforward in a Phasebased story. As can be seen in the tree in (45). although Monday c-commands the embedded verb (in the in situ sentential subject). The interpretation that the going to Ankara will occur on Monday was only possible with the SR form with a different word order within the clause of the sentential subject. This story supports Chomsky’s (1999/2001. there is a VP. NP and CP between the time adverbial and the Tense and verbal projections of the sentential subject. 50 . would be to say that adverbial modification obeys Subjacency.

” 51 . there is no ambiguity in the interpretations of the RCs in (43) when spoken with normal intonation. “that he will sell his house”.46 That is. that the girl will go to Boston t1 . I have highlighted the adverb galiba 46 That is. the adverb can only modify the embedded verb of the sentential subject. The relative clauses in (46) are passivised expressions.45) the SR form (43)b CP +Wh-man TP DP-man TP +GEN. Pauses would denote a scrambled position similar to “We found out. Let’s look at another example. and in (43)b. The phrase that was the complement of the RC verb. without pauses. the adverb can only be interpreted as modifying the matrix (RC) verb. It is the subject of this sentential subject that is the relative head. [next week (it is)]1. has become the sentential subject of the RC. in (43)a.+Wh Monday C° TP T° VoiceP VP NP DP-mani +uWh Voice° (Passive) V° discover CP N° (fact) C° T° TP DPj VP t-DPj Ankara RPi NP daughter D° V° go For those speakers whose dialects permit both RC forms.

In (46)a. In (47)a. In (47)b. Again. and is coindexed with the null resumptive pronoun subject of the sentential subject. I am abstracting away from much of the structure of this phrase. and has scrambled to TP adjoining above the sentential subject. Notice the interpretations: the adverb modifies the RC verb in the NSR form in (46)a whereas the adverb modifies the sentential subject verb in (46)b. I have highlighted the sentential subject in bold-type. [ galiba adami [Øi [proi ev-i]-ni sat-aca -ı]-nın söyle-n-di -i ] ‘the mani who that (hei) will sell hisi house was apparently announced’ apparently Ø pro house-AGR-ACC sell-FUTNSR-3s-GEN tell-PASS-NSR-3s man b. We are still assuming that the sentential subject is a DP or NP Factive. galiba was generated in vP/PassiveP. Let’s take a closer look at the positions of galiba in (47). [Øi [galiba [proi ev-i]-ni sat-aca -ı] söyle-n-en] adami Ø apparently pro house-AGR-ACC sell-FUTNSR-3s tell-PASS-SR man ‘the mani who that (hei) will apparently sell hisi house was announced’ The example in (46)a is particularly interesting because even with changed intonation and pauses. the adverb cannot be interpreted in the sentential subject. while the +Wh-expression has raised to [Spec. 47 52 . is in the verbal domain. and simply labeling it FDP or FNP. the sentential subject. FNP. In the illustrations in (47). but for ease of exposition. galiba has scrambled and adjoined to TP. I am using the Ø symbol to denote the +Wh-expression man with the understanding that this expression is generated in the specifier of the Factive phrase. we are assuming normal intonation without pauses which would reflect scrambling or discourse driven movement.‘apparently’ and the verb it modifies. FDP. such as Topic or Focus. but it must be interpreted in its base position.47 46) a. It seems galiba cannot merge in TP.

47) a. I will adopt a version of Pesetsky and Torrego (2001) (P&T). uT. As expected. Using an adaptation of 53 . More concretely. which must be deleted by T. They begin by taking do-support in English as evidence that C must contain an uninterpretable T feature. [CP Øi [TP galibaj [FDP Øi [proi evi]-ACC sataca ı]-GEN [vP/PassP tj FDP söylen-NSR]]] b. CP].TP] and then to [Spec. P & T (2001) use the principle of Economy to explain the that-trace effect. [CP Øi [TP Øi [vP/PassP [FNP Øi [vP galiba [proi evi]-ACC sataca ı]] söylen-SR]]] adami These examples seem to provide further support that the structure being suggested for the two RC forms is on the right track. 3 A minimalist account: Pestesky and Torrego (2001) In this section I would like to present one theoretical account for the facts in Table 1 within the framework of Minimalism (Chomsky (1995. (47)b is ambiguous and allows the alternative reading where galiba modifies the RC verb ‘was announced’. 2000)). but this interpretation requires a pause after galiba.

When the +Wh expression of a sentence is not also the subject. and embedded Cº that hosts successive cyclic Whmovement. ATX will force T-to-C movement. unless this other element is just as close (i. 54 . Interrogative Cº. will always precede movement of another element to check C’s uWh feature. TP]. the local movement from the TP to delete C’s uT feature.e. In effect. Attract Closest X (ACX). ATX applies vacuously. P&T propose a tighter version of Chomsky’s (1995) Attract Closest F. and either element is a candidate for movement. have an additional feature. and 2. The consequence of ATX is that C’s uT feature must always be satisfied before its uWh feature. Because the TP is the closest projection to C. the movement in (2-) can precede the movement in (1-) only when the +Wh-element is in [Spec. the uninterpretable features of C° can be satisfied by: 1. this constraint imposes ordering on feature checking of C.Travis’s (1984) Head Movement Constraint. P&T assume that uT on C must be deleted by movement of Tº to Cº. The features of C target the closest element with matching features. In P&T’s system.e. within the same minimal domain). Both these features. uT and uWh. CP] to check uWh. uninterpretable Wh (uWh). a +Wh subject that can check uWh and T° that can check uT). host an EPP feature. C° always has an uT feature which can be deleted by head movement of Tº to Cº.movement of a Wh-element to [Spec. When two elements in TP can check different features of C° (i.movement of T° to C° to check uT. which prevents a head with multiple uninterpretable features from targeting via Attract across any element that could potentially delete one of its features. In sum.

in the sentence Who bought the book? as follows. we need to revise our summary above adding the additional way that C’s uT can be checked. in fact.movement of T° to C° or movement of a nominative DP to [Spec. and 2. called “nominative”. 48 Beginning with the derivation for the sentence What did Mary buy? in (48). TP] is +Wh. even though there are two ways to delete C’s uT. But. but see fn. The uninterpretable features of C° can be satisfied in the following ways: 1.movement of a Wh-element to [Spec. I assume this is because a non-Wh expression is barred from the CP projection. the nominative subject cannot delete uT on C° by moving to [Spec. Motivated by a desire to unify nominative case on DP and agreement on T. T-to-C movement is required unless the element in [Spec.P&T explain lack of do-support. for example. according to P & T. P&T further assume that once a feature has been checked. So. in English yes-no questions. called “agreement”. and T properties borne on D. CP]. CP] in the same way that T° to C° movement can. 49 and 52. Both T and D have uninterpretable features that once checked by the other result in D properties being borne on T. uT on D. This means that although the uT features of the subject DP have already been checked by T°. let’s review P&T’s explanation of the subject/non-subject asymmetry of do-support in English interrogatives. CP] to check uWh on C°. P&T do not offer an explanation as to why. 55 . CP] to check uT on C°. it is “marked for deletion” but remains “alive” for further operations until the end of the (strong) phase. 48 The required head movement of T° to C° seems rather stipulative to me. P&T argue that nominative case is. The outcome of this analysis is that a nominative DP is able to delete the uT on C by moving to [Spec. or T-to-C movement. they are not deleted until the end of the CP phase.

uT is obligatorily deleted by head movement of T.Although there are two ways in which uT on C can be deleted..” P&T never explain why T-to-C movement is obligatory. CP]. One factor that may explain this is unique specifier positions. In discussing the ungrammaticality of “*What Mary bought” where the nominative subject has deleted uT on C. We suspect that this is not the right approach. nor the +Wh-expression’s Wh features could be checked.. P&T’s analysis (of sentences in which the Wh-element is not the subject) rests on the assumption that in matrix interrogatives. if a nominative subject occupied [Spec. movement of the +Wh-expression to [Spec.. movement feature of C. 48) What did Mary buy? CP C° did Mary T° VP V° bought what TP Central to Pesetsky and Torrego is their Economy Condition (49) based on the generalization that heads enter into Agree and Move relations only to the extent necessary. P&T state: “The obligatoriness of T-to-C movement … might lead us to search for a factor that favors T-to-C over subject movement. the Wh-element moves to [Spec. and conclude the discussion with the following statement: “We . The evidence in Turkish certainly points to unique specifier positions in the functional projections. leave it as an observation for further research…. that movement of the nominative subject to C is available as an alternative to T-to-C movement — even in matrix clauses headed by a C that contains uWh.49 After T moves to C. 49 56 . CP] would not be possible with the result that neither C’s uWh. that is.. movement . CP] to check the uWh .

(50)b. the movement of T to C is obligatory (see fns. This condition plays a crucial role in sentences where the Wh-element is the subject. The Economy Condition disallows T-to-C in (50)b because movement of the +Wh-subject results in a more economical derivation: all features of C are checked with one move. movement of the Wh-subject who from [Spec. wins out over an alternative derivation. Even though both T° and the DP in [Spec.49) Economy Condition A head H triggers the minimum number of operations necessary to satisfy the properties (including EPP) of its uninterpretable features50. TP] to [Spec. 57 . However. which requires more moves to check features. on this view. Recall that uT on C° can be deleted in one of two ways: i) by movement of the nominative DP (carrying its still “alive” uT feature) from [Spec. Who bought the book? CP C° who +uT +uWh TP T° VP V° bought the book 50 Note that. 48 and 49). 50) a. CP] can check uT as well as uWh on C. TP] are in the same minimal domain and thus candidates for Attract Closest. (50)a. or ii) by head movement of T° to C°. the EPP is an uninterpretable feature. TP]. as in derivation (50)a. The derivation which converges with less moves.

they propose that that is an instantiation of T-in-C. Again. Economy dictates that this derivation be chosen over the less economical one where the separate features of the embedded C are checked by two separate moves. *Who did buy the book? (with normal intonation) CP C° who +uT +uWh TP T° VP V° bought the book P&T also use the Economy Condition to explain the that-trace effect. the uT feature of C must be checked by T-to-C movement. 58 . the declarative C of embedded clauses that hosts successive cyclic-wh-movement bears uT and uWh features. For P&T. More 51 Although not relevant for Turkish relatives. an option not available in matrix clauses. rather than via T-to-C. Rejecting the traditional view that that is a complementizer merged as a sister to TP. C is null in English.51 51) Whati did John say [CP t-whati [T that]j+[C] [IP Mary willj buy t-whati]]? In a sentence such as Whoi did John say [CP t-whoi [TP t-whoi bought the book ] ] where the Wh-word is the embedded nominative subject. when the Wh-phrase is not the subject. movement of the subject to the embedded CP can simultaneously check the uT and uWh features of C. but T-in-C in this embedded environment is pronounced as that. Nonsubject wh-movement is demonstrated in derivation (51). each of which also bear EPP features.b. P&T account for ‘that deletion’ by allowing the nominative embedded subject to delete uT on C. In their system.

and propose that. –DIK. (Standard English) (ii. and the –K morpheme.e. i. is precluded in this instance because movement of who to [Spec. uT on the embedded C is deleted via movement.) I wonder which dish that they picked. I assume that the Turkish NSR verbal morpheme. They point to the dialectal difference in the Belfast English example in (ii). adam-ın gele-ce -i gün man-GEN come-FUTNSR-3s day ‘the day the man will arrive b. as in (52).specifically. choose the move that maximizes the number of features checked.52 3. all) features on C. adam yarın gele-ce-k man tomorrow come-FUT-3s ‘The man will come tomorrow. we can view Economy here as a local valuation: at a given point in the derivation. CP] checks more (in fact.1 Pesetsky and Torrego (2001) analysis applied to Turkish relative clauses Although Pesetsky and Torrego did not address relative clauses. –DI. (i) I wonder what Mary bought. also found in uninflected infinitival verbs. that to C. is a compound of a Tense morpheme. it is accomplished via Agree. ye -me-k eat-INF-K 53) a. This view is supported by the fact that the NSR form also allows the future tense as in (53)a. T to C movement. in (ii). their analysis can be extended to Turkish relative clauses. (Belfast English) 59 . gel -me -k come-INF-K b. 52) a. whereas in (i). Note that –cE is the future tense morpheme in matrix sentences (53)b.’ 52 P&T suggest that the lack of T-to-C movement in sentences such as (i) is due to the absence of an EPP feature on the embedded interrogative C in Standard English.

a subject that is being extracted contains both uT and uWh features. Note that for me the -K morpheme is specifically an instantiation of Tº in Cº. Movement of the subject to [Spec. +uWh] [TP [girl+uT.I propose that -DI. 54) a. *[ hediye-yi ver-di -ı ] kız gift-ACC give-NSR girl ‘the girl who gave the gift’ d.is Tense specified for Past and that the -K morpheme is a reflex of T to C movement. we would predict that whenever the subject is the relative head. The derivation in (54)b of the clause in (54)a demonstrates that the nominative subject girl can check both uT and uWh features of C. T to C movement has deleted uT on C.53 In the Pesetsky and Torrego story. +uWh ] [TP [DPgirl ] T [VP bought the gift ]]] +T. which would have been deleted in any event by the obligatory movement of girl to [Spec.e. see Kural (1993) who also argues that -DI is past tense and -K is Cº. [CP [Cº +uT. [ hediye-yi ver-en ] kız gift-ACC give-SR girl ‘the girl who gave the gift’ b. i. CP] deletes uT on C and renders the movement of T to C superfluous. If we assume that the -DIK (NSR) form is indicative of uT features of C being checked by T to C movement. +uWh] T [VP gave the gift]]] c. [ [C+uT. In the illicit (54)c. 60 . +uWh 53 For a detailed discussion. it signals T-in-C. CP] to check the uWh features of C (54)d. we would never see the -DIK (NSR) form because T to C movement would be an additional unnecessary move.

specifically T.2 The NSR -DIK form Let us now go back to the NSR form.P&T’s analysis can be straightforwardly applied to account for the NSR -DIK morpheme in non-subject RCs. The subject has both uT and uWh 61 . the NSR -DIK morpheme. There is no +Wh element in [Spec. 3 in Table 1. The subject lady is attracted to [Spec. 55) *[Øi divan-da otur-du -u] bayani Ø sofa-LOC sit-NSR-3s lady Intended: ‘the lady that is sitting on a/the sofa’ CP lady TP t-lady +uT +uWh DP t-lady +uT +uWh C° VP T° PP/DP on the sofa V° sit In (55). to check its uT feature. 2 and No. TP] to outcompete T-to-C. No. movement . The C head simply targets the closest head. 3. Using P&T’s intuition. an instantiation of T having moved to C. TP] by T’s EPP feature. we can see how Economy accounts for the unacceptability of the NSR form in sentence (2)b repeated with its derivation in (55). and is assigned nominative case. is unacceptable.

there is no alternative derivation that would converge with fewer moves. TP] to satisfy T’s EPP feature. 56) [bayan-ın Øi otur-du -u ] divani lady-GEN Ø sit-NSR-3s sofa ‘the sofa that the lady is sitting on’ CP sofa +uWh lady +uT DP t-lady +uT +uWh TP C° VP PP/DP sofa-LOC T° V° sit In the tree in (56). of T° to C° is redundant. and so disallowed. move. CP] deletes these two features. When a non-subject is being extracted. Because the subject does not have a +Wh feature it cannot check uWh on C. 62 . again the subject lady is assigned nominative case by T and is attracted to [Spec. more economical. because when the extracted element is not the subject. CP] is required. the NSR -DIK morpheme is acceptable in (3)a. On the other hand. there is no alternative.features. movement . repeated as in (56). This derivation requires two moves to delete both uT and uWh on C. Movement of the +Wh DP sofa to [Spec. The movement of T to C is obligatory for convergence as it must check C’s uT feature. in Movement movement of the subject to [Spec.

the +Wh-element in the verbal domain has moved to C. For P&T. this violates Shortest Move in P&T and should be banned. movement of the +Wh-subject deletes both features of C simultaneously. Let’s begin with the first complication. 57) [Øi divan-da otur-an] bayani Ø sofa-LOC sit-SR lady ‘the lady who is sitting on the sofa’ To be clear. as in the SR example (5) repeated as (58) (item 4 in Table 1)? Rather than T moving to C. As demonstrated in (55). [gemi yana -an] liman ship sidle-SR harbor ‘the harbor that a ship is sidling up to’ 63 . the crucial feature is uT on C. Whereas the P&T analysis explains the asymmetry in simple relative clauses in Turkish. Thus (55) is aptly the derivation for the SR sentence (2)a repeated in (57). How do we explain the acceptability of the SR form when a non-subject is relativized. Economy dictates use of the -An SR form and bars the -DIK NSR form. CP]) when the subject is not a Wh-element.3. and they rely on Economy to disallow T to C movement when the subject is a Wh-element. P&T rely on Shortest Move to explain the obligatory movement of T to C (rather than the alternative nominative-DP movement to [Spec. making T-to-C movement unnecessary. We will need to make adjustments to the P&T analysis.3 The SR form We saw that in a subject relative. 58) a. with all moves driven by the EPP. it cannot account for the exceptions (numbers 4 and 5 in Table 1) nor can it explain the optional cases.

In sum. non-specific nominals are NP’s that lack a D projection. An uninterpretable Case feature on T that remains unchecked or undeleted cannot lead to a derivational crash under this view. CP harbor +Wh TP VP NP ship C° T° DP V° t-harbor+DAT sidle. we saw that Enç’s (1991) work was of interest because she demonstrated the correlation between specificity.54 Based on this complementarity. I assume a similar structure for Turkish.b.to +Wh In section 1. 56 I realize this complicates the issue of Full Interpretability. In a sentence that has no DPs as arguments. specific nominals are DP’s.55 Conversely.up. Longobardi (1994) provides evidence for DP’s with a null D in Italian. I therefore assume that all DP’s must have their case checked/assigned but that all case-assigning heads need not necessarily discharge their case56. and because the EPP of T must be satisfied for convergence. T will always be assigning Case. 55 Turkish lacks determiners. for transitives) will have no DP on 54 We saw in (6) that case-marked objects must raise. 64 . non-specific nominal elements neither get overt case nor do they raise from their base-generated position whereas specific nominal expressions must raise and must bear overt case. raising and overt case-marking in Turkish. (9) and (10) demonstrate that the same is true for subjects. T° (and v°. Furthermore. I will discuss this option in more detail later. There is the possibility that T does assign Case to whatever element sits in its Spec. I assumed an NP/DP division based on specificity: in Turkish. so the D head in DP’s must be null. I consider the reciprocity between obligatory case assignment by functional heads (the Inverse Case Filter) and the Case Filter an unnecessary redundancy.3.

57 At this point. An example is the Turkish sentence in (59)a in which the sole argument. the subject. Therefore. being an NP. but I will return to this in greater detail later. 60) a. there-LOC one dog bark-PST ‘A dog barked [there]. being non-specific. Question: (Sokak-ta) ne ol-du? street-LOC what happen-PST ‘What happened (in the street)?’ Answer: [ora-da] (bir) köpek havla-dı.which to discharge its case feature. See Ortega (200). “What happened?” Crucially. for example in response to the question. The question remains though. has not raised to T. as in the response in (60)a. the subject. as in (60)b. [TP pro-there-LOC [VP [NP dog] bark ] PAST] This is analogous to locative inversion in Spanish. a locative must raise to satisfy the EPP of T. where there is evidence to suggest that because a bare NP subject cannot raise to a preverbal position. I am not assuming that T assigns Nominative Case to the Locative.’ b. I assume the sentence in (59)a contains a pro-form locative. It is the pro-locative that raises to T and checks its EPP feature. Note that this sentence would be quite odd without a context. how is the EPP of T checked in (59)a? As demonstrated in (59)b. this question entails a contextually relevant time and place. (Bir) köpek havla-dı. [TP [VP [NP dog] bark ] -PAST] The assumption that convergence requires that DP’s receive case but not that relevant heads assign case enables us to account for the acceptability of sentences with NP arguments. one dog bark-PST ‘A dog barked. must be an NP as in (59)b. 57 65 . 59) a.’ b.

I concur with Boeckx (2001) that Agree need not be a prerequisite for Move. whereas a non-specific one does neither. dog-GEN street-LOC bark-NSR(COMP)-AGR-ACC hear-PST-1s ‘I heard that the dog barked in the street. obligatory raising and case when it is specific. In Chomsky (2001).In Section 1. the Spec-Head relation is considered an outcome of Move which is defined as Agree + Pied-piping + (internal) Merge. Nominative case in Turkish is the Ømorpheme. EPP driven movement is comprised of Attract F (head adjunction of formal features FF) followed by pied-piping of the category for PF convergence (so that the category will be “close enough” to its FF so that the features of the category will not be scattered). thus allowing for the possibility of Move to take place in some cases under Match. we saw the correlation between displacement. An important generalization is that in Turkish a DP receiving structural case must raise to the specifier of the caseassigning head. Boeckx (2001) points out that not every DP can satisfy the EPP.58 61) a. However. These facts provide evidence that all DPs not only receive case.’ b.in sokak-ta havla-dı -ın-ı duydum. the DP must be “featurally related” to the EPP bearing head. presumably because case is assigned in a Spec-Head configuration. but must also raise to receive/check case.’ b. street-LOC dog bark-NSR(COMP)-AGR-ACC hear-PST-1s ‘I heard a dog barked in the street. 59). [pro [vP [CP [TP [DP dog]-GEN [VP street bark ] ] ]-ACC heard] 58 Referring to Chomsky (2001). Sokak-ta köpek havla-dı -ın-ı duydum. Köpe. note both the position of the subject and its case: no raising nor case on the subject when it is non-specific. In Chomsky (1995). 66 . [pro [vP [CP [TP street [VP [NP dog] bark ] ] ]-ACC heard] 62) a. specificity and case. In the sentences in (61) and (62). but similar facts were demonstrated for subjects in embedded environments where the subject is marked with overt genitive case. a looser requirement of “feature-relatedness” (see fn. Recall that a specific object must obligatorily raise and receive overt case.3 of Chapter 2.

Although not a topic of this dissertation. The proposal for Turkish then is as in (63). where absent a DP requiring structural case. Thus these sentences may have a Topic projection which selects T with an uninterpretable “Topic” feature. TP] is able to delete/check the uC feature on T61. Topic° will select for T° with an uninterpretable A-bar-like feature which itself will have an EPP feature. and uC when on T. One can imagine several plausible ways. When a non-Wh-element is in [Spec. none of which will detract from the proposal here.e. but see fn. Let us say that this feature is some sort of Wh.63 59 In Turkish matrix sentences.59 By the same token. the data we incur in later chapters suggests there may indeed be an EPP on vº as well as an A-bar projection above vP. however. I avoid a discussion as to whether v may or may not have an EPP feature. 63) Theoretical Assumptions i. iii. 67 . Wh-features are irrelevent for the operation Attract by T. The EPP feature on T is a feature of an uC feature. 60 The EPP of this feature attracts a D. let us call the matching features uT when on C. the specific subject has topic-like properties. TP] or by movement of T° to C°. it has an uninterpretable T feature that makes it select T. T has an EPP feature only when it has been selected by C. 61 It does not concern us here as to exactly how uC on T is deleted by a Wh-element in [Spec. that is. a Wh-element in [Spec. raised to T to satisfy its EPP feature. I assume that sentences such as (59) have neither a Topic projection nor an EPP feature on T. Following the reasoning for Tº. On the other hand. CP] and delete/check uWh on C. there may be an optional A-bar “topic-like” projection above vP that selects for v with an EPP feature.Deviating from Pesetsky and Torrego. TP].(i. this issue requires further inquiry. This contrasts with the embedded sentence in (61). T also has a comparable uninterpretable feature that must be checked. T must necessarily move to C because the uC feature it bears is still unchecked. That is. I propose that in Turkish relative clauses the EPP feature of T is simply a feature of some uninterpretable feature of C. I will point these out as we encounter them. For simplicity. 59. A-bar) feature60 because it was selected by C. Certainly. TP]. a featurally-related DP. The uC feature of T may be checked/deleted by movement of a +Whelement to [Spec.62 ii. T can assign case but need not. 62 This also holds for sentences with TopicP. after which it will raise to [Spec. the locative. if C is in the derivation.

a (inherently) case-marked element may be attracted by the EPP of T. when selected by Cº (or Topicº). as Tdef. vi. Let’s look closer at the consequences of assumption (ii) above. as in (i). (i) We expect there to be awarded several prizes. let’s look again at the SR in (57). The C° of a RC has an interpretable Wh feature to be checked. NP’s cannot be attracted by the EPP and so are invisible for movement. thus I will remain agnostic as to whether T actually assigns null nominative case to the already case-marked DP in its spec. has an EPP feature but need not necessarily assign case. (I take up the question of whether Tº must assign case in Chapter 4. non-specifics are NP’s. If this DP has a +Wh feature. it satisfies two features: uC on T and the EPP on uC. CP] checking uWh and satisfying C’s EPP feature. This is what is required for T°. but not uWh on C° which still must be satisfied by movement of a +Wh-expression to [Spec. I assume that finite T. As shown in (61). v. the only way for T to discharge its uC feature is by movement from T to C. We know that T’s EPP attracts the closest DP. Nominals that are specific are DP’s. 64) [Øi divan-da otur-an] bayani Ø sofa-LOC sit-SR lady ‘the lady that is sitting on the sofa’ CP lady +Wh TP C° lady+GEN +Wh VP DP t-lady +Wh T°+uC DP sofa-LOC V° sit 63 Chomsky (1999) refers to non-finite T which has an EPP feature but cannot check case.) 68 . If the closest DP is non-Wh. The exact nature of T’s case-assigning properties is not crucial here. T-to-C movement will delete uC on T. For Turkish. Only DP’s need case. repeated as (64). or the element being case-marked and “featurally-related” is able to delete T’s uninterpretable case feature.iv. With this new approach.

4 Recap: our new story Before we proceed to relative clauses with non-specific subjects. the driving feature was on the C head. the subject lady is attracted to [Spec. This contrasts with the NSR sentence in (56). CP]. In the Pesetsky and Torrego analysis. step . . where the subject is not a Wh-element and so cannot delete uC on T. T has an uC feature and an EPP feature that must be satisfied. repeated as (65). let’s review how we have deviated from Pesetsky and Torrego (2001). it only makes sense 69 . After all. is unnecessary. C merges with the TP. Movement of T° to C°. and its EPP and uWh features are checked by movement of the subject lady to [Spec. T to C movement as in is required for this derivation to converge. Notice that we have shifted the perspective from uT features on C to uC features on T. . 65) [bayan-ın Øi otur-du -u ] divani lady-GEN Ø sit-NSR-3s sofa ‘the sofa that the lady is sitting on’ CP sofa +uWh lady +uT DP t-lady +uT TP C° VP PP/DP sofa-LOC +uWh T°+uC V° sit 3. In step . the uC feature on T has been deleted. TP] satisfying T’s EPP feature. step and so disallowed. Because the subject has a +Wh feature.In the derivation in (64).

65 It does not really matter if T moves to C as soon as C merges with TP or after the +Wh element moves to [Spec. Otherwise. Thus sentences with non-specific subjects are TP’s.65 The reason we do not get the NSR -DIK morpheme (indicating T-in-C) when a subject is being relativized. by the next projection.66 In this sense. It is not far afield to assume that matrix sentences with specific subjects in Turkish have a Topic projection that selects for a T° with an uC feature that has an EPP feature. an A projection bearing Wh-like features (which perhaps is the reason. These divergences from P&T (2001) enable us to explain the use of the SR form in RC’s with no external argument as well as those with a non-subject gap in clauses a non-specific subject. [Spec. CP]. T can be thought of as a hybrid. there is no extra projection. The point is that uC on T must be checked preferably within the TP projection or at the latest. This is different from S-selection. uC on T can be checked/deleted either by movement of a +Wh DP to [Spec. it cannot check any features on C. TP] has no such restriction. Arguments could be made for both alternatives. T can undergo head-movement to C). C° cannot check its features against T° merely by virtue of this selection. all of T’s features have been checked. and that T to C is Last Resort-like. 64 70 . The cycle ensures that the former be the unmarked case. is that there is no motivation for T to move to C. whereas sentences with specific subjects have an A-bar projection. when necessary. and a T° with uC and an EPP feature. TP]. Note that only a +Wh-expression can move to [Spec. C will select a T that will ensure that features required by C will be checked64. TopicP. It follows that the same assumption must be made for the vP layer: there is some kind of “Topic-like” projection above the vP which selects for v° with an EPP feature which attracts the DP (specific) object. We saw in (26) and (27) that T° in sentences with non-specific subjects does not have an EPP feature. Thus relative clauses in Turkish have the following functional categories: a C° with uWh and EPP features. This kind of selection is analogous to the Force projection (Rizzi 1997) of C selecting a finite or non-finite T. 66 The implications for matrix sentences are that all sentences with specific subjects must have a CP layer if we are to assume that DP subjects raise from their base-generated positions. v° does not have an EPP feature and the NP (non-specific) object remains in situ. T is Greedy: it only moves to check features on itself.that if there is a C projection. CP]. In this story. or by head movement of T to C.

3. [Øi bu durak-tan bin-il-en] otobüsi Ø this stop-ABL board-PASS-SR bus ‘the bus which is boarded from this stop’ b. as in (4)a repeated as (66)a. Otobüs-e bu durak-tan bin-il-ir. Because one of them is +Wh. bus-DAT this stop-ABL board-PASS-AOR ‘The bus is boarded from this stop. require the SR form. 66) a. bleeding T to C movement.5 Clauses that lack external arguments: the SR form We can now see why clauses that lack external arguments. and bus receives inherent dative case.’ 67) 1. TP]: [TP busi T° [VP this stop [VP ti board ]]] 5. I have included the comparable matrix sentence in (66)b to demonstrate that neither of the two nominals in the sentence. TP]. item 5 in Table 1. “this stop” and “bus”. Economy dictates that the +Wh DP move to [Spec. As shown in derivation (67).T° has uC with an EPP feature. requires structural case: the PP ‘from this stop’ is rendered in Turkish as the nominal this stop with ablative case. This move will delete uC on T.the PP/DP this stop merges with (adjoins to) the VP: [VP this stop [VP bus board]] 3.T° merges with this VP: [ T° [VP this stop [VP bus board ]]] 4.C° merges with the TP: [ C° [TP busi T° [VP this stop [VP ti board ]]]] 6.C’s uWh and EPP features are checked by movement of +Wh bus to its Spec: [CP busi C° [TP ti T° [VP this stop [VP ti board ]]]] 71 . Economy will choose movement of this element to [Spec. there are only non-argument DP’s available.the +Wh-element bus merges with the verb board forming the VP: [VP bus board] 2. there are two (equidistant and non-structurally case-marked) DP’s which can satisfy the EPP of T.

Let’s take a closer look at step 4 in (67). will check uC on T. CP] to check uWh on C. How might Economy dictate movement of one DP over another? The Pesetsky and Torrego analysis provides us an answer. on both T and C. of the . In the tree in (68). 69) +Wh-DP TP DP VP +Wh-DP DP VP V° T° +uC C° . CP] to check . as in (69). they are equidistant from the point of view of T’s EPP. the element moves to [Spec. Although this element satisfies T’s EPP feature. Movement +Wh-DP to T. 68) CP +Wh-DP TP +Wh-DP VP +Wh-DP DP VP V° T° +uC C° Compare with the alternative scenario: the equidistant. there are two DP’s inside the VP. and in . T to C movement is . Being in the same minimal domain. non-Wh DP is attracted to T. Only two moves are necessary to check all the features on T and C. The +Wh DP must also move to [Spec. This derivation requires three moves to check all features CP 72 . the uWh feature on C. required to delete uC on T.

it is more economical to attract an element that will satisfy both these features as soon as possible and at once. Looked at another way. it has two features that must be checked/deleted: uC and its EPP. the system will choose the more efficient option. In (69). When the DP’s providing both options are equidistant. i.flowing ‘Water is flowing from under the door onto the floor. by the amount of work accomplished at each point in the derivation. CP]. the +Wh DP took two short moves to [Spec. water is flowing’) (ii) [[Ø alt-ın]-dan yer-in üzer-ine su ak-an] kapı bottom-AGR-ABL floor-GEN top-AGR-DAT water flow-SR door ‘the door that water is flowing under from onto the floor’ (iii) [kapı-nın alt-ın-dan [Ø üzeri]-ne su ak-an] yer door-GEN bottom-AGR-ABL top-AGR-DAT water flow-SR door ‘the floor that water is flowing from under the door onto’ They conclude “. i. the [SR] is used no matter which contains the target of relativization.e. CP] in .Derivation (68) is preferable to the derivation in (69) also because of Shortest Move. the RC is constructed with the [SR].. crosses more nodes.’ (Literally: ‘From the door’s bottom. From the point of view of T. by attracting a +Wh element to its specifier. In fact.. . uWh on C—can be satisfied with shorter moves than with longer ones. the system prefers a derivation where features—in this case.when there are two oblique phrases in an indefinite-subject construction. no matter what is relativized out of a clause with an indefinite subject. derivation (68) will trump (69) when evaluated locally. The alternative is for T to attract a non-Wh-DP to its Spec and then to undergo head movement to C to delete its uC feature..” The principle of Economy explains why this is so. All things being equal.. the movement of the +Wh DP to [Spec. onto the floor’s top. Any or all of the Economy measures sited will give preference to the SR form over the NSR. When T is merged. is longer move.e. In (68). Both DP’s are equidistant from T.67 67 This was noted in Hankamer & Knecht (1976) with the following examples: (i) [Kapı-nın alt-ın]-dan [yer-in üzer-i]-ne su akıyor door-GEN bottom-AGR-ABL floor-GEN top-AGR-DAT water is. 73 .

70) a.3S doubtful be-SR man ‘the man (such) that (he) will trust us is doubtful’ b. The tree in (71) for the RC in (70)b demonstrates the derivation of a relative clause with a “factive” DP subject. TP] of the relative clause.3S-GEN doubtful üpheli ol-du -u] ‘the man (such) that (he) will trust us is doubtful’ be-NSR-POSS. from within the subject raises to [Spec. man. CP]. [[Ø1 biz-e güven-ece -i]-nin Ø 1p-DAT trust-FUT-COMP-POSS. 74 . uC on T remains unchecked and T to C movement is required for convergence.3S man adam1 I had argued in section 2. Since the subject is not a +Wh element. [DP man [NP t-man [CP …] Nº ] D°]. the +Wh-element. Recall that it was this “factive” NP that received the theta-role from the predicate “is doubtful”. Recall further that I had assumed that the subject could be an NP or that the “fact”-clause could be embedded in a DP. We saw in (36). that relative clauses with sentential subjects permit both RC forms.10 that the subject in these RCs is something akin to “the fact that [the man] trusts us” or “such that [the man] trusts us” and that the structure of these subjects is comprised of a a null “fact” in Nº whose complement is the CP “that [the man] will trust us”. now repeated as (70). The EPP of T attracts the entire DP-subject. Once this subject is in the [Spec.6 Relative clauses with complex arguments We are now ready to extend our analysis to RCs with even more complex arguments.3. [[Ø1 biz-e güven-ece -i] üpheli ol-an] adam1 Ø 1p-DAT trust-FUT-COMP-POSS.

75 . in .+uWh NP t-mani CP TP RPi VP us-DAT t-RP VP trust T° N° (fact) C° C° T°+uC V° is doubtful D° This derivation contrasts with the tree in (72) for the RC in (70)a where the subject is an NP which cannot satisfy the EPP of T. out of the NP- subject for case or the derivation will crash. The DP-man then moves to [Spec. CP] to delete uWh on C. rendering T to C movement unnecessary. At the same time. TP]. As we saw in (31)d. Again. Thus the +Wh DP-man must raise to [Spec.71) the NSR form for RC (70)b CP +Wh-man TP VP DP DP-mani +GEN. . a DP in the specifier of an NP must raise to the spec of a case assigning head or it will violate the Case Filter. the +Wh head of the RC is basegenerated in the specifier position of the NP-subject. this +Wh expression deleted uC on T.

assigns case. +case VP NP DP-man +uWh. Rather than nominative case on C.72) the SR form for RC (70)a CP +Wh-man TP DP-man +uWh. we saw an explanation for the two different relative clause forms in Turkish that accounts for their distribution. is better able to account for the Turkish data. I argued for an EPP feature on T. I suggested a modification of the features implicated by (P&T) that while still capable of handling the Pesetsky and Torrego facts. we saw evidence that D°. -case CP C° T°+uC V° is doubtful N° (fact) 4 Conclusion In this chapter. and suggested a restricted definition of the EPP and the Case Filter. and proposed a DP/NP dichotomy that would capture the facts. Although valuable. such that these apply only to DPs. I demonstrated a correlation between specificity and displacement. but not N°. the account of Wh-movement proposed by Pesetsky and Torrego (2001) failed to adequately explain the Turkish facts. it is an uninterpretable C feature on T that needs to be 76 . TP] and then to [Spec. Furthermore. CP]. The SR form can be explained by the generalization that it is licensed whenever the +Wh relativized expression moves into [Spec.

we know the relative head must have A-moved to [Spec. TP]. we were able to predict the distribution of RC forms. and general enough to provide insight into the grammar of Natural Language. We have mentioned some of these topics above. inherent case on DPs. the ideas will simultaneously be specific enough to provide an accurate account of Turkish clauses. as we proceed to apply our diagnostic. we will examine the A-movement to T and the effects of Minimality in constructions with different verb classes. First. every time we have an SR clause. Second. In Chapter 4. TP]. we will also look into intervention effects. and in infinitival structures. then. it seems that structurally case-marked elements are “frozen” because EPP effects in Turkish seem to be sensitive to structural vs. There are others that we have not yet addressed. This gives us a vehicle for testing conditions of A-movement in Turkish sentences. Case and the EPP. we will encounter issues that have theoretical import. As we apply the SR diagnostic. This chapter provides us with two things. for example. TP]. 77 .checked. we had a vehicle for explaining the seeming optionality in RCs with complex subjects. If we are right that the SR from is licensed only when a +Wh expression moves through [Spec. The hope is that as we formulate explanations. with psych verbs. With this revised account. plus the assumptions about DPs. In addition. we have the beginnings of a diagnostic for movement to [Spec.

as in (1)b. the sentences in (1) demonstrate a direct object with and without overt case morphology. case marking and the structure of nominals. Ali kitap oku-du.’ Overt case morphology is not the only phenomenon correlated with a specific interpretation. For example. Only when the object ‘book’ is marked with accusative case. Ali kitab-ı oku-du. 1) a.Chapter 3: Specificity 1 Introduction In Chapter 2. must have) a specific interpretation. Assuming that Turkish adverbs of manner mark the left edge of the VP68. Ali book read-PST ‘Ali read a book. can it have (indeed. The sentences in (2) establish that case-marked objects are in a different structural position than their bare counterparts. We saw that Turkish does not have overt determiners. Information about specificity is encoded by case morphology and by displacement. sentence (2)c demonstrates that a 68 See Kural (1992). I made several proposals about specificity. 78 . Ali book-ACC read-PST ‘Ali read the book. In this chapter we review some of these assumptions and the arguments in their favor.’ b.

See Kornfilt (2003).70 For clarity. As we will see. *Ben hızlı kitab-ı oku-ru-m I quickly book-ACC read-AOR-1sg ‘I read the book quickly. I will refer to nominals that enter a derivation without case as ‘arguments’. and those that are inherently or lexically case-marked. Ben kitab-ı hızlı oku-ru-m I book-ACC quickly read-AOR-1sg ‘I’ll read the book quickly This is not actually the whole story. admitting that technically a Dative or Ablative expression is also an argument of the verb. 79 . To understand the behavior of Turkish nominals. then. trivially always bear overt case and are ambiguous in terms of their specificity. there are behavioral differences both in the syntax (and at LF and PF) between nominals that require structural case and those that are inherently casemarked. *Ben kitap hızlı oku-r-um I book quickly read-AOR-1sg c. For expository purposes. while an object without overt case must remain inside the VP (2)a-b. I will use the term ‘argument’ to refer to 69 70 Examples from Aygen-Tosun (1999). one must distinguish between those expressions that merge into theta positions without case.case-marked direct object cannot remain inside the VP.’ d. Ben hızlı kitap oku-r-um I quickly book read-AOR-1sg ‘I read books quickly. throughout this chapter.69 2) a. Thus. nominals that are inherently case-marked.’ b.

as in (3). specific) interpretation. TP] and bears case. TP] to check T’s EPP feature. and must receive a partitive (i. 80 . but we do observe the same pattern in subjects in embedded clauses where a specific subject must bear overt genitive case. Compare with (4)b in which the embedded subject has raised above the locative and is case-marked. and not to be confused with factive clauses a la Kiparsky (1971). and non-specifics cannot raise and are bare. the non-specific subject of the embedded existential construction in (4)a does not have case morphology. whereas in (4)a we get the existential reading because the subject has remained in situ while the locative has raised to [Spec.e. This correlation is difficult to demonstrate for subjects in matrix clauses (nominative case being the null morpheme). 3) pro [Ali-nin hasta ol-du -u]-nu söyle-di-ler. Thus.g. This regularity provides us with a useful tool: we can safely use the presence or absence of overt case as an indication of whether or not the object has raised from its basegenerated position. the subject of so-called “factive”71 clauses in Turkish has genitive case.those expressions that are typically referred to as the internal argument and the external argument. and which would be assigned structural case by v or T. For example. (4)b is analogous to (3) where the subject has raised to [Spec. we saw that specific direct objects must both raise and bear overt case. Returning to our examples in (1) and (2). pro Ali-GEN sick be-NSR-3s-ACC say-PAST-3pl ‘They said that Ali was sick’ 71 This is the term used in the literature (e. Kornfilt 1997) to identify complement clauses with the NSR –DIK verbal morpheme. However.

with a particular emphasis for Turkish. I will adopt this insight and assume that it is specificity (or rather. 72 This assumption has several implications: 1) I will assume that all case in Turkish is overt (including the null nominative Ø-morpheme). pro [bir yılan-ın yan-ın-da ol-du -u]-nu söyle-di-ler. pro [yan-ın-da bir yılan ol-du -u]-nu söyle-di-ler. the so-called definiteness effect that specific arguments raise and get case and non-specific ones do not. let us adopt the idea that structural case in Turkish is assigned in a Spec-Head configuration. and 2) case assignment via Long-distance Agree is not possible in Turkish. pro one snake-GEN side-his-LOC be-NSR-3s-ACC say-PAST-3pl ‘They said that (of the salient snakes) one (of them) was by his side’ Because overt case only appears on nominals that have raised.4) a. pro side-his-LOC one snake be-NSR-3s-ACC say-PAST-3pl ‘They said that there was a snake by his side’ b. 73 Enç reminds us that the term had already been used by Fiengo and Higginbotham (1981) and Hudson (1989) to describe constraints on NP movement. see Enç (1991) and Kennelly (2003).72 From the examples so far. noting that non-specific nominals cannot undergo overt object shift. In fact. However. 81 . non-specificity) that constrains subject and object raising. Mahajan (1992) also refers to specificity in providing an account for object movement in Hindi. For a detailed explanation of the distinction between definiteness and specificity. Enç (1991) among others demonstrates convincingly that this is actually a specificity effect73. we have seen that an argument with overt structural case has raised from its base position where it has received a -role. has been observed in other languages.

must have a DP layer. I suggested that specifics were DPs and non-specifics were NPs. and common nouns do not normally overtly raise to D (iii). pronouns are base-generated in D (ii). Longobardi presents evidence for N-to-D movement and posits the structures in (5) as the possible structures of arguments with a null D. in which case there is a null D. proper names must always raise to D (i). He shows that a singular count noun must have a lexical D unless it has a mass interpretation that allows for quantification. Having argued for the existence of a null determiner in Italian. In his story. 5) Arguments with null D in Italian (i) DP D° NP (ii) DP D° pronoun (iii) DP D° NP N° proper name N° common noun 82 . nouns as arguments. in contrast to predicative nouns. I used the insights from Longobardi (1994) to conclude that Turkish specifics have a null DP layer. Longobardi argues that. in Italian. I proposed that the difference in syntactic behavior between specific and non-specific arguments could be attributed to a structural difference.2 Toward an analysis of DP/NP structure and Case in Turkish In Chapter 1.

D Poss N the my Gianni finally called up 74 We will see these facts repeated in Chapter 5 with regard to human nominals. Viene giù acqua dalle colline. as they are useful in helping us formulate the structure of Turkish DPs. 6) a. *mio il Gianni my the Gianni b. So.Let’s see the Italian facts presented by Longobardi. *Acqua viene giù dalle colline. *vecchio il tavolo old the table We see in (9) the expected paradigm: the possible surface orders.74 Italian sentences seem to require a D in the preverbal subject position. 8) a. Il mio Gianni ha finalmente telefonato. water comes down from the hills b. what is going on with (9)c and (9)d that makes one bad and the other good? 9) a. neither an adjective nor a possessive may precede a determiner with either common or proper nouns. (a/the) great friend of Maria called me up Italian adjectives and possessives may occur between D and N or postnominally. 83 . comes down water from the hills 7) *(Un/Il) grande amico di Maria mi ha telefonato. but as the examples in 8) show. [D – Poss – N] in (9)a and [D – N – Poss] in (9)b. as shown in examples (6) and (7).

The Poss can follow D. *Mio Gianni ha finalmente telefonato. we cn understand why (9)c is unacceptable: mio precedes the D°-N° complex. the Poss should follow the name.. Assuming that Nº has raised to Dº. Il Gianni . (in (9)b) DP D° Ø NP N° Gianni D° Il NP N° Gianni If we assume that a DP projection is obligatory for proper names as subjects. there has been head movement of N to a null D. The subjects in (9)b and (9)d are almost identical except that in (9)d. but it can never precede D. Thus. there is in fact a null D head. the 84 . Gianni . DN Poss the Gianni my finally called up c. N Poss Gianni my finally called up Longobardi proposes that in sentences such as (9)d above. Gianni mio ha finalmente telefonato. (9)d is not an exception to the requirement of a DP as subject. and that there has been N to D movement.. The structural difference in the subjects is shown in (10). 10) a. with a determinerless proper name as subject.b.. Il Gianni mio ha finalmente telefonato.. (in (9)d) DP b. or can follow D and the name. beginning with the base order of [D – Poss – N]. Poss N my Gianni finally called up d. Compare this with (9)d which is acceptable because.

The grammar must find an alternative way to capture specificity/non-specificity. b. Unlike Italian. Departing from Longobardi. Turkish does not have lexical determiners. 11) [DP D°-null [NP A°-mio N°-Gianni]] [DP Gianni-D° [NP mio trace-Gianni]] G Longobardi cites Benincà’s (1980) observation that Italian bare nouns as arguments must be interpreted roughly as indefinite. I propose that these nominals do not have a D projection. Mangio patate. I always drink wine. It need not hold for Italian. I make this assumption for Turkish. One line of reasoning is as follows: Italian has lexical determiners. I eat/am eating potatoes. I suggest that the structure of ‘pretty girls’ in the existential sentence in (13)a is not as in (13)b. existentially quantified NPs. as in (12). 85 . in Italian (and generally in English. Bevo sempre vino. but rather as in (13)c. Non c’era studente in giro. There may be parametric variation in how a grammar expresses specificity and definiteness which is determined by whether a language contains lexical D’s. then the overt expression of that feature will differ based on the availability of functional items from the lexicon. as shown in (11). Thus. There wasn’t a student around Longobardi assumes an empty D even for these common nouns. whereas in Turkish.subsequent N-to-D movement yields a DP with a [D–N–Poss] surface word order. c. 12) a.75 75 To be clear. as well) it is the phonological content of the Dº that provides the specific/definite interpretation. with the proviso that N does not raise to D. If we assume that Dº is the locus of specificity. it is the absence of Dº altogether that marks non-specificity.

[NP A°-belle N°-ragazze] Taking this line of reasoning a bit further.e.13) a. there are pretty girls b. the cat) & Scratching (e) & Theme (e. Mendelbaum (1994) shows that predicate NPs are basically adjectival. By extension Szabolcsi (1987). a sentence with an NP subject like cat in (14)a. ∃e [Cat-scratching (e) & Theme (e. 77 I reject the idea that the non-specific subject incorporates into the verb for two reasons: it can be a large expression such as “hundreds of cats” and it serves as an intervener for A-movement from its ccommanding domain. Stowell (1989b) has shown that NPs are non-referential. whereas DPs are referential. kedi [ pro kol-u]-nu tırmala-dı cat arm-POSS-ACC scratch-PST ‘The cat scratched his arm’ b. *[DP [NP A°-belle N°-ragazze]] c. as shown in (14)b77. In fact. some cat or other) scratched his arm’ b. This differs from the sentence in (15)a where the subject is a DP in that we now have an external argument cat. his arm)] 15) a. the semantic interpretation for (15)b would be there is an event of ‘scratching’ which has cat as the Agent and his arm as the Theme. Ci sono belle ragazze. NPs can only be nominal predicates76. would be an event of ‘cat-scratching’ which would have his arm as the Theme. Abney (1987) and Longobardi (1994) have argued that NPs are nominal predicates (unsaturated) and do not bear a theta-role and DPs are arguments that do bear a theta role. it has been argued that only DPs can be arguments. Thus. Translating this idea into an “event-ish” semantic interpretation. 76 86 . [pro kol-u]-nu kedi tırmala-dı arm-POSS-ACC cat scratch-PST ‘A cat (i. his arm)] Higginbotham (1987) proposed that an argument is “saturated” and can thus be assigned a theta role. ∃e [Agent (e. 14) a.

where the Qmorpheme is in its canonical unmarked position post-verbally. as in the structure [VP V° N°]. 80 The anti-incorporation arguments and examples presented in this section are from Kornfilt (2003). or does Turkish allow both DPs and NPs as arguments? What must we say to account for certain arguments not moving and not bearing overt case? I assumed that non-specifics do not have a DP projection. Compare with (16)b.144). p. I would like to claim that. Kornfilt admits that certain properties of incorporation are not found in Turkish. Kornfilt (1984. rather than an argument. the Yes/No Focus Qmorpheme separates the noun from the verb. are all arguments in Turkish indeed DPs as Longobardi proposes for Italian. we see that the particle –DA. Longobardi’s suggestion that all arguments have a DP layer. 76.Longobardi’s reasoning is useful for Turkish because it provides evidence for a null D.. and (18). 78 87 . By disambiguating predicative nominals from arguments. Her motivation for adopting the incorporation account which she had argued against in Kornfilt (1984) is to explain puzzling scrambling facts.78 The second is an attempt to account for the displacement and case facts in Turkish by suggesting that non-specific arguments (either subject or object) incorporate into the verb.. In (17). See fn. Turkish does make an interesting contribution with respect to incorporation”. it may be possible for an NP (or perhaps just an N°) to be a nominal predicate. in the style of Baker (1988). demonstrates that the Turkish free morpheme bile ‘even’ can occupy this slot.80 The first problem with the incorporation account for non-specifics in Turkish has to do with the fact that a variety of focus-question and adverbial particles can appear between the supposedly incorporated noun and the verb. There are two alternatives worth considering. Longobardi’s wording does provide some flexibility: “. through the interaction of scrambling and incorporation. and Kural (1992) argue against this approach. 200379). In (16)a. The question is.a ‘nominal expression’ is an argument only if it introduced by a category D” (Longobardi 1994: 620). 79 Although Kornfilt (2003) adopts the incorporation account. her arguments against incorporation are stronger than those that support it (“In spite of these inconclusive points. ‘also/too’ can appear between noun and verb. One we have seen.

and by way of extension. Hasan pasta ye-di-mi Hasan cake eat-PST-Q ‘Did Hasan eat cake?’ 17) Hasan pasta-da ye-me-di Hasan cake-too eat-NEG-PST ‘Hasan didn’t eat cake either’ Hasan pasta bile ye-me-di Hasan cake even eat-NEG-PST ‘Hasan didn’t even eat cake’ 18) The second argument against incorporation is that we would expect to see a reflex in thematic assignment. Causatives in Turkish introduce an extra theta role to the verb. which we don’t. Hasan Ali-ye kutu-yu aç-tır-dı Hasan Ali-DAT box-ACC open-CAUS-PST ‘Hasan made Ali open the box’ The logic is that when a direct object incorporates into the verb.16) a. the causee of an intransitive verb in Turkish is assigned accusative case whereas the causee of a transitive verb is assigned dative case. Hasan pasta-mı ye-di Hasan cake-FOC eat-PST ‘Hasan ate CAKE??’ b. This is shown in (19). 19) a. Reminiscent of ergative systems which have a case assigning hierarchy based on the number of arguments introduced into a structure. the resulting syntactic structure should be analogous to an intransitive. that of causee. 88 . Hasan Ali-yi ko -tur-du Hasan Ali-ACC run-CAUS-PST ‘Hassan made Ali run’ b.

As we see in (20). This is unexpected if incorporation creates a new word resulting in the direct object losing its independent status.the causee should surface with accusative case.81 When D is [+Specific]. remain constant. taking the sentence in (14)a as an example. 89 . and accompanying case requirements on the causee. Returning to Longobardi’s idea. the DP does not raise. the subject can be a large quantified expression such as “hundreds of cats”. and can only be deleted in a SpecHead configuration with a functional head. we could say that the [+Specific] feature is uninterpretable. the DP must raise to check that feature. In addition. First. what would we have to say to account for the Turkish facts if we adopt Longobardi’s proposal that all arguments are DPs? To account for the obligatory raising of specifics only. The theta roles of the transitive verb. the causee must bear dative case. regardless of the specificity of the direct object. 20) Hasan Ali-ye / *-yi kutu aç-tır-dı Hasan Ali-DAT/*-ACC box open-CAUS-PST ‘Hasan made Ali open boxes’ These facts and others presented by Kural (1992. 1997) are sufficient for us to abandon the idea that non-specifics objects in Turkish incorporate into the verb. When the feature is missing on D. this is not borne out. we could say that D may or may not have a Specificity feature which is strong. I reject the notion that the non-specific subject incorporates into the verb for two reasons. Second. in a transitive sentence with a non-specific direct object. the non-specific subject serves as an intervener for A-movement from its c-commanding domain. 81 Or.

To resolve this. which would result in overt case morphology. as in (21)a. First. or by Move.83 This account would have consequences for the definition of the EPP. which results in overt case morphology. (specific) DP NP N b. in which case no movement occurs and the nominal does not receive overt case-marking. and that of a non-specific DP would be as in (21)b where D does not have the Specificity feature. and if the D had the additional [+Specific] feature. In our story so far.The structure of a specific DP82 would be as in (21)a. it posits two different case checking operations: case-checking via Match and Agree and case-checking via Attract and Move. 90 . Because the presence of an Agr projection is tangential to this schema. the case-checking head looks for D. the +Specific DP would be a target of Attract. In this system. there are presumably two ways that case features can be checked: either by Match and Agree. Certainly. 21) a. (non-specific) DP D+[Specific] NP D N How do these DPs get case? Based on Chomsky (1994. I avoid making a commitment. the EPP would have to be redefined so that it does not target D’s without 82 83 As the focus of here is to explain Turkish facts. while the non-specific DP would get its case checked via Agree. the case-checking head would have to Attract the DP to its specifier to check the [+Specific] feature. I will use the head-final Turkish tree structure It may be Agr that is responsible for case checking and agreement. followed by Move. Matches and Agrees with it. one could converge the operations by assuming that one operation is an extension of the other. There are several shortcomings to approach. 1995).

84 91 . This assumption does not extend to scrambling which is outside the scope of this paper. I do not adopt a null accusative. this is the only Ø case morpheme. Contrary to Longobardi. I assume an overt Ø morpheme. For the null nominative case as well. or default case. absolutive.85 In sum. but crucially. Let’s also assume for Turkish that all phrasal movement is driven by the EPP84 and that all functional heads have an EPP feature which must be satisfied by a DP merging into the specifier position of that head. Let’s take these as given for now. but we would in effect be discarding the requirement of a Spec-Head configuration for case checking and supplanting it with a rule of Spec-Head checking of a Specificity feature. let’s simplify matters and return to the Chomsky (1995) definition of the EPP as a strong D-feature. in Turkish nonspecifics are NPs which do not need case. when there is no overt case morphology. there has been no case assigned. Plus. where a case-checking head checks case without Attract. Separating case-checking from the EPP may not necessarily be a bad outcome. I have nothing to say as to what drives scrambling. 85 In fact. My proposal for Turkish is. Specifics are DPs that contain a null D. partitive. for non-specific DP’s. we would now have a language specific definition of the EPP whereby it could only be satisfied by a DP with a [+Specific] feature. Furthermore. what you see is what you have. Chomsky (1995) considers Dº to be the locus of specificity.the [+Specific] feature. the EPP of that head would not be satisfied. I suggest that (at least in Turkish) non-specific nominals lack a DP projection entirely and are only NPs. with the understanding that we will re-examine these assumptions more thoroughly in the next chapter. It being preferable to reduce assumptions rather than to create them. Thus.

which must be satisfied by a D feature. is not case-marked. The Case Filter: a convergence requirement that all DP’s in a derivation must have their case overtly checked. I have not adopted this idea entirely is because of transitive and unergative subjects which presumably merge in [Spec. Ali.and specifics are DPs which must get case. for example a feature of the Case feature on v and T. The assumption for DP arguments is that they are attracted by the EPP of a case-assigning head. satisfying the EPP. vP]. but a case assigning head need not necessarily assign case. be a feature of a feature. all heads with an EPP feature must have that feature checked/deleted. satisfying the Case Filter. in fact. and thus not require case. I am proposing narrow definitions of the EPP and the Case Filter. I have adopted the NP/DP difference for case assignment and down-played the predicative nature of verbal complements for the sake of uniformity. To be clear. in fact. and refers to non-specific men all named Ali. a DP in the specifier of that functional head. and C. I assume that names being referential must (almost) always undergo raising from N to D87 and that pronouns are base-generated in D88. [Ø Ali-yi öp-en] kız-lar Ali-ACC kiss-SR girl-pl ‘girls that kissed Ali’ 88 Postal (1966) among others has suggested that pronouns are D with features. Again. this suggestion may. [Ø Ali öp-en] kız-lar Ali kiss-SR girl-pl ‘girls that (only) kiss Ali’s’. and are assigned structural case by that head. as follows: The EPP: a feature89 of a functional head. Compare with (ii) where Ali is case-marked and must refer to a specific individual. for a derivation to converge. for the sake of simplicity. that is. v. These cannot be considered predicative. This Actually. I do not assume the Inverse Case Filter. and non-specific subjects of unaccusative verbs. may be predicative. particularly in embedded structures where specific subjects bear overt case. 86 92 . men named Ali ii.e.e. i. being complements of V. The reason. be along the lines of Longobardi in the sense that non-specific direct objects. 87 Exceptions to this generalization are cases where the name is non-referential as in the relative clause in (i) where the direct object.86 Following Longobardi. T. i. 89 The EPP may. whereas the NP/DP distinction in case requirement would account for the different behavior of these subjects. i.

In (22)b). seems to be too strong. on the other hand. be D. 22) a. in fact. pronouns may. where the negative ekki marks the edge of the VP. cross-linguistically. ] Jon bought not the-book/a-book b. The correlation between overt case and raising/non-raising cannot be a maintained in Germanic or Icelandic because all NPs/DPs bear overt case in these 90 91 We will re-examine this assumption also in Chapter 4. for our purposes) DP may raise out of a VP.92 full NPs need not shift. 88. 92 See fn. only a definite (specific. but crucially. I will assume that there is parametric variation in obligatory raising for specifics. a fact I will attribute to their specificity. raising of ‘the-book’ is felicitous whereas raising of ‘a-book’ is not. I am unable to give credit to the source of the sentences. Non-specific objects. Jón keypti [VP ekki bókina/bók. Because the data in (22) appears in so many papers on this subject. Although in Germanic and Icelandic. the shifting of pronouns is obligatory. 93 . whereas a non-specific NP cannot. they must be interpreted as definite or specific. must remain within the VP. A more accurate generalization for many languages is that non-specifics cannot raise. For example. in Icelandic Object Shift91 constructions.90 The generalization that a specific DP must raise. such as locatives or datives. if they do raise. Jón keypti bókina/*bók [VP ekki t ].assumption will have consequences for movement to T of non-subjects. Called so by Holmberg (1986).

it has a marked and contrastive interpretation. and remain in situ. we will see that whereas an NP cannot satisfy the EPP. does not entail this interpretation. 94 . (i) Hasan dün bu pastayı yedi. Let’s look at the Turkish facts to see if this analysis can be maintained. Although (i) is a grammatical sentence.94 23) a. Hasan dün /ka ık-la pasta yedi. where one or more adverbials can appear between the accusative specific object and the verb. In Chapter 4.)”. a direct object does not raise from its base-generated position unless it is specific. As demonstrated in (23)b. Hasan yesterday this cake-acc ate. One way of explaining this would be to say that whereas an NP contains the kinds of features that can satisfy the EPP. What is important here is that we are seeing evidence in different languages of a parameter that constrains the movement of non-specific NPs. and must raise at least to a projection of a case-checking head. The unmarked word order in sentence (23)c. it does however. Hasan yesterday/spoon-COMM cake ate ‘Hasan ate (some) cake with a spoon’ b. do not need case. In (2) we saw that adverbs of manner cannot appear between a nonspecific nominal expression and the verb.93 3 The EPP and case assignment I have proposed that NPs are non-specific. require case. This contrasts with (23)c. Clearly. as in “Hasan ate cake yesterday (rather than today. and case-marked. 93 It may be that the ban on movement may not be due to non-specificity but rather a function of an NP being unvalued. 94 Contrastive Focus in Turkish is the immediate preverbal position. it is unspecified for a value and therefore causes a crash. give rise to intervention effects when it is the closest nominal to the Attracting head. Hasan pasta *dün/*ka ık-la yedi. whereas DPs are specific. a verb does not tolerate any element between itself and its non-specific complement.languages.

*Hertarnir virðast mér [ t vera seinir] the-horses-NOM seem-pl me-DAT be slow 95 I remain agnostic as to whether the operation is case assignment or case checking. and must obey Attract Closest or Minimality. where the embedded subject cannot be raised over the dative experiencer.’ c. Hasan this cake-ACC yesterday/spoon-COMM ate ‘Yesterday. 24) a. Hasan ate this cake with a spoon’ For Turkish. note that the correlation between the EPP and case assignment95 is tangential. Second. Hasan bu pastay-ı dün /ka ık-la yedi. though. 95 . Let’s be clear.c. they raise to satisfy the EPP. while in other languages case may be assigned via Agree. enables us to capture the fact that overtly case-marked arguments must receive a specific interpretation. but must move to subject position if it is definite or a pronoun (24)b. on what I am committed to and the implications that follow. Það virðist einhverjum manni [hertarnir vera seinir] there seems-sg some man-DAT the-horses-NOM be slow ‘It seems to some man that the horses are slow. DPs do not raise for case. Similar to what I am proposing for Turkish can be found in Icelandic seemtype raising constructions (from Jonas and Bobaljik (1993)). classifying specific nominals as DPs that require case and non-specific nominals as NPs that do not. Evidence of a Minimality effect is in (24)c. All movement is driven by the EPP. In (24)a the dative experiencer can remain in the VP if it is indefinite. I am suggesting here that in Turkish case is assigned in a Spec-head configuration after the DP has been attracted by the EPP. although I am not totally committed to this. Mér virðast t [hertarnir vera seinir] me-DAT seem-pl the-horses-NOM be slow ‘It seems to me that the horses are slow. First.’ b.

in situ subjects and objects receive overt case whether they are specific or non-specific. we see evidence of case checking via Agree. 4 Looking at Turkish ‘Quirky’ relatives By adopting the DP/NP split above. while the dative subject was attracted to [Spec. I am thus suggesting parametric variation in how a language encodes specificity. we can explain the rather puzzling phenomenon of optionality in what are otherwise well-behaved relative clauses in Turkish. What is crucial. the embedded verb is nonfinite and cannot assign case to the embedded subject. however. In Turkish. We must assume that it was the matrix finite T that assigned nominative case to the embedded subject (triggering agreement with the verb). I leave for further research. 96 . A difference that may exist between Turkish and Icelandic is how case is assigned. Turkish makes use of both displacement and case morphology to mark specificity. As to whether there is something universal about (non)specificity and NP/DP movement. however. case may need to be checked in a Spec-head configuration after the DP has moved to the Spec of the case-assigning head. the horses. In (23)b. I include this example to demonstrate that the EPP and case assignment are separate operations. is that specificity is a feature of D. and lacking overt determiners Turkish manifests specificity with both case morphology and displacement.In (24)b. In many languages. Whereas these languages use overt determiners to mark specificity. TP] by the EPP.

as in impersonal passive constructions. [ Øi bu durak-tan bin-il-en] otobüsi Ø this stop-ABL board-PASS-SR bus ‘the bus which is boarded from this stop’ (from Kornfilt (1997)) This phenomenon does not seem too troubling at this point. [Spec. and the Non-subject Relative (NSR) (25)b. In cases where there is no surface subject. [ Øi Ankara otobüs-ü-ne bin-il-en ] duraki Ø Ankara bus-CM-DAT board-PASS-SR stop ‘the stop where the Ankara bus is boarded’ b. the SR form must be used. The EPP of T attracts a lower 97 . TP] remains vacant. 26) a. CP]. TP] to [Spec. [ kız-ın Ø gönder-di -i ] mektup girl-GEN send-NSR-3s letter ‘the letter that the girl sent’ We saw that these forms are fairly predictable with the caveat that when there is no subject in the clause.We are now familiar with the two types of relative clauses in Turkish: the Subject Relative (SR) (25)a. [ Ø mektub-u gönder-en ] kız letter-ACC send-SR girl ‘the girl who sent the letter’ b. 25) a. because one can immediately hypothesize that the SR morpheme is probably licensed by movement of an element (the relativized expression) from [Spec. The sentences in (26) are examples we saw in previous chapters where the gap is the (oblique) object of an impersonal passive.

Note now that in (27). These examples differ from the ones in (26) because the relative clauses here have overt subjects. 97 Example (13a) is from zsoy (1994) 96 98 . [[ Øi üst-ü-ne] ö renci yaslan-an ] heykeli Ø top-3s-DAT student lean-SR statue ‘the statue that students are leaning on/lean on’ Let’s make our generalization about the SR form more specific: The SR form is triggered when the EPP of T has attracted the relativized element.DP. the subjects of the clauses are non-specific. we are saying the conditions that license the SR form are met when the EPP of T has been satisfied by a +Wh-DP. is licensed when the structural position of the subject has not been otherwise occupied.out-SR hole ‘the hole which a mouse/mice comes out of’ b. From the examples in (28)97. CP] thus satisfying the licensing of the SR verbal form96. but rather for ease of remembering. mouse and student. the element further moves to [Spec. [ Øi fare çık-an ] deliki Ø mouse come. the SR morpheme is licensed. We also saw in Chapter 2. SR examples as in (27) where the external head is not the subject of the verb. within the clause. the NSR form This is reminiscent of Stylistic Fronting in Icelandic which is licensed only in impersonal constructions and in constructions with a subject gap (Maling 1980/1990). Why is the SR form felicitous when there is no notion of subjecthood of the extracted elements hole and statue? 27) a. after which the DP can move to [Spec. CP]. otherwise. we can deduce the generalization that when the clausal subject has a non-specific interpretation. an operation that would be barred for elements other than the subject. In both Icelandic and Turkish. Although this is not an explanation of the SR form. after which.

as in “at all times” b. TP]. TP] to satisfy T’s EPP and has received case. In contrast. [arı-nın [Ø1 baca -ın]-ı sok-tu -u ] kız1 bee-GEN Ø leg-POSS-ACC sting-NSR-POSS-3s girl ‘the girl whose leg the bee stung’ b. CP] from any projection other than [Spec. TP]. [üç keçi-nin otla-dı -ı] bahçe three goat-GEN graze-NSR garden ‘the garden where the three goats grazed’ Let’s look at another example with similar properties. here. [üç keçi otla-yan] bahçe three goat graze-SR garden ‘the garden where three goats graze’. 28) a. In the RC in (29)a. *[arı [Ø1 baca -ın]-ı sok-an] kız1 bee Ø leg-POSS-ACC sting-SR girl ‘the girl whose leg a/the bee stung’ 99 . the NSR form is required and the SR form is barred. TP]. 29) a. This frees up [Spec. the subject bee is specific and has raised to [Spec. It follows that when the relativized expression must move to [Spec. I include (29)b to show that a specific subject cannot remain without case. the SR form is acceptable in (29)c precisely because the subject bee is non-specific and has not raised from its base-generated position. and neither can a non-specific subject raise above an accusative object. The nominal ‘three goats in (28)a is non-specific whereas in (28)b. a +Wh-expression. the EPP of T is satisfied by the relativized element.must be used. it receives a specific interpretation. The relativized element cannot move to [Spec.

even though the relativized element is a non-subject DP. TP] to the CP projection.98 Thus. If a non-subject is relativized.e. The SR form is licensed when a +Wh-DP moves from [Spec. TP]. If the subject is the relativized expression. the subject of the relative clause will often be “old information” and will therefore be specific. For non-subject relatives. +Wh. the NSR form must be used because [Spec.c. will be in [Spec. In a construction with no subject. having been attracted there by the EPP. and movement to [Spec. as in impersonal passives. [Spec. the subject often refers to the topic. i. TP]. TP] will be occupied by the non-Wh subject. a DP. 98 It is a general property of languages that subjects tend to be about the topic. the SR form will be triggered. TP] is usually occupied by the non-Wh-subject. [[Ø1 baca -ın]-ı arı sok-an] kız1 Ø leg-POSS-ACC bee sting-SR girl ‘the girl whose leg a bee/some bees stung’ 5 The subject/non-subject asymmetry is a misnomer We have seen evidence that the subject/non-subject account for Turkish relative clauses is actually illusory. In normal discourse. it will usually be the case that the subject. in simple relative clauses. or when the subject is an NP and cannot raise to [Spec. This carries over into RCs where in normal discourse. then the SR form will be licensed. CP] will have to originate from a position lower that [Spec. TP]. This is exactly the scenario where the SR form is barred and the NSR form must be used. 100 .

As I have done throughout this dissertation. is simply raising of T to C as in .6 Repeat of Pesetsky and Torrego (2001) We saw in Chapter 2. The derivation of (31)a shown in (31)b demonstrates that the nominative subject girl can check both uT as well as uWh features of C. that Pesetsky & Torrego’s (2001) analysis straightforwardly accounts for the –DIK morpheme in non-subject RCs. making the movement of T to C gratuitous. 30) [ kız-ın Ø gönder-di -i ] mektup girl-GEN send-NSR-3s letter ‘the letter that the girl sent’ CP TP girl VP t-girl letter +Wh V° send T° C° This contrasts with the SR form where the subject bears a +Wh-feature. CP] to check C’s uWh feature. There is no +Wh element in [Spec. followed by movement of the +Wh-object letter to [Spec. Thus. what we see in derivation (30). I show movement only within the clause and ignore promotion of the relativized element beyond CP. 101 . resulting in a PF output of –DIK (the NSR form). TP] to outcompete T-to-C. Their Head Movement Generalization requires the C to T to check C’s uT feature.

This would be even more obvious in embedded structures where T° presumably assigns genitive case. preposing (i. TP] was left vacant for another element. (33)b-c. One need only say that T° always assigns nominative case.31) a. Thus. 32) a. +uWh] [TP [girl. raising to T) the subject and adding genitive case (either with or without the inherent case) on either DP results in acceptability. Under the P&T story we would have to assume that nominative case is assigned to the inherently case-marked DP. uT. [ Øi Ankara otobüs-ün-e bin-il-en ] duraki Ø Ankara bus-CM-DAT board-PASS-SR stop ‘the stop where the Ankara bus is boarded’ 102 . CP] triggers the SR form. hediye-yi ver-en kız gift-ACC give-SR girl ‘the girl who gave the gift’ b. stop+ablative and bus+dative. and that movement of a nominative expression to [Spec. prior to extraction. [ C° +uT. It is not clear to me that the T in impersonal passive constructions assigns nominative case. uWh] T° [VP bought the gift] ] The P&T story can be used to account for Turkish relatives. [Spec. However. I have included sentence (33) to demonstrate that both relativized elements had inherent case. those features remain “alive” to check uT on C. this story requires the stipulation that once nominative case is checked (note that deletion is not an option here) on the +Wh-subject. Recall the impersonal passive constructions in (26) repeated as (32) which require the SR form.e. so that it can check both uT and uWh on C and bleed T to C movement. Because the phrases in (32) lack a “subject”. the relativized DP.

pro they-GEN bus-DAT this stop-ABL board-NSR-3-ACC heard-1s ‘I heard that they boarded the bus from this stop. *pro [bu durak(-tan)-ın otobüs-e bin-il-di -i]-ni duy-dum. Thus. Second. *pro [otobüs(-e)-nin bu durak-tan bin-il-di -i]-ni duy-dum. A Turkish sentence with only non-specific arguments is not acceptable. all other tenses require that at least one nominal be referential (using this term rather loosely) or “specific”. Perhaps it is some kind of Topic-like referential requirement. The idea was that in order to ensure that the Wh-features of C get checked. a DP. C selects for a T with an uC feature. that to guarantee convergence. non-irrealis sentence has a Topic projection that selects a T with an 99 I sever (2003) argues against this idea presenting a host of sentences where both arguments are nonspecific.’ b.b. which I called uC. every non-existential. it is T that has an uninterpretable Wh feature.PASS-NSR-3-ACC heard-1s Intended: ‘I heard that [from this stop] is boarded the bus. First. This is evidence that only in restricted existential or “irrealis” TAM environments can a sentence converge with no DPs. then.’ c. each of his examples includes the evidential verbal morpheme –mi either as the sole TAM (Tense. 103 . [ Øi bu durak-tan bin-il-en] otobüsi Ø this stop-ABL board-PASS-SR bus ‘the bus which is boarded from this stop’ 33) a. even sentences such as “Some dogs barked” requires the subject “some dogs” to have a partitive construction that displays the syntactic properties of a DP.’ This is why I proposed in Chapter 2. pro bus(-DAT)-GEN this stop-ABL board. I did not go into detail as to what this feature on T might be.e. many of his examples include the +human nominal biri ‘someone’ which I show in Chapter 5 is a partitive DP. pro this stop(-ABL)-GEN bus-DAT board. however.PASS-NSR-3-ACC heard-1s Intended: ‘I heard that the bus is boarded from this stop. i. that rather than C having an uT feature. It has been noted that except for the aorist tense. Mood) marker or in addition to a future or aorist morpheme. pro [onlar-ın otobüs-e bu durak-tan bin-di -i]-ni duy-dum. Aspect. He fails to note two things.99 It may be.

We saw how this played out for the impersonal passive constructions in Chapter 2. T° is prohibited from moving to C°. 34) [ Øi bu durak-tan bin-il-en] otobüsi Ø this stop-ABL board-PASS-SR bus ‘the bus which is boarded from this stop’ CP +Wh-bus TP t-(+Wh-bus) VP t-(+Wh-bus) VP V° board T° C° DP this stop-ABL In the derivation in (34). then head movement of T to C is required to check T’s uC feature. the +Wh-Topic-expression (whichever that is) A-bar moves to [Spec. if the DP that the EPP of T is +Wh then. CP]. TP] in . uC on T is deleted and T to C movement is no to check longer motivated. A +Wh-DP must still move to C to check C’s uWh feature. 104 . This DP is +Wh. CP] in C’s uWh. An example is repeated in (34).EPP feature. uC on T is deleted. as well as an uC feature. The relativized element then moves to [Spec. the EPP of T targets the DP bus which moves to [Spec. where T-to-C is not prohibitted. In RCs. and the +Wh-DP moves to [Spec. CP] to check uWh on C. In matrix sentences. once the EPP of T is checked by the closest DP (presumably the subject). consequently. If the closest DP to T is not +Wh.

Let’s look at the derivation in (36).100 The subject is an NP. [üç keçi-nin otla-dı -ı] bahçe three goat-GEN graze-NSR garden ‘the garden where the three goats grazed’ Recall that in Section 2. In the relative clauses in (28) repeated as (35).7 Explaining the choice in RC forms Let’s review. 100 If I did not assume equidistance of the two nominals. I should point out that I am assuming that the subject three goats and the locative garden are in the same minimal domain and thus equidistant from T. we had argued that non-specifics are NPs that cannot satisfy the EPP. and cannot be attracted by the EPP. whereas the subject in (35)b is specific. the EPP of T must be satisfied by a non-subject-DP. 105 . Garden then moves to [Spec. TP] in deletes T’s uC feature. T’s EPP targets the only other nominal. See fn. 35) a. 93. This is precisely what occurs in the RC in (35)a. movement of garden to [Spec. We have two principles that determine the choice of RC form: 1) Specific nominals in Turkish have a DP projection. Because garden has +Wh features. and 2) The EPP attracts a D° feature. [üç keçi otla-yan] bahçe three goat graze-SR garden ‘the garden where three goats graze’ b. that the subject in (35)a can only receive a non-specific interpretation. In a sentence where the subject is an NP. it seemed as if both the SR and the NSR forms were acceptable. CP] and deletes C’s uWh feature. Notice. we would expect intervention effects from the higher nominal. T to C movement is not motivated. however. the DP-garden.

along the lines of Hiraiwa (2001) who suggests the same for Japanese ‘NO’ on subjects in relative clauses. the DP-subject will be left without case and the derivation will crash. if garden moves. garden must raise to [Spec. and receiving genitive case from T101. . TP] satisfying T’s EPP feature. In this derivation. either garden or three goats is a target for Attract by T’s EPP. Because the subject is non-Wh. however. as in [Spec. 106 . Thus.36) [üç keçi otla-yan] bahçe three goat graze-SR garden ‘the garden where three (non-specific) goats graze’ CP +Wh-garden TP t-(+Wh-garden) VP t-(+Wh-garden) VP V° graze C° [+uWh] T° [+uC] NP three goats Now let’s turn to the RC in (35)b with the derivation in (37). I will mention though that I believe it to be a reflex of the T-C amalgam formed by head movement of T to C. We have already seen that structural case in Turkish is assigned in a Spec-head configuration. uC on T has not been checked and T to C movement is required. CP] to check C’s uWh feature. . The +Wh-garden then moves to 37) [üç keçi-nin otla-dı -ı] bahçe three goat-GEN graze-NSR garden ‘the garden where the three goats grazed’ 101 It is beyond the scope of this paper to discuss how genitive rather than nominal case is assigned in this position.

The pairs in (38) have a sentential subject.CP +Wh-garden TP three goats-GEN VP t-(+Wh-garden) DP t-(three goats) T° [+uC] VP V° graze C° [+uWh] In the derivation in (37) we are assuming that the NSR morpheme –DIK. unlike the SR form. Let’s now look at another pair of relative clauses which seem to allow either RC form. 38) a. [[Ø1 biz-e güven-ece -i]-nin Ø 1p-DAT trust-FUT-COMP-POSS-3s-GEN doubtful üpheli ol-du -u] ‘the man that it is doubtful will trust us’ be-NSR-POSS-3s man adam1 107 . This is better shown in (39). not a bad consequence when we are assuming that in this sentence case was assigned to the DP-subject. I assume that one has an NP subject which prevents it from raising to [Spec. The outcome is that structural case-assignment and agreement with the verb seem to go hand-in-hand. while the in the other. TP]. To account for the acceptability of both forms. [[Ø1 biz-e güven-ece -i] üpheli ol-an] adam1 Ø 1p-DAT trust-FUT-COMP-POSS-3s doubtful be-SR man ‘the man that it is doubtful will trust us’ b. TP]. One other point worth mentioning is that. the subject is a DP which raises to [Spec. is the instantiation of T in C. leaving that position open for the +Wh expression. the NSR verb shows agreement.

I had posited a null resumptive pronoun (RP) inside the sentential subject co-indexed with the +Wh-expression in the specifier position of the nominal phrase. DP subjects must move to [Spec. The phrase being non-Wh. the NSR form is required. Thus. On the other hand. [CP Ø1 [TP [DP/CP Ø-RP1 biz-e güven-ece -i]2-nin [VP t2 üpheli ol-du -u] adam1 us-DAT will. [CP [TP Ø1 T° [VP [NP/CP Ø-RP1 biz-e güven-ece -i] üpheli ol-an] adam1 us-DAT will. T-to-C movement is necessary to delete T’s uC feature. either NP or DP. recall that in Chapter 2. This move by the relativized expression to [Spec. the +Wh DP that is to be relativized.trust-GEN doubtful be-NSR-3s man It is evident that there is really no “optionality” in the choice of RC.First. Note that in example (38)b where we assumed the subject was a DP. the entire sentential subject is marked with genitive case. 39) a. T to C movement will be required to save the derivation. If the DP subject does not have +Wh-features. TP] or they will violate the Case Filter. TP] checks uC on T. the phrase cannot be an element in-situ and must have raised to [Spec.trust doubtful be-SR man b. thereby bleeding T to C movement. This is instantiated as the NSR form. that is. NP subjects cannot satisfy the EPP and so T’s EPP feature will attract another DP in the structure. TP]. 108 . The movement is motivated by the EPP on T.

I have shown that it is the interaction of specificity and movement that give us the alternations in Turkish relatives. we look at how movement to [Spec. The generalization is not one of subject/non-subject asymmetry but rather a combination of the category of the subject. and we will continue to address these in the proceeding chapters. that do not need to have their features checked.8 Summary Determining that specific arguments in Turkish must raise to be assigned case. This chapter was intended to review the fundamental claims regarding specificity and the categorial structure of nominals that are important to the overall research proposal of this work. although it seems clear that there will be parametic variation as to how languages encode specificity. In Chapter 5. It may be that in Turkish -features are uninterpretable. -features need checking via (structural) case. TP] is affected by the hierarchical positions of nominals. I think can be maintained for Turkish. and prepositions. 109 . adjectives. the proposal that NPs are not subject to the Case Filter also needs further investigation. Many issues regarding the nature of DPs still need clarification. we again address the structure of nominals. etc. In Chapter 4. and the syntactic reflex of a D° head. Further cross-linguistic research is needed to support this claim. An innovation in this chapter and in the previous one. adverbs. The idea that specifics are DPs and nonspecifics are NPs. is the proposal that there is a +Wh-like feature on T. The assumption that the EPP can only be satisfied by D is fairly well accepted. however. Ns may be lexical atoms like verbs. feature checking and case assignment. that is.

however. indefinites can be either specific or non-specific. There are two women in the room. women-pl-ABL two-agr know-1s (ii) a. The syntactic behavior of Turkish nominals clearly falls along the specific/non-specific divide. I show that movement to [Spec. (i) a.Chapter 4: The EPP on T and Minimality 1 Introduction and Background In this chapter. 1.103 As we saw in Chapter 3. as shown in (i). *There are two of the women in the room. tanıyor-um. and displacement102 Let’s review the behavior of arguments in Turkish. 103 The items marked as bad in (1) are in fact acceptable with marked stress and other contrastive or focus intonations. * Bayan-lar-dan iki-si tanıyor-um. b. I present evidence for the EPP in Turkish. 102 110 . the structure of some of which induce Minimality effects. Assuming that adjuncts adjoin to a verbal projection (either vP or VP). Burzio (1986). TP] obeys Minimality and that the behavior of DP movement across NP subjects is constrained along verb classes (as suggested by Perlmutter (1978). Bayan-lar-dan iki-si-ni women-pl-ABL two-agr-ACC know-1s ‘I know two of the women’ b. (1) is evidence that specific. I then use the SR relative clause as a diagnostic for movement to [Spec. case. is that they are unacceptable as unmarked cases. case-marked DOs must raise out of their thematic positions in VP. TP]. among others). The point. Enç (1991) disambiguates definiteness and specificity noting that whereas all definites must be specific.1 Specificity. Two of the women is indefinite but is unacceptable in an environment restricted to non-specifics.

Ali dün Ali yesterday /with a spoon/quickly cake(-ACC) ate ‘Ali ate (some) cake yesterday/quickly/with a spoon’ /ka ıkla /hızlı pasta(*-yı) yedi. Ali this cake-ACC yesterday /with a spoon/quickly ate ‘Ali ate this cake quickly/yesterday/with a spoon’ c. but certainly in embedded constructions where the specific subject bears overt genitive case. Yan-ın-da bir kız ol-du -u-nu söyle-di-ler.e. and those that do not raise must be bare. However. b. Bir kız-ın yan-ın-da ol-du -u-nu söyle-di-ler. side-his-LOC one girl be-NSR-3s-ACC say-PAST-3pl ‘They said that there was a girl by his side’ b. one girl-GEN side-his-LOC be-NSR-3s-ACC say-PAST-3pl ‘They said that (of the salient girls) one (of them) was by his side’ 111 . Ali-GEN sick be-NSR-3s-ACC say-PAST-3pl ‘They said that Ali was sick’ 3) a. Thus. The subject of socalled “factive” clauses in Turkish has genitive case. the nonspecific subject of the embedded existential construction in (3)a does not have case morphology. Ali cake yesterday/with a spoon/quickly ate In Chapter 2. Ali-nin hasta ol-du -u-nu söyle-di-ler. Compare with (3)b in which the embedded subject has raised above the locative. Ali pasta *dün /*ka ıkla /*hızlı yedi. as in (2). 2) a. The same correlation can be demonstrated for subjects. we can use the presence or absence of overt case as an indication of whether or not the object has raised from its base position.1) a. we also saw that direct objects that raise must bear overt case. specific) interpretation. and must receive a partitive (i. is case-marked. Ali bu pasta-yı dün /ka ıkla /hızlı yedi. perhaps not in matrix clauses because nominative case is the null morpheme.

The conclusion is: case-marked arguments must have raised. 112 . As shown in (iii). More specifically. After scrambling of the locative PP/DP (iva). the EPP attracts a D feature. [CP t1 [TP t1 [PP/DP [DP t ]1 üst-ün-de]2 bayan t2 otur-an] ] divani top-AGR-LOC woman sit-SR sofa Lavine and Freidin (2002) also show that for Russian and Ukrainian A-movement and scrambling have different properties and consequences. TP] (rather than just the +Wh-expression sofa-GEN from the specifier of the PP/DP). and thus can only be satisfied by movement of a DP. the PP/DP ‘on the sofa’s top’ has a possessor/possessee structure. and non-case-marked ones cannot have raised. The scrambled position of the PP/DP must be different from the EPP position because if the entire PP/DP had moved to [Spec. the possessor first moves to [Spec. EPP driven movement is sensitive to minimality such that the EPP feature attracts the closest element from its c-commanding domain that can satisfy it. TP] and then to [Spec. Let’s now look at arguments that raising is a consequence of an EPP feature on v and T. the SR form would not be licensed. we can scramble the locative above the subject and then extract sofa though TP licensing the SR form (ii).1 The EPP on T in Turkish There are several versions of the EPP. but for now. triggering the SR form (ivb). although the focus throughout this chapter will be on T. I would like to adopt the version that defines the EPP as a feature on a functional head that attracts a morphologically contentful element to its specifier. but if we embed sofa in a locative PP/DP as in (iii). [CP [DP divan-ın]1 [TP [DP divan-ın]1 [PP/DP [DP t ]1 üst-ün-de]2 bayan t2 otur-an] C°] sofa-GEN sofa-GEN top-AGR-LOC woman sit-SR c. (i) *[bayan Øi otur-an] divani woman Ø sit-SR sofa Intended: ‘the sofa that a woman is sitting on’ (ii) [Øi üst-ün-de bayan otur-an] divani top-AGR-LOC woman sit-SR sofa ‘the sofa that a woman is sitting on (top of)’ (iii) [PP/DP [DP divan-ın] üst-ün]-de sofa-GEN top-AGR-LOC ‘on the sofa’ (Literally: ‘on the sofa’s top’) (iv) a. We have seen evidence that structural Case in Turkish must be assigned in a Spec-Head 104 My use of the term “movement” precludes scrambling. where sofa is the specifier or possessor (and bears genitive case). [TP Spec-empty [PP/DP [DP divan-ın] üst-ün-de]2 [vP bayan t2 otur] T° ] sofa-GEN top-AGR-LOC woman sit b. Importantly. CP]. I assume that all movement104 of nominals is to satisfy the EPP feature of the head into whose projection they move. The SR form is unacceptable in (i). 1. the Turkish facts indicate that scrambled elements do not satisfy the EPP.1.

107 See Kural (1991). For the moment. with case assignment occurring when the proper configuration is achieved. three potential factors have been identified as motivating this displacement: Case (Boškovi 2002. for now let’s adopt the view that the case-assigning functional heads T and v (as well as the functional head C°) have an EPP feature that must be satisfied by a DP moving to the specifier of that head. and the requirement that certain functional heads must have a syntactically realized specifier phrases (Chomsky 1981. i. the subject dog. the locative street must have moved to a position higher than the base position of the subject. Chomsky (1995) considers Dº to be the locus of specificity. 282. in (4)a. 1982. and rely on a notion of an EPP feature as “the thing that ‘causes’ movement”. to the right of the locative street. “movement is driven by case” proponents could explain this by saying that DPs must satisfy the Case Filter and so Case drives movement of argument DPs to structural case-assigning heads. having been attracted to [Spec. I have presented evidence that in Turkish non-specifics are NPs. agreement. In sentence (4c). as a free rider. is an NP which cannot satisfy T’s EPP. Chomsky 1995.106 To demonstrate.105 We have seen that DPs must raise and be case-marked. has not raised and receives a non-specific interpretation.e. 2005). Thus. dog is a DP which can be attracted by T’s 105 106 In fact. p. TP] by the EPP. let’s disentangle Case from displacement. and specifics are DPs.configuration. Within the generative framework. Assuming locatives mark the edge of VP107. Compare (4)a with (4)c where the subject must receive a specific interpretation. 113 . 1999). We will see shortly that this argument is probably not correct for Turkish.

‘A dog/dogs are barking in the street’ b. Burada [bir tavuk] pi iyor. *[Bir tavuk] pi -iyor a chicken cook-PRES ‘A chicken is cooking’ b. as in (6). ‘Some dog (or other) is barking in the street’ c. ?*[Bir adam] uyu-mu a man sleep-PST ‘A man slept’ 6) a. Köpek sokak-ta havl-ıyor.CONT. Note that these sentences become acceptable with the addition of a locative or temporal phrase.EPP. 5) a. ??[Bir bardak] kır-ıl-mı a glass break-PASS-PST ‘A glass is broken’ c. *Bir köpek sokak-ta havl-ıyor. street-LOC dog bark-PRES.CONT. an NP subject cannot raise over a locative. ‘The dog is barking in the street’ *A/some dog is barking in the street’ The unacceptable or marginal examples108 in (5) provide further support that NPs do not check the EPP of T. dog street-LOC bark-PRES. 114 .CONT. Note that (4)b is unacceptable with a non-specific interpretation for dog. 4) a. here a chicken is-cooking 108 Examples from Kural (1992). Sokak-ta köpek havl-ıyor. hence its position to the left of the locative. a dog street-LOC bark-PRES.

193-194) bars A-movement from an inherent Case position (fn. we will see that the only DP that cannot move to [Spec. 55). Clearly. whereas the examples in (6) are felicitous because T’s Case feature has been discharged. so that if the DP moves to another case position (satisfying the EPP). Burada [bir adam] uyumu . multiple case checking is excluded under Last Resort. otherwise there would be no reason for the derivation to crash if the [-Interpretable] Case feature of the DP that satisfied the EPP on these functional heads were erased. Burada [bir bardak] kırılmı . a DP’s [– Interpretable] feature. here a glass was-broken c. another interpretation is possible. bura-da is a DP with locative case that is deleting T’s EPP feature. in (6) T° is assigning structural case to an inherently case-marked DP. the Turkish facts show this conjecture to be incorrect. here a man slept I conclude from these sentences that T has an EPP feature which must be satisfied for convergence. Furthermore. 109 115 . In this account.b. Case checking of a DP erases its case features. Notice that this account requires that T° (and v°) have case features that must be discharged (the Inverse Case Filter). Is this because structural case-marked elements are frozen for further A-movement109 or is it because case-stacking of structural case on structural According to Chomsky (1995). we can reason that the sentences in (5) are unacceptable because T has not discharged its Case feature.281). In Section 2. whereas multiple satisfaction of the EPP and multiple agreement is possible. The adverbial bura-da ‘here’ in Turkish is actually a determiner-like nominal bu ‘this’ with locative case -da. 284). It is difficult to know if T must assign case. it “offers no Case feature to be checked. so the derivation crashes” (p. for example bura-dan with ablative case means ‘from here’. such as Case.280). This is because “checking [of features] is deletion and is followed by erasure without exception” (p. and bura-ya with dative case means ‘to here’. Thus. is “frozen in place” once checked (p. More specifically. Of course. If we assume the Inverse Case Filter. Chomsky vaguely suggests that some version of the uniformity condition (Chomsky 1986: pp. The nominal can take other cases as well. TP] is one with accusative case.

so it does not need to move for case. i. This gives us two ways to prevent the accusative object from further A-movement: an accusative expression will be frozen because of Last Resort. TP]. perhaps for PF reasons?110 Let’s look at the implications of each assumption. 111 In Chapter 7. On the other hand. there is no way to prevent an accusative-DP from moving to T. 116 . 36). adopting the Inverse Case Filter will mean that T will always have a case feature it must discharge. 285).case is disallowed. it cannot enter into further agreement relations” (fn.e. In this story. In a story where T does not need to assign case. 114 as well. it is not clear how structural case marking would make D features invisible for Attract by the EPP. and T did not check the case of the element in its Spec. the derivation would still converge. 110 Chomsky (1995) also makes reference to “Case conflict” (p. This has the unhappy result of conflating the EPP and Case. we see that the latter of these options seems to be the better account. See fn. The locative that moves to [Spec. is “frozen”. The problem is we now have no way of preventing an accusative DP from moving to [Spec. In adopting an EPP story. 11). Chomsky (2000) also maintains that “once Case of is checked. TP] “does not satisfy the requirement that the features of T can be checked only by Nominative”. We would need to invoke [±Interpretable] features such that once a DP is structurally case-marked. it is “frozen”. if T’s EPP were satisfied. although Icelandic facts regarding Quirky Dative raising to check T’s EPP feature are mentioned.111 The facts in (5) and (6) are a problem for the pure Case drives movement account. Chomsky makes clear that the Quirky Dative in [Spec. TP] already has case. its D feature is no longer available to the EPP. But. Either default inflection or long distance Agree with a lower Nominative is required for convergence (p. Furthermore. I assume that T assigns Case to the DP in its Spec but that movement of the DP is triggered by the EPP. or T will have to assign case to whatever element is in its Spec with the consequence that structural case stacking leads to a crash. I forego the need for the Inverse Case Filter.

in assuming that nominative case in Slavic (as well as in English) is licensed by agreement. A defective category is defined as in (7). 2001). among others. the T° cannot assign nominative case.112 The Russian constructions usually contain two nominal expressions. An argument against the Inverse Case Filter is found in Lavine and Freidin (2002) (L&F). Examples of defective categories in English are infinitival T and passive v which are both -incomplete appear in (8)a and (8)b respectively. In the Russian and Ukrainian examples below. Pesetsky and Torrego (2001). citing Modern Greek and Japanese as examples of languages where T’s features and nominative case can be checked independently. the fact that a locative must move if and only if the subject is non-specific is theoretically unstatable without recourse to the Inverse Case Filter. 117 . They admit that this correlation does not hold universally across all languages. One of these nominals must always 112 The authors (L & F) point out that they follow Chomsky (2000. an accusative direct object and an instrumental. and cannot show agreement. if one rejects the EPP. George and Kornfilt (1981). 7) A category that lacks a full set of -features is defective.Certainly. He was [vP v-DEF attacked by the visitor]. They argue that Russian and Ukrainian Accusative Unaccusative Constructions have a “defective” T. We expect Len [TP T-DEF to finish his book this summer]. The idea is that a T that is -incomplete cannot enter into an Agree relation with a nominal and also cannot assign nominative case. 8) a. b.

when T is -complete. Of interest to us in these constructions is the following: First. 9) a.SG.SG. Compare (9)a-b with (11)a-b. as in (9). flooded-M. Liven' zatopil podvaly downpour-NOM. soldier-ACC ‘A bullet wounded a soldier’ b. Soldata ranilo pulej soldier-ACC wounded-[–AGR] bullet-INST ‘A soldier was wounded’ b.SG. no matter which argument fronts. the accusative nominal must always move to preverbal position. Volnoj oprokinulo lodku wave-INST overturned-[–AGR] boat-ACC ‘A wave overturned a boat.appear preverbally. wounded-F. Podvaly zatopilo livnem basements-ACC flooded-[–AGR] downpour-INST ‘Basements were flooded by the downpour’ c. In the Ukrainian passive-like construction. the subject surfaces with nominative case. 11) a.SG.’ 10) Inozemcja bulo posadženo do v’jaznyci foreigner-ACC was-[–AGR] placed-[–AGR] to prison ‘A foreigner was put into prison’ Let’s focus on the Russian case. basements-ACC ‘A downpour flooded basements’ 118 . Pulja ranila soldata. as in (10). bullet-NOM. the word order is discourse neutral. Second.M.F.

we have evidence that a T that is not assigning case. unlike Scandinavian. Further evidence that Case and the EPP operate independently come from Quirky Case subjects in Icelandic and Germanic case facts. not nominative. This is evidence against the Inverse Case Filter because in the Accusative Unaccusative Construction expressions which already have case move to a functional projection that does not seem to be assigning case. as in Turkish. TP] bears dative. TP] for case. for example). not with the lexical -features of the subject. I assume that T° has not assigned case in (13).114 113 The Icelandic examples in this section are from Freidin and Sprouse (1991). case. As this argument already bears case. and conclude that Russian and Ukrainian. This is contra Burzio (2000) and Frank (2002). Furthermore. it does not need to raise to [Spec. L&F take the fact that either the accusative expression or the instrumental can front as evidence of accusative case being valued via ling-distance agree. T° has an EPP feature which can be satisfied by movement of a non-case-marked nominal to [Spec. TP] where it is assigned nominative case as in (12) or by movement of an inherently case-marked nominal expression to [Spec. the agreement between subject and verb is with the outer shell only. The dative expression in [Spec. and that because the nominative is assigned to the ‘outer’ shell of the DP. Here. TP] as in (13). are not object-shift languages. In order to account for the agreement between the verb and the 114 119 .2 Evidence for the EPP in Icelandic and German It looks like in Icelandic.Third. nevertheless attracts a nominal to its specifier. Burzio argues that (abstract) nominative case is stacked upon Quirky subjects (datives. 1. nominative case in Icelandic has a distinct morpheme.113 I assume that the Icelandic verb in (13). batnaði ‘recover’ is unaccusative-like with the difference being that the expression bearing the Benefactive/Experiencer theta role first-merges as the complement to V° with inherent dative case.

Haraldii virðist [ ti hafa batnað veikin] Harald-DAT seems to-have recovered-from the-disease object as shown in (i) from Sigur sson (1996) and Taraldsen (1995). deleting only its EPP feature. me-DAT failed-3pl all the-attempts-NOM ‘I failed all the attempts. For a complete version of his account. its -features against the object. with default verbal agreement. as in Turkish. and now deletes. it seems clear that matrix T° has not assigned case. Thus. This second checking operation is optional. he fails to account for simpler examples such as (iii) from Jonas and Bobaljik (1993). Haralduri virðist [ ti hafa lesið bókina] Herald-NOM seems to-have read book b. in (i). Again. as evidenced by example (ii).’ (iii) Mér virðast t [ hertarnir vera seinir] me-DAT seem-pl the-horses-NOM be slow ‘It seems to me that the horses are slow. see Frank (2002: 138-153). These examples also demonstrate that. an inherently case-marked expression can move to [Spec. Although Frank provides a lengthy account to explain the restriction against agreement between between a matrix verb across a dative to the object of an infinitival. TP]. Frank adds the notion that checking of -features does not necessarily entail their deletion. Note that whereas in (14)a the subject has been assigned nominative case. ‘It seemed to me that they work well. where the matrix verb agrees with the embedded subject. T first checks its EPP and -features against the dative subject. and then again checks.12) Haralduri las bókina sínai Herald-NOM read book his(+REFL) Haraldii batnaði veikin hjá bróður sínumi Harald-DAT recovered-from the-disease at-the-home-of brother his(+REFL) 13) Dative expressions behave in an identical fashion to the nominative subjects in raising verbs as shown in (14). 14) a.Jonas and Bobaljik (1993).’ 120 . in (14)b. (i) Mér mistókust allar tilarunirnar. the subject bears the inherent dative case which was assigned in the embedded clause.’ (ii) Mér virtust / virtist þær vinna vel. me-DAT seemed-3pl/-3s they-NOM to-work well.

In this structure. Egill drap Harald í gær Egill-NOM killed Harald-ACC yesterday b. some other feature must be implicated for the movement. Let’ s assume that passivization is simply the absence of a vP projection in a derivation with the result that the theta position for an external argument is unavailable. where the internal argument 115 In fact. If passivization is an operation that bleeds accusative case-assignment. therefore. there is no reason to assume that it would have any effect on inherently case-marked DPs. there is no evidence that nominative case has been assigned in (15)b. the EPP of T must still be satisfied. In (16)a. 15) a. Haraldur var drepinn í gær Harald-NOM was killed-PPP yesterday b. The same phenomenon seems to hold in Icelandic passive constructions. Unlike in (15)a. respectively. 121 . Egill hjálpaði barninu Egill-NOM helped the-child-DAT 16) a. whereas in (16)b. In (17).Expressions merged with inherent case retain their case-marking even in passive constructions. there is evidence of long distance case assignment in Icelandic. the dative argument retains its case. The sentences in (15)a and (15)b select for a bare (non-case-marked) argument and a dative argument. the direct object Harald (which needs case to avoid a Case Filter violation) raises from its theta position to [Spec. Thus. Barninu /*barnið var hjálpað the-child-DAT/the-child-NOM was helped We saw in Turkish that non-specifics cannot satisfy the EPP. it may be that T° assigns nominative case in Icelandic via Agree prior to movement. TP] and is assigned nominative case115.

is non-specific, some other expression must satisfy the EPP. In (17)a, the time adverbial ‘yesterday’ raises to [Spec, TP], and as shown in (17)b-c, where no other expression exists in the clause, a pleonastic (or expletive) ‘it’ must be inserted to satisfy T’ s EPP feature. Note that neither of these strategies is acceptable when the internal argument is specific, as shown in (18). I assume this is because the strategies in (17) are both Last Resort, in order to save a derivation. When the derivation contains a “ proper” candidate that will satisfy the EPP (i.e. a specific argument), it is the argument that must raise.

17) a. Í gær var hjálpað barni yesterday was helped-PPP a-child-DAT b. Það var hjálpað barni it was helped-PPP a-child-DAT c. *var hjálpað barni was helped-PPP a-child-DAT 18) a. *Í gær var hjálpað barninu yesterday was helped-PPP the-child-DAT b. *Það var hjálpað barninu it was helped-PPP the-child-DAT Similar evidence appears in the German examples below.116 Whereas T assigns nominative case to canonical subjects, it does not seem to discharge/assign case to inherently case-marked DPs that move to subject position. The German verbs ‘observe’ and ‘help’ subcategorize for an accusative Theme and a Dative Theme, respectively, as shown in (19).
116

The German examples are from Freidin and Sprouse (1991), who following convention give the examples in the form of subordinate clauses to abstract away from German matrix V2 effects.

122

19) a. daß der Polizist den Spion beobachtete that [the policeman]-NOM [the spy]-ACC observed ‘that the policeman observed the spy’ b. daß der Polizist dem Spion half that [the policeman]-NOM [the spy]-DAT helped ‘that the policeman helped the spy’

In German passive constructions, the formerly accusative object behaves as expected, ending up with nominative case in subject position, as in (20)a. Contrast this with the dative Theme in (20)b which retains its lexical (inherent) case marking, appearing in subject position [Spec, TP] with dative case.117 As shown in (20)b, nominative case is not acceptable on a passivized dative expression.

20) a. daß der Spion beobachtet wurde that [the spy]-NOM observed-PPP was b. daß dem Spion /*der Spion geholfen wurde that [the spy]-DAT/[the spy]-NOM helped-PPP was One might argue that the subject in (20)b bears null nominative on top of the dative. This is unlikely for the following reason. Verbs that have subjects with nominative case can be coordinated in German, as in (21)a. However, as demonstrated in (21)b, a passivized dative-DP subject is not possible in a coordinate structure with a nominative-DP. I conclude from this that there is a case mismatch in the two DPs which should not be the case if the dative bore (null) nominative case.
Franks and Levine (2005) point out that in Genitive under negation constructions in Lithuanian, “ [i]n the competition between structural and Lexical Case, lexical Case normally wins.” The same facts hold for Russian. In Genitive under Negation constructions and Numerically Quantified NPs, expressions that would normally bear structural case surface with genitive case, whereas inherently case-marked expressions retain their lexical case.
117

123

21) a. daß der Spion Angst hatte und beobachtet wurde that [the spy]-NOM fear had and observed-PPP was ‘that the spy was afraid and was observed’ b. *daß dem Spion Angst hatte und geholfen wurde that [the spy]-NOM fear had and helped-PPP was Intended: ‘that the spy was afraid and was helped’

The combination of these arguments suggests that the EPP is separate from Case. Committing ourselves to the EPP, let us now turn to determining what elements satisfy the EPP on T and under what circumstances. The evidence will show that the Turkish subject relative (SR) clause form provides a reliable diagnostic for movement to [Spec, TP].

2 Review of Turkish Relative Clauses
We saw that the subject relative (SR) requires the -An verbal morpheme with no agreement as in (22), whereas a non-subject relative (NSR) as in (23) requires the -DIK verbal morpheme with possessive agreement. Crucially, the NSR form requires the (overt) clausal subject to bear genitive case morphology. Thus, in (23)a, the subject must be case-marked and must receive a specific interpretation; the NSR forms in (23)c and (23)d where the subject bears no overt case are unacceptable. 22) a. [Ø i divan-da otur-an ] adami Ø sofa-DAT sit-SR man ‘the man who is sitting on the sofa’ b. *[ Ø i divan-da otur-du -u] adami Ø sofa-DAT sit-NSR-3s man

124

23) a. [adam-ı Ø i otur-du -u ] divani n man-GEN Ø sit-NSR-3s sofa ‘the sofa that the man is/was sitting on’ b. *[adam Ø i otur-an] divani man Ø sit-SR sofa c. *[adam Ø i otur-du -u] divani man Ø sit-NSR-3s sofa Intended: ‘the sofa a (non-specific) man is sitting on’ d. *[Ø i adam otur-du -u] divani Ø man sit-NSR-3s sofa Intended: ‘the sofa a (non-specific) man is sitting on’ We also saw, as in (24)a, that when the subject is non-specific, i.e. is an NP under our assumptions, the Subject Relative form is permitted even though a non-subject has been relativized. In the minimally different (24)b the subject is specific, and thus a DP, and the NSR verbal form is required. When the subject is a DP, the SR form is barred, (24)d. Note though the position of the subject vis-à-vis the dative “ top” (a nominal expression that takes case) which I assume is an adjunct. As expected, when the subject is an NP, it cannot raise, i.e. be in a structurally higher position than the dative “ top” . Compare the unacceptable (24)f with the acceptable (24)a.

24) a. üzeri-ne ö renci yaslan-an araba top-DAT student lean-SR car ‘the car that a student is leaning on’ b. üzeri-ne ö renci-nin yaslan-dı -ı araba top-DAT student-GEN lean-NSR car ‘the car that the student is leaning on’ c. ö renci-nin üzeri-ne yaslan-dı -ı araba student-GEN top-DAT lean-NSR car ‘the car that the student is leaning on’

125

d. *üzeri-ne o ö renci(-nin) yaslan-an araba top-DAT that student-GEN lean-NSR car e. *o ö renci(-nin) üzeri-ne yaslan-an araba that student-GEN top-DAT lean-NSR car f. *ö renci üzeri-ne yaslan-an araba student top-DAT lean-SR car

We formulated a generalization, or rather, a mnemonic, to handle the facts: the SR form in Turkish relatives is licensed when the EPP of T attracts a +Wh DP to its specifier. When a non-Wh expression occupies [Spec, TP], the SR form is barred. Since the SR form is the instantiation of a +Wh element having moved into (and then out of) [Spec, TP], we can use the acceptability of the SR form as a diagnostic of attraction of a DP (in this case albeit a +Wh one) by the EPP on Tº.

3 The SR and Movement through [Spec, TP]
We are assuming that the SR form is the instantiation of a +Wh element having moved through [Spec, TP]. Use of the SR form with NP subjects when extracting non-subjects is not completely free, however. Using the acceptability of the SR form as a diagnostic, let us examine when the SR form is acceptable and when it is not, with the aim of discovering the constraints on the EPP of T118 in Turkish.

118

This paper examines the EPP only in embedded contexts which may have different properties than matrix contexts. For example, left dislocation is degraded in embedded environments. i) a. Down the street ran the boys. b. *He announced/believed/promised that down the street ran the boys.

126

the interpretation of the subject must be a specific type of “ incorporation” . the subject. incorporation ol-an dil-ler-de incorporation be-SR language-pl-LOC ‘languages that have incorporation’ Literally: languages in which there is incorporation b. cannot raise. ?incorporation-ı ol-du -u dil-ler-de n incorporation-GEN be-NSR-3s language-pl-LOC ‘languages which have [this kind of] incorporation’ c. As these examples show. while the NSR form is required in the (c) examples with DP subjects.3. being an NP. This is indeed the case. bu incorporation-ı ol-du -u dil-ler-de n this incorporation-GEN be-NSR-3s language-pl-LOC ‘languages which have this [kind of] incorporation’ Let’ s see what happens with different verb classes.1 Relative clauses with non-specific subjects and the SR form Subjects in existential constructions are non-specific. The (b) sentences demonstrate the case on the relativized element prior to movement: the element that was attracted by the EPP of T in (26)a is 127 . In an existential construction. Note that the example improves with the addition of “ this” to the subject as in (25)c. The (a) examples with the SR form have non-specific NP subjects. The verb in examples (26) and (27) is unaccusative. 25) a. We would thus predict that relativization of a non-subject DP in an existential construction would license the SR form. In the marginal (25)b. both RC forms are acceptable depending on whether or not the subject is specific. as in (25). and the EPP of T must target another DP.

’ c. 26) a.prog. Liman-a gemi yana -ı yor.prog. the SR form is acceptable (when extracting non-subjects) with unaccusative verbs when the subject is an NP.a locative. such as (28) and (29). gemi-nin yana -tı -ı liman ship-GEN sidle-NSR harbor ‘the harbor that the ship is sidling up to’ This contrasts with unergative verbs. where only the NSR form is permitted. 28) a. harbor-DAT ship sidle-pres. ya mur-un ya -dı -ı bölge rain-GEN rain-NSR-POSS-3s region ‘the region where it is raining/rained’ 27) a. bare subject is unacceptable in either RC form. *gençler dans ed-en salon youth dance do-SR club Intended: ‘the club where young people dance’ b. gençler*(-in) dans et-ti -i salon youth-GEN dance do-NSR club ‘the club where young people dance’ 128 . Bölge-de ya mur ya -ı yor. ya mur ya -an bölge rain rain-SR region ‘the region where it rains/where (typically) there is rain’ b.-3s ‘It’ s raining in that region. Notice that the subject must be specific (a DP) and receive overt case. and in (27)a. gemi yana -an liman ship sidle-SR harbor ‘the harbor that a ship is sidling up to’ b. a non-specific.-3s ‘A ship is sidling up to the harbor’ c. region-LOC rain raining-pres. Crucially. a dative.

a genitive can satisfy the EPP on Tº. Let’ s now turn to transitive constructions. *Ø 1 arı sok-an kı 1 z Ø bee sting-SR girl Intended: ‘the girl some bee stung’ [Actual: ‘the girl who stung some bee’ ] 129 . the relativized element. -nı n-ı sok-du -u]] kı 1 z 30) a. TP]. i. (30)c. was the genitive possessor girl of the accusative object. as in (30)b. we see that locatives. the relativized element is the possessor of the object DP. girl’ s leg. and datives can satisfy T’ s EPP in unaccusatives with NP subjects. Note that the non-specific subject being an NP cannot raise from its base-generated position to a position higher than the accusative direct object in [Spec. [ [Ø 1 baca -ı ] n-ı arı sok-an ] kı 1 z leg-POSS-ACC bee sting-SR girl ‘the girl whose leg some bee stung’ c. but. Thus.29) a. vP]. In (30)b. this is not acceptable with unergatives. atlet-ler*(-in) ko -tu -u saha athlete-pl-GEN run-NSR-3s track ‘the track where athletes run’ So far.e. *[arı[Ø 1 baca -ı ] sok-an ] kı 1 n-ı z bee leg-POSS-ACC sting-SR girl d. Use of the SR form is also acceptable. and as expected the non-subject relative (NSR) verbal form is licensed when the subject is specific. genitive possessor DPs. the element that moved through [Spec. when the subject is non-specific. *atlet-ler ko -an saha athlete-pl run-SR track Intended: ‘the track where athletes run’ b. In (30)a. [CP arı n [VP [DP Ø 1 baca -ı ] bee-GEN leg-POSS-ACC sting-NSR-3s girl ‘the girl whose leg the bee stung’ b.

In the NP-Deletion example in (i).Note that. But I don’ t know who’ 130 . the specific subject in the embedded clause bears genitive case. of course. 114). (i) Bu kim-in araba-sı Ahmet’ in arabası . the SR form is unacceptable when relativizing an accusative object even when the subject is non-specific. Ahmet one-GEN you-ACC call-COMP-ACC said-pst ‘Ahmet said someone called you’ b. we must assume that. genitive case on the subject is ungrammatical. in the sluiced structure (iib). (30)d.e. 117. genitive case remains on the possessor. I assume that the genitive on RC subjects in Turkish is structural case assigned either by T or by a 119 Although I have argued against the idea that T° has a case feature it must discharge. Ahmet birin-in sen-i ara-dı ı -nı söyle-di. genitive case on a possessor is inherent. Recall that it is only structural case that causes this restriction. This would distinguish possessor genitive from clausal genitive on a subject. the move obeys minimality). As we noted. perhaps because they can get a second case (Hong 2002). inherently case-marked expressions are free to undergo A-movement. at least in Turkish. .) provides the following data as evidence that Turkish has two types of genitive case: inherent and structural. TP]. Ama kim(*-in) bilmi-yorum.119 Because in (30)b a genitive possessor is patterning like the dative and locative DPs in these SR constructions. Thus. a non-subject +Wh-DP can raise to [Spec. A.c. in the NSR form with overt subject) bears genitive case. Ahmet birin-in sen-i ara-dı ı -nı söyle-di. this who-GEN car-poss. and I leave the option open that the Inverse Case Filter may hold. 120 Chomsky (1995) considers genitive case inherent because it is associated with theta-marking (p. as expected. nothing in the story I am presenting rests on this assumption. 2000). But see fn. Ince (p.120 Recall that the RC subject (i. In (iia). However. TP] in a transitive construction as long as the DP does not bear accusative case (and. This is in line with Chomsky (1995. I point to arguments in favor of one view or the other as a way of expanding the discussion. Ahmet-poss ‘Whose car is this? Ahmet’ s’ (ii) a. Ahmet one-GEN you-ACC call-COMP-ACC said-pst but who(*-GEN) know-pres-1s ‘Ahmet said someone called you. See also fn. 118. an accusative case-marked element cannot move to [Spec. it can no longer undergo A-movement to a structural case assigning/checking position. once structural case has been checked on a DP. not structural.

I will not open the can of worms about an EPP feature on D.V-T-C amalgam. This is consistent with Hong’ s 2002 observations in Korean. the possessor genitive assigned by D is inherent case because it patterns with all other inherently case-marked elements in being able to raise to T in unaccusatives with NP subjects. if one wanted to assume that genitive case assigned by D° is also structural. movement of the structural genitive possessor to [Spec.123 The unergatives in (28) and (29) had animate subjects. these may be interpreted at LF whereas structural case must be PF-interpretable. an ablative. TP] does not lead to PF crash because of homomorphism. Note that structural case does not seem to allow a second case. I must admit that I find the idea that the case assigned by D° is structural tempting because it is not theta-related and does seem to require raising to a Spec-Head configuration. 122 Of course. unergatives with inanimate NP subjects also bar movement of another element to T. 123 Note that a possessor genitive as in the response to the question in (i).122. Since inherent or lexical case is thematic. But. structural or inherent. TP].1s ‘I am afraid of yours’ (iii) *sen-in-i gördüm you-GEN-ACC saw-1s ‘I saw yours’ 121 131 . Not surprisingly. This would mean that both the accusative object and its possessor are candidates for Attract by T’ s EPP. These facts suggest that possessor genitives are structural. who-GEN dog-AGR ran.away you-GEN ‘Whose dog ran away? Yours. ablative in (ii). as demonstrated in the unacceptable unergative examples in (31) and (32).’ (ii) *sen-in-den korkuyor-um you-GEN-ABL fear. (i) Kim-in köpe -i kaçtı ? Sen-in.’ 32) a. one could argue that nominative case on top of accusative leads to a PF crash. *ta -lar dü -en tepe rocks fall-SR hill ‘The hill rocks fall from’ See Hiraiwa (2001) for arguments that a possessor genitive and RC subject genitive are different types of case and that the subject genitive case is assigned by a V-T-C amalgam.121 On the other hand. does not allow an additional inherent case. (iii). 31) *araba-lar gid-en cadde cars go-SR street ‘the street that cars go on. One might wonder if sentiency or animacy might be a factor. or a structural accusative case. and (32). The former leads to a PF crash whereas the latter will lead to convergence. Example (31) is an attempt to raise a locative to [Spec. Since the case assigned by the RC T° is genitive.

TP] over an accusative object. In a transitive construction. 125 But see fn.PROG. Thus we can conclude that accusative object induces intervention effects for A-movement. T can. ‘Rocks fall from this hill. 124 132 . Tepe-den ta lar dü ü-yor. vP] will already have been assigned structural case. 119.124 We would assume that another (non-structurally case-marked) DP in the structure should be able to satisfy the EPP of T as long as the move obeys minimality. but the derivation would crash at PF. the DO in [Spec.b. as example (33) demonstrates. a locative cannot move to [Spec.125 The question is. a transitive with an NP-subject. 122 where I proposed that perhaps the DP could move to just such a position. and will no longer be a candidate for Attract by the EPP of T. assign case to a DP without case. This is consistent with Hong’ s (2002) observation about multiple case on DPs in Korean: structural case can be added onto an inherently case-marked element. See fn. it can no longer undergo A-movement to a case-checking position. it can also assign nominative case to an inherently case-marked DP.’ 4 Toward an explanation We are assuming that once structural case has been checked on a DP. However. This is demonstrated in the tree in (34) for example (33). hill-ABL rocks fall-PRES. trivially. At this point in the derivation. will the accusative object be invisible for Attract or will it serve as an intervener for movement of another nominal element lower in the structure? In (30)b. I assume that case is assigned in a Spec-Head configuration. TP]. the genitive possessor of the accusative DO raised to [Spec. the EPP of v must be satisfied before T° merges with vP. but stacking of structural case onto structural case does not seem to be possible.

vP]. if they do assign (overt) case. or in the absence of one. and v° has an EPP feature when selected by v*°. the case-marked element must be the 126 A footnote in the previous chapter contained a discussion about the EPP on v° where I speculated that the vP projection might be analogous to the TP projection.126 A DP direct object can satisfy v’ s EPP. another nominal element must move to [Spec. It is important to note that with both Tº and vº. How do we account for the difference? 35) [Ø 1 çocuk-lar-a kurt saldı r-an] orman1 Ø children-DAT wolf attack-SR woods ‘the woods where a wolf/wolves attack children’ Recall that we had assumed that vº also has an EPP feature that must be checked/deleted.33) *[Ø 1 çocuk-lar-ı arı sok-an] orman1 Ø children-ACC bee sting-SR woods ‘the woods where a bee/bees sting children’ 34) woods +Wh CP TP C° t-woods +Wh vP T° +EPP DP-children-ACC NP-bee VP woods-LOC +Wh DP children v° V° sting The unacceptable SR RC in (33) contrasts with the acceptable example in (35). 133 . That is. T° has an EPP feature when selected by C°. the head of a (perhaps A-bar) v*P projection above vP.

and so equidistant from v. leaving the locative in VP. As demonstrated in the tree in (36).one Attracted to the case-assigning head by the EPP. the locative woods moves to [Spec. and . If the dative DP children moved to [Spec. so. (Actually. but is not assigned structural case. In the derivation in (36). the derivation will crash because of a Case Filter violation. the internal argument is assigned inherent dative case which does not need to move for case. the locative being in the specifier is closer. This leaves [Spec. -Case] V° attack The derivation in (36) has implications for the “ movement for case” story. Thus. the dative internal argument children and the locative woods are in the same minimal domain. under one definition. on the other hand.) We can assume that the Turkish verb attack “ saldı ran” in (35) is quirky and does not assign/check accusative case. If the direct object does not raise. can further undergo A-movement to [Spec. vP]. TP] as in 36) woods +Wh t-woods +Wh CP TP C° vP T° [+EPP] woods-LOC +Wh NP-wolf VP woods-LOC +Wh PP/DP children-DAT v° [+EPP. it would block the +Wh locative from raising to 134 . In example (35). vP] and deletes v’ s EPP feature in . the direct object must raise out of the VP. vP] available for the +Wh-locative. in (33). Let’ s assume that v° had a case feature that needed to be deleted or checked.

it can cannot move to [Spec. vP] in (36) might not be scrambling. We also see evidence that a scrambled expression cannot move again because a DP cannot scramble around an intervener and A-move. once scrambled. once the +Wh-locative had been assigned structural accusative case. 2) a scrambled expression is porous for movement (contra Chomsky (2005)). it cannot move again (See fn. TP] and then to [Spec.e. Furthermore. you cannot say “ *?the top that the child spilled juice on” . There is no way to tease apart the cause of the derivational crash when we try to relativize these scrambled expressions. could not move to [Spec. 135 . only a constituent can move once an expression has scrambled. where the RC head is an N° that is promoted from [Spec. and that within the confines of this work. CP] to a theta position in the matrix sentence. we will see many examples where a scrambled expression ameliorates minimality effects by carrying a +Wh-DP constituent around an intervener. It seems that this example is another piece of evidence against the movement is for case analysis. Throughout this work. Tomo Fujii (p. we will observe effects and properties of scrambling. once having scrambled.) points out. in a context where the child spilled juice on top of one of the sofas.c. and 3) once an expression scrambles. we are able to note three properties that emerge with respect to scrambling which are important for the analysis in this work: 1) a scrambled expression does not delete an EPP feature (because it does not move to the specifier of the functional head bearing an EPP feature).[Spec.127 127 It is not possible to deal with the larger issue of scrambling in this thesis. we are not able to determine whether. an element is frozen. A scrambled DP adjoins below TP. A constituent may raise out of a scrambled expression. bleeding the SR form. in a context where a factory worker is painting tops of cars. Japanese also has constructions such as [sofa’ s top]+case. to merely state that these larger DP’ s cannot be relativized. This is consistent with the Turkish facts: Turkish RCs do not permit a possessor-possessee relative head. TP] and the SR form would be barred. TP]. but the expression itself is locked in place for the rest of the derivation. that the scrambled expression may not be frozen. a DP cannot be relativized using the SR form. however. and [table’ s under]+case ‘the bottom of the table’ . it would be assigned structural accusative case on top of the inherent locative case. CP] which would license the SR form. However. 11 in Chapter 7 for speculations as to why this may be so. they are pseudoPPs) and cannot be relative heads either in Turkish or Japanese. Even in English. they can actually move to TP. One might ask whether the movement of the locative to [Spec. Thus. This is not likely as there is evidence that once scrambled. and like other accusative objects. it does seem okay to say “ the top that the worker painted (green)” .) My assumption that scrambling freezes an expression was based on the fact that. The scrambled elements we are looking at in this work have just this configuration. I assume the raising/promotion analysis of RCs. but now. Crucially. This would be permissible. however. it would be frozen for further A-movement. It seems more prudent then. but although “ top” and “ bottom” bear case (and in Turkish they bear possessive agreement with the possessor in the Spec) these terms to be semantically weak (i. the freezing effect may be epiphenomenal to structural and semantic constraints. If the locative woods moved to delete v°’ s case feature.

Why was the SR form acceptable for unaccusatives but not for unergatives? Previous research has suggested that the base position of the subject. 136 . the subject is generated in [Spec. using WCO. but not in (37). the expression that picks up the outermost theta position. that in an unaccusative structure. the subject is generated as the complement of Vº. There are suggestions that there may be an A-bar projection above the verbal domain but below the inflectional domain (also proposed by Baker & Stewart 2002). ak-an] dami 37) [Ø i su water pour-SR roof ‘the roof water pours/drips from’ CP roof +Wh t-roof +Wh roof-ABL +Wh TP VP NP water C° T° [+EPP] V° pours Kural (1992). Note that the difference in these structures is the position of the NP-subject which intervenes between the +Wh-expression and T in (38).128 i. but crucially not to a C projection. scope and binding phenomena as diagnostics. in these verbs is different. Kural’ s idea is appealing because the idea that scrambling is A-bar movement to a projection immediately above vP provides us with a landing position as well as a reason for the freezing effect. whereas in an unergative. as in (38) for example (29)a. argues that scrambling in Turkish is A-bar movement. vP]. i.e. as in (37). 128 Perlmutter’ s (1978) “ Unaccusative Hypothesis” and Burzio (1986) for evidence in Italian. This is hierarchically equivalent to the position the scrambled expressions here seem to occupy.1 The base position of subjects and intervention by NPs Let’ s return now to our questions regarding the unacceptable unergative examples in (28)-(29) and (31)-(32).e.4.

The implication is that there are three v heads in Turkish: one with both EPP and case features. I had assumed that the v in this construction had an EPP feature but did not assign case.38) *[atlet-ler ko -an] saha athlete-pl run-SR track Intended: ‘the track where athletes run’ CP TP track +Wh NP-athletes VP track-LOC +Wh V° run v° [-EPP. it would be frozen and could not move to T to 137 . I’ d like to side-track momentarily to point out thatthe derivation in (36) is an argument against the non-EPP story that movement to functional heads is for Case. then the +Wh-locative in (36) could not have moved for case. one with only an EPP feature. This differs from the transitive construction in (35) with the tree in (36) which had an internal argument that was assigned inherent dative case.e. -Case] vP C° T° [+EPP] Note also that I am making one more assumption about unergatives: that the v° head does not have an EPP feature. For if it had been assigned accusative case on top of its inherent locative case (which is theoretically possible in the story here). If we are right in assuming that an element that is assigned structural case is no longer a candidate for A-movement. and one with neither an EPP nor a case feature. movement to another case assigning functional projection. i.

track ( ). unaccusatives provides us with insight to answer such questions. but it does serve to block the locative from being attracted to T. The structure of (38) meets the requirements of the DIC: (athletes) c-commands (track). while an NP may bear the kind of features that would be picked up 138 . blocks the movement of a DP from its c-commanding domain. This contrasts with (37) where there is no interevener. which itself cannot satisfy the EPP of T. Another possible explanation for the movement of the locative to the vP is scrambling. where > is c-command. We must conclude that the DIC plays a role in the unergative examples: an NP. The locative could scramble above the subject and then move to [Spec. the locative DP. The differences with respect to RCs in unergatives vs.e. but > > . matches probe Tº ( ).129 129 Another way of explaining the intervention effects of NPs might be to propose that they are unspecified. What we see in (38) is evidence for what Chomsky (2000) describes as a defective intervention constraint (39). probe . i. Thus the NP subject is an intervener for attraction of the locative by T’ s EPP. In terms of the DIC.e. TP]. 39) Defective Intervention Constraint (DIC): In the structure. lack a value. the effects of matching are blocked. The facts throughout this thesis indicate that this option is not feasible. in (38). and and match is inactive. i. athletes ( ) is “ inactive” . The NP-subject. the NP subject cannot satisfy the EPP on T.trigger the SR form. Thus. not a candidate to move to T. The question that should be asked is why does the locative need to move to vP in the first place? After all.

the features are unvalued and thus unable to delete the EPP on that head. is the EPP (in Turkish) a feature or a structural requirement? If the EPP were merely a structural requirement as argued by Lasnik (2001). Having argued throughout this chapter that there is an EPP. If the EPP was only a requirement that the Spec position of a functional head be occupied. The first. precluding the motivation for movement. there are three core versions of the EPP each with different implications. This idea is radically different from what I have been proposing thus far. The data indicates that not only do c-commanding DPs intervene in the raising to [Spec. Chomsky (2000. and NPs such as the subject of an unaccusative. based on Chomsky (1981. an NP subject of an unergative verb could not block another DP in its by the EPP of an Attracting head. 1982) is configurational: clauses must have subjects. these expressions would not induce intervention effects. we would not see intervention effects. In the second version. TP]. NPs do move. It is only when specific features must be checked can the head be selective in admitting or barring movement to its specifier. but they fail to delete or check the EPP.5 Versions of the EPP Within the generative framework. Note that this version requires the idea of unique Spec positions. These include DPs such as accusative objects. In this version. In this version. 2001) relies on Match and Agree operations for -feature checking. Once an NP has been attracted by the EPP. but even those expressions that cannot themselves satisfy the EPP intervene. a DP cannot come along and check the EPP on a head by merging in a second Spec position. If the EPP were not sensitive to features. the EPP is a strong D feature of a functional head and the subsequent pied-piping of the phrase entire phrase. the EPP is a requirement that certain functional heads must have a specifier. they are attracted by the EPP. Recently. 139 . based on Chomsky (1995). the question is. the derivation will crash.

the DP/DP phrase on the street with the structure [street’ s top-AGR]-LOC functions as a carrier of the +Wh-street around the NP subject. [[Ø 1 üst-ün]-de araba gid-en] cadde1 top-AGR-LOC car go-SR street ‘the street that cars go on (top of)’ b. but at this point note that the unacceptable (31) repeated as (40). there is some featural requirement that an NP cannot satisfy. what would prevent an NP from satisfying it by moving to [Spec. We will see this operation in greater detail in later chapters. is acceptable in (41). After. TP] licensing the SR form. it is free to move to [Spec. As demonstrated in (41)b. Although limited to subordinate clauses.c-commanding domain from moving to T.’ 41) a. If the EPP were merely a structural requirement. Obviously. street has gotten around the subject. [CP [TP [DP/PP Ø 1 üst-ün]-dek [vP araba tk gid-en] cadde1 top-AGR-LOC car go-SR street 6 Conclusion This chapter has looked at the consequences of and constraints on the EPP of T in Turkish. the findings prove interesting because they support previous observations about the syntactic structure of different 140 . cars. 40) *araba-lar gid-en cadde cars go-SR street ‘the street that cars go on. We are able to deduce that the NP subject is an intervener for movement because it is possible to skirt the NP via scrambling. TP].

Using the acceptability of the SR relative clause form as a diagnostic of movement to [Spec. some other expression. however. a DP. This is because neither the accusative object nor the NP-subject c-commands this possessor DP. a DP-direct object raises to vP. the former cannot satisfy the EPP and the latter cannot move to a structural case position. which is generated in a structurally higher position. they block movement of a lower DP. Nominals that cannot satisfy T’ s EPP feature do. TP]. These facts 141 . Any inherently casemarked expression can be relativized using the SR form in unaccusatives because the subject will not intervene. Although the accusative object itself cannot satisfy T’ s EPP feature. NP-subjects and accusative objects induce intervention effects. in transitive constructions. does block the raising of DPs from its ccommanding domain. case-marking of arguments and intervention effects. Structural hierarchy rather than grammatical functions such as subject or object determines the ability of a DP to raise to T. the NP subject of an unergative verb. Similarly.verb classes. An NP subject of an unaccusative verb because of its position as the complements of V° does not block another DP raising to T. A genitive possessor from within the accusative object. can raise to T. serve as interveners for the raising of other DPs in their c-commanding domain. We have seen arguments for an EPP feature which can only be satisfied by a DP (a specific nominal expression). both are barred from moving to T. In clauses with an NP-subject. Although neither can satisfy the EPP on T. must raise to delete T’ s EPP feature for convergence. By contrast. it serves to block the raising of a lower DP such as a locative. however. Approaches based on theta-roles are also not adequate. we have seen intervention effects of Attraction by the EPP of T.

give support to Chomsky’ s (2000) Defective Intervention Constraints which stipulate that elements that cannot themselves delete features serve to block other elements in their c-commanding domain from checking those features.e. any DP can check T’ s EPP feature with the proviso that it not be structurally casemarked (i. In sum. the EPP seems to be a featural requirement. not bear accusative case). as long as there is no intervening c-commanding nominal. 142 .

As we will see later in this chapter. the subject of an unaccusative verb is its complement. *[araba(-lar) gid-en] cadde car(s) go-SR street ‘the street that cars go on. another way of denoting non-specificity in Turkish is by use of a partitive construction. vP]. (i) [araba-lar-dan git-ti -i] cadde car-pl-ABL go-NSR street ‘the street that (some (of the)) cars go on’ 143 . Note that this excludes inherently case-marked expressions which. so example (i) is also a possibility. it must be stated with a specific subject using the NSR form. 131 Note that example (6) must do double-duty. When relativizing a non-subject. and is used to mean both “ the specific cars” and “ the cars in general” . while the subject of both transitive and intransitive verbs is the expression merged in the theta position in [Spec. the NP-subject cars intervenes between the Wh-element and T. the SR form is generally not acceptable in clauses with unergative.131 1) a. analogous to PPs. Thus.Chapter 5: Human/Non-human Distinctions in Turkish 1 Background In Chapter 4. or intransitive. verbs. As demonstrated in the tree in (1)b for (1)a.’ 130 By subject I mean the expression that merges into the outermost theta position of the verb. There is no way to express this clause with a non-specific subject. are lexically theta-marked. we saw that the movement of a DP to T obeys Minimality. as in (2). This means that the “ subject” 130 is the complement of V° and other nominal expressions are generated higher than the subject. The example in (1) has an unaccusative verb.

car street-GEN top-AGR-LOC go-AOR Lit: ‘Cars go on the street’ s top’ (i.b. can scramble around the subject. TP] without intervention from the subject as in the now acceptable SR clause in (3)d. track CP TP t-street +uWh NP-cars VP DP-street +Loc V° go C° vP T° v° 2) [araba(-lar)-ı git-ti -i] cadde n car(s)-GEN go-NSR street ‘the street that the cars go on’ I have argued that (1)a is a Minimality violation. There is another derivation that literally gets around the Minimality issue by letting cadde ‘street’ become a “ free-rider” on a scrambled expression. A larger expression. [DP/PP [DP cadde-nin] üst-ün]-de street-GEN top-AGR-LOC Lit: ‘on the street’ s top’ b. (3)c. [araba [VP [DP [DP cadde-nin] üst-ün]-de gid-er]]. for example the DP/PP [on the street’ s top] with street as the specifier as in (3)a and (3)b. the NP subject c-commands the locative and is an intervener.e. The specifier street is now free to move out of the scrambled expression raising to [Spec. ‘Cars go on/over the street’ ) 144 . 3) a.

[CP . *[kedi tı rmala-yan] çocuk cat claw-SR child Intended: ‘the child that a cat clawed’ [Actual: ‘the child that clawed a cat’ ] b. [ [DP [DP cadde-nin] üst-ün-de]1 [vP araba [VP t1 gid-er ] street-GEN top-AGR-LOC car go-AOR ‘It is cars that go on/over the street’ d.. First.. [DP [DP Ø 2] üst-ün-de]1 [vP araba [VP t1 gid-en] cadde2 top-AGR-LOC car go-SR street ‘the street that cars go on (the top of)’ We see the same phenomena with transitive verbs.c. TP]. 145 . and in (6)b we see that the relativized expression in (6)a bears inherent ablative case prior to promotion. 4) a. TP] position. notice that as shown in (4)b for the unacceptable example in (4)a. as shown in the sentence in (5)b. On the other hand. an accusative direct object cannot move to T. CP TP +WH-DP-child+ACC vP +Wh-DP-child+ACC NP-cat VP +Wh-DP-child C° T° v° V° claw It is not surprising that a structurally case-marked element cannot move to another structural case-assigning position. we have seen that Turkish allows inherently case marked elements in the [Spec. The relativized expression in (5)a bears inherent locative case prior to movement to [Spec.

the possessor child is free to raise out of the DP to T. (Bu) tarla-da mı r yeti iyor. the DP [child’ s arm] is barred from moving to [Spec. TP]. 7) a. ii) although an accusative direct object is higher than an in-situ subject. The derivation of (7)a appears in (7)b. ‘A mouse/mice is/are coming /came out of the hole. the subject no longer intervenes between the object and T. Whereas the accusative object itself. Once the direct object has raised to [Spec.out-SR hole ‘the hole that mice come out of’ b.prog. and iii) inherently case-marked expressions are permitted in [Spec. Delik-den fare çı -yor. The example in (7)a demonstrates that a genitive possessor of an accusative object can also move to [Spec. TP]. as in move . [mı r yeti -en] tarla sı corn grow-SR field ‘the field where corn grows’ b. vP].’ Returning to the issue of Minimality. we have so far determined that i) an intervening NP subject blocks raising of a +Wh-expression to T. [ [Ø 1 kol-u]-nu kedi tı rmala-yan] çocuk1 arm-POSS-ACC cat claw-SR child ‘the child whose arm a cat clawed’ 146 .5) a. TP] as evidenced by the acceptability of the SR form. sı (this) field-LOC corn grows ‘Corn grows on (this) field’ k-an] delik 6) a. the object is barred from moving to T. [fare çı mouse come.out-pres. kı hole-ABL mouse come.

the relativized expression can move to [Spec. The conclusion is that as long as the relativized expression can get around the subject. 2 The Problem of Human Subjects Having determined that the SR form is only licensed when the relativized expression can move to. TP]. it is only the genitive possessor of the accusative object that raises in from the vP to [Spec.b. CP]. indeed. TP] and license the SR form. TP] after which it again raises to [Spec. the direct object [child’ s arm] is not +Wh. [Spec. Thus. is. triggering the SR form. and the subject itself does not have to. CP TP +Wh-child vP DP+ACC NP-cat VP DP +Wh-child +GEN arm C° T° v° V° claw D° Crucially. and through. child. cannot raise to T. only the element in its Spec. we saw that this move must obey Minimality. We also saw that raising of a possessor out of an accusative object which itself has raised above a non-specific NP-subject avoids a Minimality violation. The problem 147 .

*[[Ø 1 yavru-su]-nu avcı (-lar) yiy-en] geyik1 young-POSS-ACC hunter(s) eat-SR deer ‘the deer whose young a hunter/hunters ate’ c. [[Ø 1 ev-i]-ni fı na yı rtı k-an] aile1 house-POSS-ACC tornado raze-SR family ‘the family whose house a tornado razed’ b. [[Ø 1 çocu -u]-nu arslan yiy-en] anne1 child-POSS-ACC lion eat-SR mother ‘the mother whose child a lion/lions ate’ b. [[Ø 1 kayı -ı ]-nı nehir sürükley-en] balı 1 kçı boat-POSS-ACC river drag-SR fisherman ‘the fisherman whose boat a river dragged’ b. the SR form is acceptable only when the subject is –human. The near-minimal sets (8) through (10) are transitive constructions where the relativized element is the possessor of the accusative direct object. *[[Ø 1 ev-i]-ni i çi/asker(-ler) yı k-an] aile1 house-POSS-ACC worker/soldier(-pl) raze-SR family ‘the family whose house (a) worker(s)/soldier(s) razed’ 10) a. it results in unacceptability when the subject is +human. [[Ø 1 yavru-su]-nu avcı -lar-ı n ye-di -i] geyik1 young-POSS-ACC hunter-pl-GEN eat-NSR-3s deer ‘the deer whose young the hunters ate’ 148 . a +human subject results in unacceptability of the SR form. 8) a.is that whereas this strategy generally holds. *[[Ø 1 kayı -ı ]-nı köylüler kı -ya sürükley-en] balı 1 yı kçı boat-POSS-ACC villagers shore-DAT drag-SR fisherman ‘the fisherman whose boat villagers dragged onto the shore’ 9) a. [[Ø 1 ev-i]-ni belediye yı k-an] aile1 house-POSS-ACC municipality raze-SR family ‘the family whose house a municipality razed’ c. However.

Furthermore. which it cannot get unless it has raised to T. [pro bir kı n yan-ı z-ı n-da ol-du -u]-nu gör-dü-ler.’ b. vP]. one girl-GEN side-his-LOC be-NSR-3s-ACC see-PAST-3pl ‘They saw that (of the salient girls) one (of them) was by his side. I argue that the genitive subject in the NSR form is in [Spec. as long as the subject is specific. z side-his-LOC one girl be-NSR-3s-ACC see-PAST-3pl ‘They saw that there was a girl by his side. 2002)and Chomsky (2005) to which the scrambled expression adjoins. (i) [avcı -lar-ı n [vP Ø 1 yavru-su]-nu ye-di -i] ] geyik1 hunter-pl-GEN young-POSS-ACC eat-NSR-3s ] deer ‘the deer whose young the hunters ate’ 132 149 . We assumed that locatives mark the edge of the VP (Kural 1992). TP] and the subject intervenes. this order is not possible with the SR form because the Whgeyik is A-moving to [Spec. but I will point out that the unscrambled version of the NSR form is also acceptable as in (i) for example (13c) because the Wh-expression geyik undergoes A-bar movement from with the accusative DP in [Spec.) It is beyond the scope of this work to discuss this scrambling further. TP]. We assumed case is assigned in a Spec-Head configuration and that the EPP of a case assigning head attracts a DP to its Specifier. (11)b becomes acceptable in the NSR form (11)c. The subjects in both (11)a and (11)b are animate. (There may be a projection v*P above vP but below TP as suggested by Lasnik (1998.’ Subjects of embedded clauses in Turkish generally bear overt genitive case. and has moved to T132. but in the unacceptable (11)b. We can assume this to be so because the subject bears genitive case. Support for this assumption appears in (12).As demonstrated in (11). so I assume that the possessorpossessee DP-object scrambles out of the vP. the subject is +human. We also assumed that non-specifics are NPs that neither satisfy the EPP nor require case. the embedded subject of the copular Note that the accusative direct object in the NSR form is higher than the subject in these examples. the difference does not lie in animacy. 11) a. As expected. to a position above the subject. In (11)a. pro [yan-ı n-da bir kı ol-du -u]-nu gör-dü-ler.

No other animacy heirarchy among the arguments is playing a role in the grammaticality of the examples. In (11)b the locative is above the non-specific subject which bears no case morphology. And. In (12) below. In all these examples. the same pattern obtains. TP]. 133 I assume that the subject of a copular structure merges as a complement of V° and that locatives adjoin to VP. In examples (8) and (9) above.134 As we see (12)c-d. it must be a DP. where the embedded clause is an existential construction.” (i) ??bir doktor otur-an ev one doctor live-SR house ‘a house in which a doctor lives’ (ii) bir köpek bul-un-an ev one dog find-PASS-SR house ‘a house in which there is a dog’ (Literally: ‘a house in which one finds a dog’ ) 150 . The SR form becomes unacceptable when the non-specific subject is +human. it must receive a specific interpretation. possessor and possessee are both human. In (10)a. in (13). 134 Zimmer (1987) notes that the RC with the human subject in (i) is quite marginal whereas (ii) is fine. whereas genitive case on the subject can be viewed as evidence that the subject is in [Spec. TP].133 It must also receive a specific interpretation. Returning then to the acceptable (10)c. they are both animate. Thus a locative in a copular sentence merges above (and linearly to the left) of the subject. because the subject bears genitive case. the possessor is human and the possessee is animate. the NSR form is acceptable for both human and non-human subjects with the stipulation that the subject be specific. the direct object is composed of a human possessor and an inanimate possessee. and based on our assumptions. and in (10)b. This contrasts with (11)b.clause has raised above the VP and bears genitive case. A non-specific subject must remain in-situ and must be bare. it must be in [Spec. He suggests that this is “ presumably because humans are inherently more individualized and topicworthy than non-humans. the possessor is inanimate and the possessee is human. Let’ s look at several alternatives.

In Turkish.12) a. I include the minimal pairs in (13).house/publisher search-SR article ‘the article whose author publishing. In Turkish. as shown in (13)b. [[Ø 1 ine -i]-ni tren ez-en] köylü1 cow-POSS-ACC train run. the +human ‘publisher’ is not. 13) a. and ‘publish’ plus the word ‘house’ ev denote the ncı non-human organizational ‘publisher’ . the words ‘publisher’ and ‘publishing house’ are similar to the English in that they contain the same root plus a morpheme. whereas ‘publishing house’ is acceptable as the subject of the SR clause.over-NSR-3S peasant ‘the peasant whose cow the train ran over’ To rule out as much as possible the other factors that may be responsible for the difference. [[Ø 1 ine -i]-ni oför-ün ez-di -i] köylü 1 cow-POSS-ACC driver-GEN run. or ‘publishing house’ . and case-marked. +human noun ‘publisher’ yayı . [[Ø 1 yazar-ı ]-nı yayı nevi /*yayı ncı aray-an] makale1 author-POSS-ACC publishing. *[[Ø 1 ine -i]-ni oför ez-en] köylü 1 cow-POSS-ACC driver run. but only if specific. the purpose of this chapter is to find a principled syntactic explanation as to why this may be so— it demonstrates that the phenomena has been previously remarked upon. 151 . yayı nevi.over-SR peasant Intended: ‘the peasant whose cow a driver ran over’ c.houses/*publishers are looking for’ Although I do not find this explanation adequate— indeed. [[Ø 1 ine -i]-ni tren-in ez-di -i] köylü1 cow-POSS-ACC train-GEN run.over-NSR-3S peasant ‘the peasant whose cow the/*a driver ran over’ d. As demonstrated in (13)a. ‘publish’ yayı plus the ‘-er’ morpheme -cı denote the n . Both are possible as clausal subjects.over-SR peasant ‘the peasant whose cow a train ran over’ b.

in intransitive sentences and 2. A aç-tan bir çocuk dü -tü tree-ABL one child fall-PST ‘A child fell from a tree’ b. rather than animacy and definiteness. *Bir elma a aç-tan dü -tü one apple tree-ABL fall-PST 15) a. I show that the contrast is more aptly described in terms of human/non-human features and specificity. Bir çocuk a aç-tan dü -tü one child tree-ABL fall-PST 135 Most of the examples (14)-(20) are from Erguvanlı (1984) which includes the following observation based on animacy: [-animate] indefinite subjects are restricted to the immediate left of the verb 1. [[Ø 1 yazar-ı ]-nı author-POSS-ACC ‘the article whose author the publishing. i. A aç-tan bir elma dü -tü tree-ABL one apple fall-PST ‘An apple fell from a tree’ b. a +human subject leads to unacceptability.135 As shown in (14).b. 14) a. a non-human nonspecific subject is restricted to the immediate preverbal position. Interestingly. 152 . must remain in situ.e.house-GEN search-NSR-3S article /publisher-GEN 3 Toward a Solution In the previous section. we saw that in otherwise acceptable transitive SR relative clauses. there are other instances in Turkish where a +human nominal behaves differently from its nonhuman counterpart. The examples in (15) show that this restriction does not hold for +human nonspecific subjects.in transitive sentences with a definite [+animate] DO. in matrix sentences.house/the publisher is/was looking for’ yayı nevi-nin /yayı -nı ncı n ara-dı -ı ] makale1 publishing.

This contrasts with +human direct objects which must bear accusative case. I include the past tense example in (16)e to demonstrate that the demonstrative is not necessary to denote specificity.e.137 A definite or specific interpretation is achieved by means of a demonstrative. (i) has the interpretation of ‘I like apples [not some other fruit]’ and can only be uttered as a contrastive response to someone else’ s utterance about some other fruit. 16) a. *Ben insan-lar sev-er-im I human-pl like-AOR-1s 136 Use of the aorist tense in (16)d would have been odd without the demonstrative ‘this’ . Ben elma sev-er-im I apple like-AOR-1s ‘I like apples’ b. the sentence would mean ‘I like the apples (we grow rather than the pears)’ . Note that overt case on ‘apples’ in (16)d-e)136 yields a specific interpretation. *Ben elma-lar sev-er-im I apple-pl like-AOR-1s c. *Ben elma-lar-ı sev-er-im I apple-pl-ACC like-AOR-1s d. i. (17)d-e.Non-specific non-human direct objects cannot have case (16)b (or a plural marker (16)c). as shown in (17)a-c. *Ben insan sev-er-im I human like-AOR-1s b. the non-plural example with accusative case is also unacceptable with the unmarked interpretation of ‘I like apples’ . (i) Ben elma-yı sev-er-im I apple-ACC like-AOR-1s 153 . Ben elma-lar-ı sev-di-m I apple-pl-ACC like-PST-1s ‘I liked the apples’ 17) a. Ben bu elma-lar-ı sev-er-im I these apple-pl-ACC like-AOR-1s ‘I like these apples’ e. 137 Although Erguvanlı (1984) did not include it.

’ 154 . Ben bu insan-lar-ı sev-di-m I these human-pl-ACC like-PST-1s ‘I liked these people’ Although Turkish has relatively free word order. *[bir bahçe] [bir adam] suluyor one garden one man watering f. even a non-specific one.c. *[bir bahçe]-yi [bir adam] suluyor one garden-ACC one man watering e. [bir adam] [bir bahçe] suluyor one man one garden watering ‘A man is watering a garden’ b. [bir adam] [bahçe]-yi suluyor one man garden-ACC watering ‘A man is watering the garden’ d. bahçe-yi [bir adam] suluyor garden-ACC one man watering ‘The garden. a man is watering (it). Ben bu insan-lar-ı sev-er-im I these human-pl-ACC like-AOR-1s ‘I like these people’ e. The exception is (18)f where the object is in Topic position. Ben insan-lar-ı sev-er-im I human-pl-ACC like-AOR-1s ‘I like people’ d. the transitive constructions in (18) demonstrate that a non-human object can never precede a +human subject. must be D-linked. i.e. 18) a. [bir adam] [bir bahçe]-yi suluyor one man one garden-ACC watering ‘A man is watering a (specific) garden’ c.

the human direct object cannot have its -features checked by AgrS because it is barred from moving to T.138 No such restriction applies to non-human specific objects. iv. in (19) the +human object must be the highest expression in the verbal domain so that its -features can be checked by AgrO. bir araba yol-u tı kamı one car road-ACC blocked ‘A car has blocked the road’ Let’ s take stock: i. ii. *bir sürpriz Ali-yi ev-de bekliyor one surprise Ali-ACC home-LOC waiting 20) a. a non-human non-specific subject cannot precede a +human specific object. +human expressions have -features. Yol-u bir araba tı kamı road-ACC one car blocked ‘A car has blocked the road’ b. it does not matter whether it is the object or the subject that checks the Agr features in (20). iii. and will argue that a partitive construction is used to denote non-specificity for humans. Ali-yi ev-de bir sürpriz bekliyor Ali-ACC home-LOC one surprise waiting ‘A surprise is waiting for Ali at home’ b. Whereas non-specific –human direct objects cannot have case (or a plural morpheme). it is in its base position. This point is purely speculative. Perhaps. +human direct objects must bear overt case. Being marked with structural accusative case. Whereas –human non-specific subjects must remain in situ. it seems must) raise from their base positions. however. 155 . As we will see later in this chapter.As shown in (19). 138 I will shortly demonstrate that a non-specific human nominal is not possible in Turkish. 139 I am not sure what position bir araba ‘a car’ occupies in (20)b. I assume that in (20)a. Although I have not said anything throughout about Agr projections.139 19) a. +human nonspecific subjects can (in fact. it may be the AgrO projection in the verbal domain that causes the difference. A non-human object can never precede a +human subject. though. A non-human non-specific subject cannot precede a +human specific object. or rather are -complete.

the +human Wh-expression who must always bear accusative case when it is a direct object. in the examples below. (iii) and (iv) above.140 Let’ s look at the contrast between who and what. just like DPs +human arguments must bear case.c. let’ s first look at more evidence and the consequences of such an assumption. The facts in (i) also follow if we assume that +human subjects are DPs: they must raise from their base positions. and be case-marked. *pro kim unut-tu-n? pro who forget-PST-2s 140 from Kural (p.) 156 . As shown in (21) and (22). and someone and something. According to Kural (1992) +human direct object Quantifiers (QPs) and Wh-phrases must bear accusative case. pro kim-i unut-tu-n? pro who-ACC forget-PST-2s ‘Who did you forget?’ b. no such requirement holds for what. and just like DPs. they must be DPs. we must conclude that +human nominals can never be NPs. they must raise to case assigning positions. The assumption that +human arguments are obligatorily DPs is supported by evidence from Quantifiers and Wh-phrases. Thus. Before I explain the nature of the +human DP.These facts combined lead us to conclude that +human arguments must be DPs. Extending this assumption to the +human facts. (ii) above. Recall that I had proposed that only DPs can raise and receive Case. 21) a. including the issue of specificity.

pro [ne-yin kı l-dı -ı rı ]-nı biliyor-mu? pro what-GEN broke-NSR-3s-ACC know-Q-3s ‘Does he know what (specific thing) broke’ When a Q-morpheme is added to (26b). This contrasts with (24).e. No such requirement exists for the non-human wh-subject ne ‘what’ in (24)141. pro [kim-in gel-di -i]-ni gör-dün? pro who-GEN arrive-NSR-3s-ACC saw-PST-2s ‘Who did you see arrive?’ 141 The example in (23)b is an interrogative even though the matrix verb does not bear the Question morpheme –mI. making (ii) an interrogative in contrast to (27). the +human wh-subject kim ‘who’ must bear overt case. This can be verified by embedding the question in a complement clause. which can only be a question if the matrix verb has a Q-marker. *pro [kim gel-di -i]-ni gör-dün? pro who arrive-NSR-3s-ACC saw-PST-2s ‘Who did you see arrive?’ b. pro ney-i unut-tu-n? pro what-ACC forget-PST-2s ‘What (specific thing) did you forget?’ b. (i) a. pro [ne kı l-dı -ı rı ]-nı biliyor-mu? pro what broke-NSR-3s-ACC know-Q-3s Does he know what broke?’ b. i. the interpretation becomes ‘Did you see who arrived?’ I will not speculate here as to what may motivate this difference. (ii) Ne kı l-dı -ı san-ı r-ı -nı yor? what break-NSR-3s-ACC believe-3s ‘What does he think/believe broke?’ 157 . as in (i). pro ne unut-tu-n? pro what forget-PST-2s ‘What did you forget?’ Nominative case is null. In (23). it raises to T and bears the Ø nominative morpheme. 23) a. In the ECM example in (ii) with the verb san ‘believe/think’ . the bare what takes matrix scope.22) a. but we can assume the same holds for who when it is a subject.

pro [ne kı l-dı -ı rı ]-nı biliyor pro what broke-NSR-3s-ACC know-3s ‘He knows what broke’ b. pro something forget-PST-2s ‘You forgot something’ 27) a. pro something-ACC forget-PST-2s ‘You forgot some (specific) thing’ b. pro [bir ey(-in) kı l-dı -ı rı ]-nı biliyor pro something(-GEN) broke-NSR-3s-ACC know-3s ‘He knows some (specific) thing broke’ 158 . whereas something need not (26). And. In (25). pro bir ey-i unut-tu-n. 25) a. pro someone forget-PST-2s 26) a. pro someone-ACC forget-PST-2s ‘You forgot someone’ b. pro bir ey unut-tu-n. pro biri-ni unut-tu-n. pro [ne-yin kı l-dı -ı rı ]-nı biliyor pro what-GEN broke-NSR-3s-ACC know-3s ‘He knows what (specific thing) broke’ The same requirement holds for Quantifiers. *pro biri unut-tu-n.24) a. as expected the QP someone must bear genitive case when it is the embedded subject (27)a whereas no such requirement exists for something (27)b. pro [biri*(-nin) gel-di -i]-ni biliyor pro someone-GEN come-NSR-3s-ACC know-3s ‘He knows someone came/is coming’ b. we see that the +human QP someone must also obligatorily bear case.

in Persian (an SOV Indo-European language). it would have to raise to T and get case. a direct object ‘someone’ without the –ko marker is unacceptable. p. che /yek chizi did-im b. *ki /*kesi did-im b. as shown in (29). as in (i).c. TP] is occupied by a non-Whelement. h r s b h m k h iz bhul jati h every morning I somethings forget PROG be ‘every morning I forget some (specific) thing’ (ii) a. We saw that the SR relative clause is licensed for non-subject extraction when the subject remains low in the structure. the NSR is triggered. p. Repeating the tree in (7) as (28). A non-human universal quantifier direct object can raise or remain in situ.’ ‘We saw some (specific) thing. but a human direct object must obligatorily raise (T.) ‘something’ as a direct object can optionally be marked with the specificity marker –ko. we can conclude that they must obligatorily have raised to a case-assigning head. (ii).c. Chandra. On the other hand. When [Spec. 159 . the direct object case morpheme –ra is obligatory on Wh-words and Quantifiers denoting humans (ii) but is not permitted for non-human nominals unless they are D-linked and specific (i). (i) a. *h r partime m kIsI bhul jati h every party I someone forget PROG be b. h r partime m kIsI-ko bhul jati h every party I someone-KO forget PROG be ‘every party I forget someone’ In Afrikaan. 142 Similar facts can be found cross-linguistically.142 We saw that Case in Turkish is assigned in a Spec-head configuration. If the subject were a DP.The QP and Wh-facts support our earlier conclusion that +human arguments must always bear case. direct objects raise when they are specific. note that we had assumed the nonspecific subject cat was an NP that must remain in-situ. che-ra /yek chizi-ra did-im? what/one thing see-1p what-ACC/one thing-ACC see-1p ‘What did we see?’ ‘What (specific) thing did we see?’ ‘We saw something. leaving [Spec. Because +human arguments are always case-marked. (i) a.’ In Hindi (P. TP] vacant for the Whexpression. h r s b h m k h iz bhul jati h every morning I somethings forget PROG be ‘every morning I forget something’ b. ki-ra /kesi-ra did-im? who/someone see-1p who-ACC/someone-ACC see-1p ‘Who did we see?’ ‘We saw someone.’ (ii) a.). For example. Biberauer.

[ [Ø 1 kol-u]-nu kedi tı rmala-yan] çocuk1 arm-POSS-ACC cat claw-SR child ‘the child whose arm a cat clawed’ b.DP-cat VP DP +Wh-child +GEN arm C° T° v° V° claw D° Thus.28) a. CP TP +Wh-child vP DP+ACC NP-cat VP DP +Wh-child +GEN arm C° T° v° V° claw D° 29) a. By assuming that +human nominals are obligatorily DPs. the SR form is not possible for non-subjects when the clausal subject is a DP. CP TP DP-cat +GEN vP DP+ACC t. we can account for the fact 160 . [ [Ø 1 kol-u]-nu kedi-nin tı rmala-dı -ı çocuk1 ] arm-POSS-ACC cat-GEN claw-NSR child ‘the child whose arm the cat clawed’ b.

otobüs-ler geldi-ler bus-pl came bus-pl came-pl ‘the buses came (at one time)’ ‘the buses came (series of comings/*all at one time)’ Thus. Plural agreement with an overt non-human subject usually denotes a plurality of events.143 Adopting Postal (1966). 161 . TP] will always be occupied by the +human DP-subject. otobüs-ler geldi b. yolcu-lar geldi-ler traveler-pl came traveller-pl came-pl ‘the travelers came’ ‘the travelers came’ ii. as in (iiib) and (ivb). I suggest that +human nominals in Turkish must always first merge with a null pronoun which gives the nominal its -features. Adopting a version of Longobardi (1994). Turkish actually has a “ poor” agreement system and that the verbal agreement that does exist is triggered by the presence of (or perhaps required for interpretability by) -features which appear only on +human nominals. a. and is default or defective agreement for +human arguments. i. a. 143 Interestingly. number agreement on the verb in Turkish is permitted only when the subject is +human. TP]. and that the null 3rd person singular verbal agreement is in fact no agreement at all for non-human arguments. a. köpek-ler hı rladı -lar dog-pl growled dog-pl growled -pl ‘the/some dogs growled’ ‘the/some dogs growled (on different occasions)’ iv. Kornfilt (2005) suggests that in Turkish “ agreement has pronominal features” . yolcu-lar geldi b. a. ku -lar ötüyor b. Consequently. I assume that D˚ and pronouns are essentially the same thing. it may be that contrary to what is believed. specifically -features. 4 Explaining the Behavior of Human Nominals We have determined that +human nominals in Turkish have the hallmarks of DPs. *ku -lar ötüyor-lar bird-pl chirping bird-pl chirping-pl ‘the/some birds are chirping’ iii. a bundle of features.that the SR form is banned when the subject is human: [Spec. They must be case-marked and they must raise to functional projections. I propose that +human nominals in Turkish have the structure as in (30). the +Wh relativized element will not be able to move to [Spec. and the SR form will be barred. It seems reasonable to assume that -features must obligatorily enter into an Agree relation. köpek-ler hı rladı b.

water comes down from. An exception is (34).the hills b. *Acqua viene giù dalle colline. when the proper name is the last name of a female. the article is obligatory. we looked at some of Longobardi’ s arguments. comes down water from. Gianni called me up b. the Callas/Callas sang 162 . (a/the) great friend of Maria called me up However. Let’ s revisit some of them here. Gianni mi ha telefonato. 31) a. 33) a.the hills 32) *(Un/Il) grande amico di Maria mi ha telefonato. Viene giù acqua dalle colline. Il Gianni mi ha telefonato.30) DP NP+human D° (= Ø -pronoun= -features) athlete In Chapter 3 on Specificity. we saw that Italian sentences require a D in the preverbal subject position. the Gianni called me up 34) La Callas/*Callas ha cantato. as in (31) and (32). as in (33). First. an article is optional when the subject is a proper name.

would be an event of ‘cat-scratching’ which would have Higginbotham (1987) proposed that an argument is “ saturated” and can thus be assigned a theta role.. Translating this idea into an “ event-ish” semantic interpretation.. which is neutral. DP D° Il D° Ø NP NP N° Gianni N° Gianni It becomes understandable why an overt determiner is obligatory for last names of females.. Il Gianni . there is in fact a null D head. Stowell (1989) has shown that NPs are non-referential. I had adopted Mandelbaum’ s (1994) proposal that predicate NPs are basically adjectival. a sentence with an NP subject like cat in (36)a). the feminine gender feature is not carried on the last name. 144 163 . Thus. 35) a. By extension Szabolcsi (1987). Gianni . with a determinerless proper name as subject. (33)a) and (33)b) are almost identical except that in (33)a). The structural difference in the subjects is shown in (10). whereas DPs are referential. there has been head movement of N to a null D. and that by extension NPs can only be nominal predicates144. DP b.. with the result that there is too much -feature information on the D for it to be null (or vacuous). I had suggested a strong hypothesis: that only DPs can be arguments. I assume that this gender feature is borne by D. Abney (1987) and Longobardi (1994) have argued that NPs are nominal predicates (unsaturated) and do not bear a theta-role and DPs are arguments that do bear a theta role. (33)a) is not an exception to the requirement of a DP as subject. For Turkish. and that there has been N to D movement. Whereas a name like Marie carries a gender feature.Longobardi assumes that in sentences such as (33)a) above.

some cat or other) scratched his arm’ b. c.e. The flour has weevils. b. The boy has a needle. kedi [ pro kol-u]-nu tı rmala-dı cat arm-POSS-ACC scratch-PST ‘The cat scratched his arm’ b. My proposal is that in Turkish (and perhaps in other languages as well146). The tree has branches. the cat) & Scratching (e) & Theme (e. (ii). When the subject of ‘have’ is non-human. his arm)] 37) a. See Chapter 3 and reasons cited there. x) & Scratching (e)] & Theme (e. the semantic interpretation for (37)b) would be there is an event of ‘scratching’ which has cat as the Agent and his arm as the Theme.his arm as the Theme. the theme must be inalienably possessed (or characteristically treated as such). He also includes the following data from English in regards to the interaction between +human subjects and locations and the verb ‘have’ . as in (i). The flour has a ring *(in it). This differs from the sentence in (37)a) with a DP subject in that we now have an external argument cat. = There is a ring in the flour. it is the lexicon. b. ∃e[∃x:Cat (x) [Agent (e. or more aptly. The tree has a nest *(in it. 146 Freeze (1992) includes interesting cross-linguistics data for existentials. The boy has a cousin/nose. ∃e [Agent (e. as shown in (36)b)145. 36) a. (ii) a.) = There is a nest in the tree. (i) a. No such restriction holds for human subjects. his arm)] The question remains as to why a +human subject is not allowed by the grammar to be predicative or adjectival as in (36)a. nominals in the lexicon marked as having +human features 145 I point out again that I reject the idea that the non-specific subject incorporates into the verb. Thus. c. 164 . d. The boy has fleas. [pro kol-u]-nu kedi tı rmala-dı arm-POSS-ACC cat scratch-PST ‘A cat (i. Does the restriction reside in the syntax or the semantic component? I suggest that it is the syntax that drives this constraint.

Pronouns are inherently human because they contain. As for possession. In addition. *The shelf has a mongoose. that has consequences in Turkish. which if we take Postal’ s work seriously is equivalent to human+D. it is only that such a requirement exists that is germane here. c. I am not committed as to the exact nature of the selectional requirement and how it is satisfied.) (iv) a. which is consistent with what I am proposing for Turkish. There is a ring in the flour. Lupe has a book. it is also possible that a +human nominal must merge with a null-pronoun. A mongoose is on the shelf. TP] of the verb ‘have’ . features. (i) a. absence of -features means absence of person and number features.149 Note that whereas a non-human location permits existential predication with a theme. in overt syntax forming a DP which then merges into a theta position. 148 Of course. This human+D expression. Thus. It is noteworthy that number features seem to add structure to a nominal. in fact.148 The +human element that enters the derivation. There is a mongoose on the shelf.have a selectional requirement for -features. but that non-human possessors can not. and so must first-merge into a theta position. is prevented (either by the syntax or the LF interface) from being predicative/adjectival. I assume this is so because this pronoun lacks -features. if we assume a thematic heirarchy with inalienable possession being the lowest. they. is a composite of human+ -features which is identical to human+(null)pronoun. *A book is with/at/by Lupe. ( The boy has a needle. (v) a. b. I assume that a +human nominal must first merge with -features in the lexicon before it can enter a derivation. as shown in the existential/’ have’ constructions in (v) and (vi). c. Thus. Note that this excludes the 3rd person plural pronoun. b. There is a needle on the boy. (iii) a.e. in the good SR examples where the RC subject is an NP. There is a nest in the tree. 147 Note that the 3rd person singular pronoun in English it cannot refer to humans: *It came. the human location prefers to be the subject of a ‘have’ construction. 149 Higginbotham (1987) argues that the copula in “ John is the director” is different from the copula in “ John is a director” in that the former expresses identity and is referential whereas the latter is predicative. he argues that in English definite descriptions can be predicative as the embedding of [John (is) the man] without an overt copula in (ia) demonstrates. changing the subject to a plural leads to degrading or unacceptability. (iii). human “ locations” require the ‘have’ construction in contrast to non-human locations which can only appear in existential or copular constructions. *There is a book with/at/by Lupe. Although discussing the nature of human/non-human nominals in English is beyond the scope of this thesis. b. I consider [John [the man (for the job)]] 165 .147 As for non-pronominals. are -features. i. and that they must raise to [Spec. then. as with all DPs. c. In fact. for example a NumP. one could argue that the data on possession in (i) and (ii) suggests that human possessors can raise within this heirarchy. the data here demonstrates that +human locations are not acceptable in existential “ there” constructions.

In (38)c. It would also explain why an overt determiner is required for female last names in Italian as we saw in (34). As we will shortly see. a DP must also enter into an Agree relation. CP]. This means that it must have first merged with a null-pronoun (or D°). on the other hand. it can satisfy the EPP. TP] in order to receive case. As a DP. which would explain why a proper name may raise to D° in (10)a. For a non-subject to move to [Spec. and so must move to [Spec. we are now in a position to account for the unacceptable SR form in (38)b. Thus I will stick to the strong hypothesis that a DP cannot be a predicate. The unacceptability of (38)b is due to a Case Filter violation. the non-Wh-subject oför ‘driver’ is +human. TP]. singular (perhaps masculine) -features. I assume the accusative object [DP [DP peasant’ s] cow] has scrambled over the subject. after which the possessor [DP peasant] raises from the specifier of the DP to [Spec. *I consider [John [the man (standing over there)]] I would argue that (ia) actually contains a small clause. the feminine feature on the proper name (assuming it exists only on last names of females) would create a feature mismatch on D°. as in “ the man (who is best) for the job” . 166 . In (38)b. the +human subject has raised to [Spec. Returning to the Turkish examples in (12) repeated as (38). I should point out that the nature of D° must be parametric cross-linguistically. Before we return to Turkish.Being a DP. may be either an NP. A null D° in Italian may have default 3rd person. TP]. A non-human subject. Recall that the SR form is licensed when a +Wh-expression moves to [Spec. it must satisfy the Case Filter. the subject had to be an NP so that it would not need case. and it must receive case. as b. TP] and received case.

in (38)a, leaving [Spec, TP] vacant for the Wh-expression, or a DP, as in (38)d, in which case it must raise to [Spec, TP] to be assigned case. 38) a. [[Ø 1 ine -i]-ni tren ez-en] köylü1 cow-POSS-ACC train run.over-SR peasant ‘the peasant whose cow a train ran over’ b. *[[Ø 1 ine -i]-ni oför ez-en] köylü 1 cow-POSS-ACC driver run.over-SR peasant Intended: ‘the peasant whose cow a driver ran over’ c. [[Ø 1 ine -i]-ni oför-ün ez-di -i] köylü 1 cow-POSS-ACC driver-GEN run.over-NSR-3S peasant ‘the peasant whose cow the/*a driver ran over’ d. [[Ø 1 ine -i]-ni tren-in ez-di -i] köylü1 cow-POSS-ACC train-GEN run.over-NSR-3S peasant ‘the peasant whose cow the train ran over’ We saw in (25) and (26), repeated as (39)and (40) respectively, that there is a contrast with respect to case-marking and specificity based on whether an argument is +human. A case-marked non-human argument, as in the accusative ‘something’ bir ey-i in (40)a, must receive a specific interpretation (a specific “ thing” ). This is expected from what we know about overt case and specificity. However, the bare someone as a direct object is unacceptable in (39)b, and the case-marked someone does not denote a specific person. How is it that a case-marked expression can yield a non-specific interpretation, in this case, a non-specific human? 39) a. pro biri-ni unut-tu-n. pro someone-ACC forget-PST-2s ‘You forgot someone’ b. *pro biri unut-tu-n. pro someone forget-PST-2s

167

40)

a. pro bir ey-i unut-tu-n. pro something-ACC forget-PST-2s ‘You forgot some (specific) thing’ or ‘You forgot a/one thing’ b. pro bir ey unut-tu-n. pro something forget-PST-2s ‘You forgot something’

I suggest non-specificity for human nominals is achieved in Turkish by means of a partitive construction. Support for this proposal comes from the fact that the word someone in Turkish is in fact bi-morphemic: one plus possessive agreement, as shown in (41)a. I assume that the structure of bir-i is actually “ of them, one” as in (41)b, with a null pro denoting the of them in the specifier of D°. This null pro triggers the possessive agreement on bir ‘one’ . This DP-NP structure is analogous to an existential IP with the NP-one in the restrictor and the pro defining the scope. Because D° is a case-assigner, the pro in the Spec of the DP is assigned case. The NP does not require case, but the entire DP must be case-marked. The partitive itself is specific, i.e. a DP, but the expression in its restrictor can be non-specific. 41) a. Bir-i gel-di. one-3s come-PST ‘Someone came’ b.
pro NP one+agr D° = Ø -pronoun +human DP

168

There are two partitive constructions in Turkish: one with genitive case on the superset nominal, and the other with ablative case, as in (42)a and (42)b respectively150. 42) a. Ali kadı n-lar-ı n iki-si-ni tanı yordu. Ali woman-pl-GEN two-AGR-ACC knew ‘Ali knew two of the women’ b. Ali kadı n-lar-dan iki-si-ni tanı yordu. Ali woman-pl-ABL two-AGR-ACC knew ‘Ali knew two of the women’ Let us look closer at the derivation of these partitives. Beginning first with (42)b, I assume that the structures begin by merging the superset DP-women with the subset NP two X’ s, where X is of the same type as the superset, as in (43)a where the index merely denotes “ of the same type” . In (42)b the DP-women is a PP-like expression “ from/of the women” and as with most Turkish PPs is inherently case-marked, in this case with ablative case. As shown in (43)b the merge of the PP/DP and the NP projects to an NP which merges with a D°. D° has an EPP feature and attracts the ablative PP/DP-women to its specifier triggering agreement on the NP151, as in (43)c. Note that I refer to the phrase “ of the women” as a PP/DP-women. This is because the structure of this expression is actually a DP with ablative case; there is no PP projection above the DP. In this structure, there is no issue of minimality in the movement of the PP/DP to [Spec, DP]. In an English PP, the DP would be raising

150

151

Examples from Enç (1991). I remain agnostic as to where and how agreement is actually triggered. There may be an Agr projection (both in the nominal and verbal domains), but I have ignored this completely as it would take me too far afield. What is relevant, is that an agreement morpheme shows up on the phonologically contentful N°. Kornfilt (2005) offers an account of Agr in these constructions, but the analysis is at right angles with what I am proposing here.

169

from the complement position of P°, and thus be c-commanded by the NP two. However, the Turkish PP has the structure of a DP, i.e. it is not embedded in a PP projection. This is shown in the tree in (44) for the derivation in (43). 43) a. [PP/DP women1]-ABL + [NP two X1] b. [NP [PP/DP women1-ABL] + [NP two X1]] + D° c. [DP [PP/DP women1-ABL] [NP [PP/DP women1-ABL] + [NP two X1]] + D°+AGR]

44)

DP women-ABL NP DP women-ABL

D°= Ø -pronoun +human NP two+agr

The derivation of the partitive in (42)a is almost identical to (43)-(44) except that the PP/DP women does not have inherent case and needs to be assigned structural case. Looking at the derivation for (50)a in (45), we see that a possessive PP/DP (of) women merges with the NP two X’ s in (45)a. This structure then merges with D°, and the PP/DP (of) women raises to [Spec, DP], triggering agreement. In this derivation, D° assigns genitive case to the DP in its Spec.152 45) a. [PP/DP (of) women1] + [NP two X1] b. [NP [PP/DP (of) women1] [NP two X1]] + D°

152

This is basically a possessor construction. It may be that there is a null of in the PP/DP (of) women. A possibility that I find appealing is that the specifier of NP may be a possessor theta position and that an element merged here picks up a possessor theta role.

170

c. [DP [PP/DP (of) women1-GEN] [NP [PP/DP (of) women1] [NP two X1]] D°+AGR]

Let’ s now return to the sentence in (39)a repeated as (46). Note that the structure of the direct object is actually polymorphemic, consisting of one plus possessive Agreement plus the accusative morpheme. I propose that this structure is identical to the partitives we saw above, specifically the derivation in (45). The derivation in (47) shows the structure of bir-i-ni ‘someone’ . The superset denoted by (of) them+human, is actually a null pronoun, or more specifically a D° with -features, as in (47)d. The NP one X picks up its identity from whatever the identity of the superset is, but it does not itself need to bear the same features. Thus one X can remain an NP. In this way, we can explain the non-specific interpretation of someone+case; the larger partitive construction is a DP which gets case, but the NP bir is non-specific. Thus, in (46), the direct object is a partitive with the non-specific (NP) lexical item one embedded in a DP shell that serves as its scope. Non-specificity obtains because someone, i.e. [NP one X] remains in the restrictor. Of course, the partitive, being a DP, requires case; (39)b repeated as (48), is therefore a Case Filter violation. 46) pro bir-i-ni unut-tu-n. pro one-3s-ACC forget-PST-2s ‘You forgot someone’ a. [PP/DP (of) them+human1] + [NP one X1] b. [NP [PP/DP (of) them+human1] [NP one X1]] + D° c. [DP [PP/DP (of) them+human1-GEN] [NP [ t ] [NP one X1]] D°+AGR] d. [DP [PP/DP D°= =null pronoun X1+human-GEN] [NP [ t ] [NP one X1]] D°+AGR]

47)

171

48)

*pro bir-i unut-tu-n. pro someone-3s forget-PST-2s

To be sure we are on the right track, let’ s look closer at the Turkish “ one” bir. Perlmutter (1969) demonstrates that the indefinite article in English behaves like the numeral one. Yükseker (2003) argues that bir ‘one’ in fact behaves differently from the other numerals. First, as demonstrated in (49), numerals (other than one) can alternate in word order with adjectives with no semantic reflex. On the other hand, bir denotes numericity when separated from the nominal it modifies, as in (50). 49) a. iyi yeni iki kitap good new two book ‘two good new books’ b. iki iyi yeni kitap two good new book ‘two good new books’ c. iyi iki yeni kitap good two new book ‘two good new books’ 50) a. iyi yeni bir kitap good new one book ‘a/*one good new book’ b. bir iyi yeni kitap one good new book ‘*a/one good new book’ c. iyi bir yeni kitap good one new book ‘*a/one good new book’

172

Second, numerals can co-occur with a demonstrative as long as the nominal is casemarked, as in (51)153, whereas bir is incompatible with a demonstrative as shown by (52). In fact, bir seems to be incompatible with specificity, as shown by the unacceptability of the accusative case on (53)b. Example (53)c demonstrates an exception where bir is focused and pronounced with stress and can appear with an overtly accusative, i.e. specific, object.154 51) a. Patricia bu kitab-ı oku-du Patricia this book-ACC read-PST b. *... bu kitap ... this book c. ... iki kitap ... two book

155 d. *... iki kitab-ı ... two book-ACC

e.

... bu iki kitab-ı ... this two book-ACC

52)

a. *Patricia bu bir kitab-ı oku-du Patricia this one book-ACC read-pst b. *... bu bir kitap ... this one book

53)

a. Patricia bir kitap oku-du Patricia one book read-PST ‘Patricia read a book’ b. *... bir kitab-ı ... one book-ACC

153

Numerals seem to be [–specific] because they lead to unacceptability when modifying a casemarked nominal without an overt demonstrative element, as in example (51)d. 154 This issue of focus will be addressed later as part of a larger discussion. 155 This example is acceptable when in response to a D-linked set, but this is orthogonal at this point.

173

c. Exception: Patricia her gün BIR gazetey-i oku-r Patricia every day ONE newspaper-ACC read-AOR ‘Patricia reads one particular newspaper every day’ Finally, whereas numerals exhibit adjectival properties as part of a larger DP, (54)a, bir ‘one’ is predicative; it does whatever the indefinite article “ a” or the indefinite quantifier “ some” do in English.

54)

a. [DP bu iki yeni kitap] this two new book b. [DP bu(*-nlar) iki yeni kitap(*-lar)] this(*-pl) two new book(*-pl) c. *[DP bu bir yeni kitap] this one new book d. *[TP [DP bu] [VP iki yeni kitap] ] this two new book e. [TP [DP bun-lar] [VP iki yeni kitap] ] this-pl two new book ‘These are two new books’ f. [TP [DP bu] [VP bir yeni kitap] ] this one new book ‘This is one new book’ g. [TP [DP bu] [VP yeni bir kitap] ] this new one book ‘This is a new book’

For expository purposes I am going to refer to bir denoting non-specificity (translated as the indefinite article in English in the glosses above) in lower case letters and the numeral BIR ‘one’ in upper case letters. I conclude from the data that once bir raises from a position as sister to N°, it is interpreted as the numeral BIR. To be more

174

and 2) it induces minimality effects and blocks raising from its c-commanding domain. to build the structure of the DP subject with the non-specific 156 I adopt the view that phrasal labels are for expository/mnemonic purposes only and that the syntax only cares about categories as identified by their features or properties. it is the head of a phrase. One° causes syntactic (non-)specificity effects which will be interpreted in the semantic component. 55) a. In this way. A pronoun is a D° and is composed of -features. I assume that bir. in fact. we might say that (55)a is analogous to [one X1] [book1] whereas (55)b is simply [one [new [book]]]. Thus. OneP. cannot raise at all. bir ‘a new book’ OneP new One° bir N° book b. Repeating (41) as (56)a. and is adjectival. as far as the syntax is concerned OneP is an NP. I also assume that One° is the antithesis of a pronoun. One° is devoid of features. let’ s use our analysis of the possessive partitive we saw in (47).156 This contrasts with the numeral BIR that adjoins to NP. and is the locus of specificity. The difference is shown in the diagrams in (55). 175 . and serves as a “ drag” on nouns: by this I mean that 1) it is an NP and cannot satisfy the EPP (OneP and its complement N cannot raise). D-features raise in the syntax. which is syntactically an NP.precise. and thus non-specific. the numeral BIR ‘one new book’ NP BIR NP new N° book Pursuing our analysis of partitives above.

or an inherently case-marked expression may move there to satisfy the EPP. and (59)b for (58) where the 176 . First. the subject must be obligatorily case-marked. 56) a. as in (57). I assume the subject bir-i is a partitive structure where the superset is a pro with the meaning akin to ‘of the group’ . As we saw. as in (56)b. Bir-i-Ø gel-di. Susan one-3s-GEN come-COMP-3s-ACC hear-EVID ‘Susan heard that someone had come’ In (56)a. onlar-dan. partitives can be denoted in Turkish by either a nominal expression with ablative case. to be assigned case.interpretation ‘someone’ . 57) on-lar-dan bir-i (dir) 3-pl-ABL one-3s (copula) ‘It/He is one of them’ on-lar-ı bir-i (dir) n 3-pl-GEN one-3s (copula) ‘It/He is one of them’ 58) I assume that D° is a functional head that assigns case and also has an EPP feature. onlar-ı as in (58). This is shown in (59)a for (57) where the ablative moves to [Spec. one-3s-NOM come-PST ‘Someone came’ b. DP] to check the EPP feature of the (null pronoun) D°. note that the subject biri is indeed a DP with null nominative case because when we embed the sentence. Susan [bir-i*(-nin) gel-di -i]-ni duy-mu . As with other case assigning heads a non-case-marked DP may move to its specifier. n. or a possessive with genitive case.

the effect is such that the superset group identifies the scope and the OneP delineates the restrictor. In (56)a. The superset in (56)a could also be pronounced as in (60).3rd person plural pronoun them without case raises to [Spec. Because bir is in the restrictor. the superset DP “ (of) them” remains unpronounced. the movement triggers agreement on bir. but would entail a D-linked group. In both instances. DP] and receives genitive case from D°. DP them+GEN OneP DP them One° bir D° = Ø -pronoun In both these structures. DP them+ABL OneP DP One° them+ABL bir D° = Ø -pronoun b. 59) a. 3-pl-ABL/GEN one-3s-NOM come-PST ‘One of them came’ 177 . 60) On-lar-dan/ı bir-i-Ø n gel-di. with subject bir-i. it receives a non-specific reading. a reflection of the “ Avoid Pronoun Principle” which operates quite systematically in Turkish.

DP] to be assigned genitive case. Furthermore.Thus. (i) çocuk-lar-dan iki kı z child-pl-ABL two girl ‘two girls of (the group of) children’ (ii) çocuk-lar-ı iki kı n z*(-ı ) child-pl-GEN two girl-3s ‘the children’ s two girls (as in daughters)’ 158 See fn.157 We had originally said that the expression denoting the superset could be first-merged as the complement of the subset (in this case One°) or as the Spec of One°. the pro denoting the superset “ of them” merges into [Spec. 61) pro(on-lar-ı bir-i n) DP pro-GEN OneP pro(3pl) ( =“(of) them”) D° = Ø -pronoun One° +Agr bir -i 157 The genitive possessor requires agreement and the superset-subset relationship must be one that allows normal “ possessivity” . I will assume the possessive structure for pro + bir because there is evidence that this partitive structure requires the subset to be of the same type.158 I revise this view somewhat and suggest that in the ablative construction. as shown in (61). the superset first merges as the specifier of the subset-NP. OneP]. I assume the partitive construction has the genitive possessor as the superset. after which OneP merges with D°. I assume the superset “ (of) them” of (56)a is pro. 156. Because [bir X] must pick its identity from the pro superset. Thus. the superset ablative adjoins to the subsetNP. but (ii) indicates that the genitive possessor induces familial identity. and in the possessor partitive. triggering possessive agreement on One. 178 . and “ of them” -pro raises to [Spec. As shown in (i). the ablative can denote a group to which one belongs.

The result is that by being in the restrictor of this construction. so how was the non-specific reading obtained? The polymorphemic structure of bir-i indicated that it was part of a partitive structure. the requirement that +human nominals be DPs is satisfied. biri. +human subjects triggered unacceptability in relative clauses that were otherwise acceptable. However. while its superset possessor defines the scope. *Adam gel-en gün. the ungrammaticality of this example is predictable because the subject is +human and must move to [Spec. At the same time. this study serves to identify another variable that needs to be considered when evaluating grammaticality. In addition to the theoretical implications of human-non-human distinctions in specific grammars and in Universal Grammar. By proposing an account where the bir remains in the restrictor of a complex DP. the contradiction between overt case and non-specificity was resolved. ‘someone’ . and saw that the grammar required it to be obligatorily case-marked.64) Based on what we saw in this chapter. 5 Summary We began with a problem: in Turkish. examples with non- 179 . Kennelly (1997) states that “ Time expressions do not function as a ‘Locative’ argument in terms of relativizing using the [SR] strategy. Our analysis showed that +human nominals could not be NPs. We saw evidence from other data that supported this conclusion. TP] for case. For example. bir remains nonspecific. We had determined that overt case denotes specificity in Turkish.’ ” (p. ‘The day when a man came. on a pragmatic level. in general. We then looked at the non-specific +human QP.

the NSR form is acceptable for both these examples with the condition that the subject be specific. 62) a.over-NSR peasant ‘the peasant whose cow *a/the driver ran over’ 180 . the scrambled position above the subject in [Spec. TP] is also available for the accusative object. but note that the casemarking and interpretation (modulo the discourse effect of the scrambling) is identical. a. [bomba patlay-an] gün bomb explode day ‘the day a bomb exploded’ c.human subjects are perfectly grammatical with the same time expression as the relative head. Of course. [çöp al-ı n-an] gün trash take-PASS-SR day ‘the day the trash is taken’ 6 Contrastive Focus and Human Subjects We saw that extraction from possessive accusative objects using the SR form is unacceptable when the RC subject is human. Compare (63) with (64).159 159 As we have seen in other cases.over-NSR peasant ‘the peasant whose cow *a/the train ran over’ b. as shown in (62). tren ‘train’ and öför ‘driver’ must be overtly case-marked with genitive case and must receive a specific interpretation. [[Ø ine -i]-ni öför-ün ez-di -i] köylu cow-AGR-ACC driver-GEN run. [ya mur/kar ya -an] günler rain/snow rain-SR days ‘the days it rained/snowed b. i. [[Ø ine -i]-ni tren-in ez-di -i] köylu cow-AGR-ACC train-GEN run. as shown in the examples in (65). where the clausal subjects. Thus the examples in (i) are also acceptable.

But this seems to be exactly what is happening in example (66).over-NSR peasant ‘the peasant whose cow *a/the train ran over’ b. [ öför-ün [Ø ine -i]-ni ez-di -i] köylu driver-GEN cow-AGR-ACC run. A +human subject cannot remain in-situ.into-SR man ‘the man whose ship a captain ran into’ 160 I sever (2003) shows that focus and contrastively focus are distinct phenomena in Turkish that exhibit different properties. as well as example (67) with a dative object. we determined that a +human nominal must be a DP and must therefore always raise and be overtly case-marked.over-SR peasant ‘the peasant whose cow THIS DRIVER (rather than that one) ran over’ In the previous section.160 66) [[Ø ine -i]-ni BU öför ez-en] köylu cow-AGR-ACC THIS DRIVER run. and example (68). 181 .over-NSR peasant ‘the peasant whose cow *a/the driver ran over’ But note that the unacceptable (64) becomes acceptable when the +human subject is contrastively focused. [tren-in [Ø ine -i]-ni ez-di -i] köylu train-GEN cow-AGR-ACC run. *[[Ø gemi-si]-ne kaptan çarp-an] adam ship-AGR-DAT captain run.63) [[Ø ine -i]-ni tren ez-en] köylu cow-AGR-ACC train run.over-SR peasant ‘the peasant whose cow a driver ran over’ 64) 65) a. 67) a. as in (66). The SR form with a +human subject is unacceptable unless the subject is contrastively focused.over-SR peasant ‘the peasant whose cow a train ran over’ *[[Ø ine -i]-ni öför ez-en] köylu cow-AGR-ACC driver run.

We had determined that human subjects cannot remain low in the structure. In the good SR examples (66). [[Ø gemi-si]-ne BU kaptan çarp-an] adam ship-AGR-DAT THIS captain run. they must always raise to [Spec. *[[Ø 1 ev-i-ni i çi/askerler yı k-an] aile1 house-AGR-ACC worker/soldiers raze-SR family ‘the family whose house (a) worker(s)/soldiers razed’ b.into-NSR-3s man ‘the man whose ship THIS CAPTAIN (rather than that one) ran into’ [[Ø 1 ev-i-ni BU i çi-nin/askerler-in yı k-tı -ı ] aile1 house-AGR-ACC THIS worker-GEN/soldiers-GEN raze-NSR-3s family 70) 71) ‘the family whose house THIS WORKER/SOLDIERS (rather than that/those) razed’ Returning to the SR examples. we know that the SR form is licensed when [Spec. TP] is available for the +Wh-expression. TP]. TP] will always be occupied by the non-Wh human subject. [[Ø 1 ev-i-ni BU i çi/askerler yı k-an] aile1 house-AGR-ACC THIS worker/soldiers raze-SR family ‘the family whose house THIS WORKER/SOLDIERS (rather than that/those) razed’ We know that the NSR form is acceptable (in fact. so non-subject SR clauses are expected to be bad because [Spec.b.into-SR man ‘the man whose ship THIS CAPTAIN (rather than that one) ran into’ 68) a. 69) [[Ø 1 ine -i]-ni BU öför-ün ez-di -i] köylu1 cow-AGR-ACC THIS driver-GEN run.e. as in (69) through (71) below. and contrastive focus is also possible on the NSR clauses. i. the 182 . required) for non-subject extraction with human subjects.over-NSR-3s peasant ‘the peasant whose cow THIS DRIVER (rather than that one) ran over’ [[Ø 1 gemi-si]-ne BU kaptan-ı çarp-dı -ı n ] adam1 ship-AGR-DAT THIS captain-GEN run. (67)b and 68)b.

I assume that case is. All case-marking has been quite predictable and falls within standard theoretical assumptions.the subject is case-marked by T under long distance agree and the contrastive focus takes care of the movement at LF. TP]? There are three possible answers: 1. Let’ s look at these one at a time. In fact. The problem with the first option is that so far we have not assumed Long Distance Case Assignment. the human subjects in the examples do not bear case morphemes. if anything. The specifier of T was available only when the subject was an NP that did not require case. Allowing human subjects to have no case would completely negate what we have had to assume based 183 . 3. 2. How is it that contrastive focus on a human subject allows it to remain out of [Spec. a PF phenomenon. In Turkish. it would be odd to say that that for this circumstance only. The second option is also unappealing because we are coming up with exceptions to what has been a very consistent state of affairs. i. case is assigned via Agree and that PF does not require the case morpheme to have phonetic content. We have seen no evidence to presuppose the existence of a default case. absence of case on DPs leads to a PF crash.e.Contrastive Focus allows default case on the subject or perhaps no case.Case is a PF phenomenon for visibility. expressions that are scrambled or sluiced keep their case morphology. much of the evidence pointed to case assignment only in a [Spec-Head] syntactic configuration. Again. TP]. Furthermore.relativized expression must have moved to [Spec.

By assuming this. The intonation provides the visibility for the phrase at PF and the focus features drive the raising of the expression at LF. In this way. human nominals must always be case-marked and can never remain in-situ. whereas non-human nominals can be bare and remain in-situ. What do we know about human nominals and DPs in general? We know that they must raise in the structure and be overtly case-marked. I assume that. just as with Wh-movement for interrogatives. We have just seen evidence that. The implication is that other than the EPP which is a requirement in narrow syntax. We know that in these examples. If we assume that casemarking is a PF phenomenon required for visibility/interpretability. I had assumed that specifics “ had to” raise to give us the Diesing-style mapping.on the behavior specific to human nominals. to have case and to raise. we can account for the acceptability of the SR form for contrastively focused human subjects. TP] had to be vacant to license the SR form. That is. in fact. The issue seemed to be that the grammar requires specifics. But it is not clear that the focused expression must raise at 184 . both interface requirements of a DP are met. then we can argue that contrastive focus also gives us a PF reflex. perhaps at LF. This marked pronunciation is obviously visible at PF. in general. contrastive focus is a feature that must be checked in the C domain. in particular. raising and case are LF and PF interface requirement respectively. and +human subjects. there is no raising of the focused expression in the syntax because [Spec. movement for contrastive focus occurs in covert syntax. Turkish is a Wh-in-situ language. There is also a slightly different way to look at the phenomena here. In addition. contrastively focused elements must receive intonational stress.

More interestingly. At the same time. without raising.LF. this being the flip-side of the second structural case marking in Korean which is presumably interpreted as focus marking at PF (and LF). the contrastive focus intonation on the subject does the work of structural case-marking at PF. there is also a view from Korean that focus and structural case may be doing the same work. Let’ s assume that this prohibition holds for Turkish. A crucial requirement is satisfaction of the Case Filter. This would follow nicely with our expectation that there would be a difference in the scope of a contrastively focused SR vs. NSR clause. even at LF. 185 . then the “ Case” requirement would be met. it could also be possible to interpret something as specific. we would have to assume that the focus marking in the NSR examples with genitive subject is not on the subject (structural case and focus being incompatible) but rather on the T projection. or to be more precise. It is also possible that the interpretation of specificity is not because of heirarchy per se but simply a coincidental by-product of DPs— which are interpreted as specific at LF— having to raise. That would mean that in the SR examples. all other requirements being met. Now as to the PF Case requirement. Schutze (2001) and Hong (2002) present evidence from Korean case-stacking phenomena that the structural case marking on top of inherent case-marked expressions is actually a focus marker. Under this story. their data shows that focus and structural case marking are incompatible. if it did at the PF interface whatever case does. Now if contrastive focus were also a case. And the specific interpretation would follow from the features on D°.

This contrasts with the NSR examples where only the Quantifier/Demonstrative of the subject raises at LF. the prediction is that there will be a difference in intonation patterns between these two types of clauses. and also to point out topics of interest. but really they are orthogonal to this thesis. In my account. the entire DP is focused and raises at LF. I assumed that in the SR examples. It is difficult to tease apart these two or indeed to develop diagnostics to tease apart the different interpretations between these structures. but it may be worthwhile. to make a prediction. One further issue remains. at least. I have made predictions. the entire subject phrase will be stressed whereas in the NSR version only the demonstrative will receive stress. How do we explain the contrastively focused NSR examples (69) through (71)? Herburger (2000) notes that (contrastive) focus can target as little as a quantifier or as much as an entire phrase.I bring up these issues both to unpack what is fundamental to the spirit of this research project and what is peripheral. I do want to add. and it does not matter here what explanation one may wish to adopt. though. In the SR version. that in addition to the scopal differences. 186 . I have avoided a detailed discussion of the different interpretation we expect in these two constructions. Again. but the question is tangential to this work.

which canonically requires the Non-subject Relative (NSR) form. is possible with the Subject Relative (SR) form in unaccusatives. *at-lar ko -an saha horse-pl run-SR field Intended: ‘the field where horses run’ 161 As noted in Chapters 2 and 3. [Ø i su ak-an] dami water pour-SR roof ‘the roof water pours/drips from’ b. as in (1) but not in unergatives. [su-yun ak-tı -ı ] dam water-GEN pour-NSR-3s roof ‘the roof that the water is dripping/pouring from’ 2) a. and thus an NP in this account. Dam-dan su ak-ıyor. We saw for example that relativization of a non-subject. but not in unaccusatives.161 We concluded this was because the NP subject intervened between the Wh-expression and Tº in unergatives.-3s ‘Water is pouring/dripping from the roof’ c. the subject must be non-specific. 187 . roof-ABL watership pour-pres. as shown in the tree in (3) for (1). 1) a. and in (4) for (2).Chapter 6: Relativization in Psych Verb Constructions 1 Background The central claim in this thesis is that the properties observed in Turkish relative constructions can be explained by A-movement and minimality.prog. as in (2).

b. we have seen the SR form is licensed in another derivation with scrambling. the SR RC is acceptable. thereby circumventing the intervention effects of the subject. the complex locative DP roughly equivalent to ‘the field’ s inside’ first scrambles (and adjoins) to a projection below TP in move 188 . When the Wh-expression is embedded in a larger DP which scrambles around the subject. In the tree in (5). -Case] C° vP T° [+EPP] However. at-lar*(-ı n) ko -tu -u saha horse-pl-GEN run-NSR-3s field ‘the field where the horses run’ 3) [Ø i su ak-an] dami water pour-SR roof ‘the roof water pours/drips from’ CP roof +Wh t-roof +Wh TP C° VP T° [+EPP] V° pours roof-ABL +Wh NP water 4) *[at(-lar) ko -an] saha horse-pl run-SR field CP TP field +Wh NP-horses VP field-LOC +Wh V° run v° [-EPP.

Second. It is only when a +Wh expression moves to [Spec. so called psych verbs. any time the clausal subject is +human. to even get the analysis off the ground. let’ s refresh our memory as to potential confounding factors. 5) [[ Ø 2 iç-in-de]1 at t1 ko -an] saha2 inside-AGR-LOC horse run-SR field CP TP field +Wh DP-LOC C° vP vP T° [+EPP] NP-horses VP DP-LOC V° run v° [-EPP. human nominals cannot be NPs. we must control for the human features of the subject. TP] that the SR form is triggered. Thus. First. This means that for non-subject extraction from psych verbs using the SR form. after which the +Wh genitive expression ‘field’ in its Spec raises to [Spec. Thus. for non-subject relatives. -Case] [[field+Wh]-GEN inside] There is a class of verbs that does not seem to be as well-behaved in this respect: the class of predicates that denote psychological states. This derivation licenses the SR form. Before we begin.. the subject can never be +human. the SR form is licensed for non-subjects when the RC subject is an NP (non-specific). the SR form will be barred because the DP subject must move to [Spec. 189 . TP] and checks T’ s EPP feature. TP] for case.

the Experiencer is the Accusative object or Dative object respectively. that is. that is the play that didn’ t please Mary might be an excellent play. A preliminary evaluation of Turkish psych verbs does not yield such a tidy division. Compare (i) and (ii) with (6) and (7). whereas the Group 1 Subject Experiencer zlverb kı ‘to get angry (at)’ permits both Target and Subject Matter as Themes. Subject Experiencer that takes a PP Theme (marvel). but sı ‘bore/frustrate’ requires the Causer to be the Subject Matter. the verb bayı ‘love/get a kick out of’ seems to require a Target Theme. as the semantic denotation of the Theme seems to be rather arbitrary. Object Experiencer (amuse). (i) The play didn’ t appeal to Mary. while nefretet. This suggests that the Causative is introduced in the vP. Thus. or structurally casemarked. In Group 1 and 2. (ii) The play didn’ t please Mary. respectively. Thus. a la Pesetsky (1995).e. In Group 3. the verb kork. where the impression is that Mary gave the play a bad evaluation. In Groups 4 and 5. but it was written by her rival. 164 Pesetsky (1995) notes that the Theme of psych verbs can be divided into two semantic types: the Target of emotion and the Subject Matter of the Emotion. it seems that languages allow psych verbs to range over all verb classes. Likewise with the Object Experiencer verbs which differ as to the interpretations they permit for the Causer.‘fear’ permits both Target and Subject Matter Themes. I mention these to point out that there seems to be no regularity in interpretation that can be deduced from the various types of verbs or the Case that the arguments bear. those that take PP complements) become transitivized (assign accusative case to the direct object) with the addition of a Causative morpheme.164 162 Turkish allows a Causative morpheme in many of these psych verbs. The Subject Matter Theme is shown in (ii) where “ play” only participates in a linking to the Experiencer.162 These are shown below. the subject is the Experiencer with an inherently case-marked Theme. The Group 4 verb rezil etk‘disgrace’ allows both semantic roles for the Causer. (i) Bu ben-i aı r-t-tı . without the Causative morpheme.With this much introduction. we have an Experiencer subject and an accusative. The Target of emotion object is evaluated by the Experiencer. Theme. Similarly.1 Classes of Turkish psych verbs Turkish psych verbs fall into several types. in Group 2. 163 It may be more than coincidental that Levin (1993) identifies four classes of psych verbs in English: Subject Experiencer that takes an object Theme (admire). this me-ACC surprise-CAUS-PST ‘This surprised me’ (ii) O ben-i kı r-dı z-dı S/He/that me-ACC anger-CAUS-PST ‘S/He/that made me mad/angered me’ Many of the intransitive psych verbs (i. and Object of preposition Experiencer (appeal). either Target or Subject Matter. 1.‘despise’ imposes a Target interpretation on the Theme. let’ s now look at the behavior of these verbs with respect to relativization. as in (i). These can be viewed as analogous to the kinds listed for Turkish.163. The discussion in this Chapter is limited to psych verbs in what I assume is the base form. 190 .

surprise to this’ ) Ben on-a kı m z-dı I him/that-DAT got. r-dı I this-DAT surprised ‘This surprised me’ (Literally: ‘I felt. pta I you-DAT envy do-PRES ‘I envy/emulate you’ Ben san-a bayı l-ıyorum I you-DAT faint-PRES ‘I love you’ (Literally: ‘I swoon over you’ ) Ben o-na güven-irim I that-DAT trust-PRES ‘I trust him’ Ablative Theme 7) 8) 9) 10) Group 2: Subject Experiencer 11) Ben on-dan kork-tum I that-ABL fear-PST ‘I got scared by that’ Ben bun-dan zevk al-dı m I this-ABL got-pleasure ‘I enjoyed this’ 12) 13) Ben sen-den nefret ed-iyorum I you-ABL hatred do-PRES ‘I despise you’ 191 .angry ‘I got mad at him’ Ben san-a gı ed-iyorum.Group 1: Subject Experiencer 6) Dative Theme Ben bun-a a ı m.

14) Ben sen-den bı m k-tı I you-ABL fed.up ‘I am fed up with you’ Accusative Theme Group 3: Subject Experiencer 15) Ben sen-i sev-iyorum I you-ACC love-PRES ‘I love you’ Ben sen-i özlü-yorum I you-ACC miss-PRES ‘I miss you’ 16) 17) Ben o-nu / sen-i arzulu-yorum I s/he/it-ACC/you-ACC desire-PRES ‘I desire s/he/it/you’ çocuklar-ı köpek özle-yen adam -nı I you-ACC miss-PRES ‘I miss you’ 18) Group 4: Accusative Object Experiencer 19) O ben-i etkile-di It me-ACC affected ‘It affected me’ O ben-i rezil et-ti He/It me-ACC disgrace do-PST ‘He/It disgraced me.’ O ben-i sı yor kı He/It me-ACC bores ‘He/It bores/frustrates me’ 20) 21) 192 .

the SR is acceptable only for subject extraction and the NSR is required when relativizing non-subjects.22) O ben-i boz-du He me-ACC humiliated ‘He humiliated me’ Group 5: Dative Object Experiencer 23) O ban-a dokun-uyor s/he/that me-DAT upset-PRES ‘S/he/That upsets me’ O ban-a tuhaf gel-di that me-DAT odd come-PST ‘That seemed strange/odd to me’ O ban-a eziyet et-ti s/he/that me-DAT bother/disturb do-PST ‘S/He/That bothered/disturbed me’ O ban-a malum ol-du that me-DAT obvious be-PST ‘It/That became known to me’ (as in ‘I found out’ ) 24) 25) 26) 2 Turkish Psych Verbs and Relativization All psych verbs behave as expected with respect to canonical relativization. As the examples in (27) demonstrate. 27) a. [haber-e ası r-an] adam news-DAT surprise-SR man ‘the man who felt surprised by the news’ 193 .

at first glance. the example in (28) is marginal.1 Experiencer subjects The complication with these Subject Experiencer verbs is that. where the DP subject is case-marked. 28) ??at tren sesi-ne aı r-dı horse train noise-DAT felt-suprised Intended: ‘The horse was surprised by the noise of the train’ köpek/??at e etimci-nin yeni düdük çal-ma-sı -na aı r-dı dog/horse trainer-GEN new whistle blow-INF-3POSS-DAT felt-surprised ‘The dog/??horse was surprised by the trainer(’ s) blowing a new whistle’ 29) 194 . that is. [adam-ı a ı n r-dı -ı ] haber man-GEN surprise-NSR-3s news ‘the news that the man felt surprised by’ d. that the subject ‘dog’ is being endowed with human characteristics. Thus. If the Experiencer subject is being encoded as “ human-like” . or is (30)a bad because nonsubject extraction using the SR form is not possible with human subjects. they seem to require human subjects. And the example in (29) can be viewed as an instance of coercion of sorts. and here ‘dog’ is behaving as if it were syntactically +human. *[haber-e ası r-dı -ı adam ] news-DAT surprise-NSR man c. it would account for the RC in (30)a. where the SR form with an NP subject is unacceptable. Is this because there is something peculiar about the structure of psych verbs.b. Note that example (31) with human Experiencer as the subject is perfect. *[adam(-ı n) a ı r-an] haber man(-GEN) surprise-SR news 2. Contrast this with the corresponding but good NSR clause in (30)b.

This would be analogous to the alternation we see in English between the +human pronouns him/her and the –human it when referring to animals. that horse lovers would find the marginal (28) and (29) with ‘horse’ acceptable. then. But recall that we had determined that this was a selectional requirement imposed by the lexicon. is that psych verbs require the Experiencer subject to have a D-feature (which is the same as -features). Let’ s therefore keep our analysis at the syntactic level and just assume at this point that Subject Experiencer Psych verbs seem to require DP subjects. People who love dogs refer to their pets as “ he” or “ she” while a nondog-lover may use “ it” to refer to the same animal. and that there is some leeway in the acceptability of this feature on animals which is posited idiosyncratically. then. It is no surprise.surprised-NSR-3S trainer b. [[Ø 1 yeni düdük çal-ma-sı -na] köpe -in a ı r-dı -ı ] e etimci1 ‘the trainer who the dog felt-surprised by (his) blowing a new whistle’ 31) ö renci müdür-ün yeni düdük çal-ma-sı -na aı r-dı student principal-GEN new whistle blow-INF-3POSS-DAT felt-surprised ‘The student was surprised by the principal(’ s) blowing a new whistle’ Surely. prior to evaluation or interpretation by the semantic component. 195 . we are looking at syntactic reflexes here. -features are required for “ the quality of being human” .surprised-SR trainer Intended: the trainer who dogs felt-surprised by (his) blowing a new whistle’ new whistle blow-INF-3POSS-DAT dog-GEN feel. it cannot be possible for a language to encode dogs as having “ human-like” qualities and horses not. This will account for the facts so far. We had determined that -features rest on D. The proposal. We can assume that this quality can be extended idiosyncratically to animals depending on the individual user. Furthermore. In Turkish.30) a *[[ Ø 1 yeni düdük çal-ma-sı -na] köpek a ı r-an] e etimci1 new whistle blow-INF-3POSS-DAT dog feel.

Notice now the acceptable SR example in (35) where the verb has been passivized. 32) a. TP] unavailable for the +Wh-expression. 196 . making [Spec. With the Experiencer former-subject out of the way in an adjunct clause. .do-SR country Look what happens when the +Wh-expression is embedded. the +human subject must raise to [Spec. the verb gı ‘to envy/covet/emulate’ . TP] for case.do-NSR-3s country ‘the country where the peasants envy educated women’ b. the parallel SR form in (33)b is also acceptable. Not surprisingly from what we saw in previous chapters regarding extraction from complex subjects. the relative head ‘villagers’ has been extracted from the complex DP-subject ‘villagers’ daughters’ .do-SR country Intended: ‘the country where peasants envy educated women’ c. the subject ‘villagers’ daughters’ may not remain in situ or without case.Let’ s look at another example. (32)a. pta Relativizing a locative using the NSR using gı is fine. *[köylü(-ler) e itimli kadı nlar-a gı ed-en] ülke pta villager(s) educated women-DAT envy. TP] is now 165 Example (34)c is an attempt to avoid intervention effects by scrambling the dative object around the subject. pta relativizing a non-subject using the SR form is bad in (32)b-c because the clausal subject ‘villagers’ is +human and must obligatorily raise to [Spec. As demonstrated in (34)b-c. *[e itimli kadı nlar-a köylü(-ler) gı ed-en] ülke pta educated women-DAT villager(s) envy. In the NSR example in (33)a. note that the NSR is acceptable (34)a. Here the Experiencer has been demoted. When extracting from within the Dative object though.165 Again. whereas the parallel SR form is unacceptable. As expected. TP] for case. [köylüler-in e itimli kadı nlar-a gı et-ti -i] ülke pta villagers-GEN educated women-DAT envy. [Spec. it is now the complement of an adjunct ‘by-phrase’ with ablative case.

available for the +Wh specifier of the Dative DP. [[Ø 1 kı zlar-ı n] n-ı e itimli kadı nlar-a gı et-ti -i] köylüler1 pta girls-AGR-GEN educated women-DAT envy. 33) a.do-SR villagers Intended: ‘the villagers1 who [such that] educated women envy (their1) daughters’ c. [[Ø 1 kı zlar-ı e itimli kadı ] nlar-a gı ed-en] köylüler1 pta girls-AGR educated women-DAT envy. extraction of a non-subject is fine using the NSR form (36)a and bad using the SR form (36)b. Let’ s take a look at another verb that in principle should permit a non-human subject. As expected. *[[ Ø 1 kı zlar-ı n-a] e itimli kadı n(-lar) gı ed-en] köylüler1 pta daughters-AGR-DAT educated woman(-pl) envy. let’ s look at the behavior of this verb with a human subject as in (36). a dog “ envying” the food or the collar of another dog is a little weird. It seems though that the same facts hold for the Subject Experiencer verb güven ‘trust’ .do-SR villagers ‘the villagers1 whose [such that (their1)] daughters envy educated women’ 34) a. and the SR form is again licensed (36)d. a configuration which licenses the SR form. 197 .do-PASS-SR soldiers ‘envy/covet’ to have a non-human subject. First.do-SR villagers 35) [[Ø 1 babalar-ı -na] genç adamlar taraf-ı n-dan gı ed-il-en] pta askerler1 ‘the soldiers1 whose [such that (their1)] fathers are envied by young men’ Let’ s say we accept the lexico-semantic fact that it doesn’ t make sense for gı pta fathers-AGR-DAT young men viewpoint-ABL envy. *[e itimli kadı n(-lar) [Ø 1 kı zlar-ı n-a] gı ed-en] köylüler1 pta educated woman(-pl) daughters-AGR-DAT envy.do-NSR-3s villagers ‘the villagers1 who [such that] educated women envy (their1) daughters’ b. After all.do-NSR-3s villagers ‘the villagers1 whose [such that (their1)] daughters envy educated women’ b. Passivization renders the human subject out of the way. [e itimli kadı nlar-ı [Ø 1 kı n zlar-ı -na] gı et-ti -i] pta köylüler1 educated women-GEN daughters-AGR-DAT envy.

çocuklar [okulun müdür-ü]-ne güvenirler children school-GEN principal-AGR-DAT trust ‘(The) children trust the school’ s principal’ b. This is demonstrated in the tree in (37)b for the unacceptable SR in (37)a. as in (38)a. [[ Ø 1 müdür-ü-ne] güven-il-en] okul1 principal-AGR-DAT trust-PASS-SR school ‘the school whose principal is trusted’ Now let’ s look at this Experiencer subject verb with a non-human subject. the meaning in (i) is technically ‘the person who [of the animals] (some) animals trust’ .166 As demonstrated in the tree in (38)b.AGR] Dº ]-GEN 198 . the SR form is barred when relativizing a non-subject because the subject intervenes. (i) [hayvanlar-ı n güven-di -i] insan -nı animals-AGR-GEN strust-NSR person ‘the person who animals trust’ (ii) [DP [DP animals-GEN] [NP animals-POSS. whereas the Dative Wh-expression can satisfy the EPP. the EPP of T fails to be satisfied. (i). In (37)a.36) a. *[[ Ø 1 müdür-ü-ne] çocuk(-lar) güven-en] okul1 principal-AGR-DAT child(ren) trust-SR school Intended: ‘the school whose principal children trust’ d. I showed that the denotation of non-specificity on a nominal expression where a Dfeature is imposed on it (as in the case of humans. The only way to relativize the Dative Theme of the verb ‘trust’ is with the NSR form with a specific subject. the subject raises to [Spec. TP] satisfying T’ s EPP and is assigned genitive case while the +Wh-Dative long-distance A-bar moves to [Spec. 166 In Chapter 5. As expected. CP]. I assume that the underlying structure of the subject in (i) is as in (ii) where the non-specific (existential) ‘animals’ is in the restrictor of the null DP-animals. psych verb Experiencers) is achieved in Turkish by the use of a partitive construction. [[Ø 1 müdür-ü-ne] çocuklar-ı güven-di -i] okul1 n principal-AGR-DAT children-GEN trust-NSR-3s school ‘the school whose principal the children trust’ c. and as we will see. and. it is blocked from doing so by the intervening subject. The non-specific subject hayvan ‘animal’ cannot satisfy it. Thus.

the DP scrambles around the subject. CP TP vP animal VP person-DAT +Wh V° trust v° C° T° [+EPP] 38) a. CP TP DP-animal-GEN vP DP-animal VP person-DAT +Wh V° trust T° [+EPP] v° C° We saw in (5). an example of an alternative derivation using the SR form: the +Whexpression is a “ free-rider” in a larger DP and literally gets carried around the blocking element. TP]. This alternative 199 . Recall how this worked: the relativized expression is embedded in a DP (as the specifier). The relativized expression is now free to raise to [Spec. and adjoins to a position lower than [Spec. and the SR form is triggered.37) a. TP] because there are no interveners. *[hayvan güven-en] insan animal trust-SR person Intended: ‘the person who animals trust’ b. [hayvan-ı güven-di -i] insan n animal-GEN trust-NSR person ‘the person who the animal trusts’ b.

for example. and the other verbal.angry-NSR people ‘the people whose children the animals trust’ The conclusion that Experiencer psych verbs “ select” for human subjects is too strong as evidenced by the acceptability of (39)b. 200 . this is the flip-side of the requirement that +human nominals have a D feature. that human nominals must merge with a D. in sentence (40) we get the reading there were some children who played in this park. It certainly doesn’ t mean all children in general. that psych verbs select for a DP as Experiencer. played in the park.angry-SR people Intended: ‘the people whose children animals trust/get angry at’ b. 39) a. We must therefore assume that Turkish psych verbs require a subject with a D feature. [[Ø 1 çocuklar-ı -na] hayvanlar-ı güven-di -i/kı n z-dı -ı insanlar1 ] children-AGR-DAT animals-GEN trust-NSR/get. prior to LF. this requirement about Experiencers may be universal. that produce identical intervention effects in the syntax. That is. 167 In essence. As shown in the examples in (39). one nominal. Contrast this with sentence (41) where the reading seems to be that all (contextually relevant) children. in general. 40) 41) Children played in this park Children trusted this policeman.derivation does not seem to be possible with a Subject Experiencer psych verb. *[[Ø 1 çocuklar-ı -na] hayvan(-lar) güven-en/kı z-an] insanlar1 children-AGR-DAT animal(s) trust-SR/get. trusted the policeman. In English. the SR form is still unacceptable. we have two instances of selection. Note that we are making a claim that psych verb selection is evaluated in overt syntax. Experiencers need not be definite but they cannot be existential.167 Although we are looking at Turkish facts here.

(44)c-d-e. as we have seen repeatedly.168 As shown in the table in (42). extraction of a locative is allowed with the NSR. and that semantic notions such these and their counterparts are encoded in the lexicon. This is typical behavior of a clause with a +human subject. Extracting from the Ablative Theme is permitted only with the NSR form. either lacks a referent (=kind-level) or is existential. (44)a-b. absence of a D-feature is interpreted as existential. and that this requirement is satisfied in unique ways in varying grammars. the relative head is the inanimate okul See Enç (1991) and Chapter 3 of this dissertation.169 And. In Turkish. the presence of a D-feature on a nominal has syntactic consequences. Karimi (2003) offers a revised version of Enç (1991) according to which a nonspecific nominal (an NP. In (44). The implication is twofold: that neither of these interpretations are permitted by the selectional requirement of the Experiencer of a psych verb. Not surprisingly. for example. Let’ s look at Subject Experiencer psych verbs that take an Ablative Theme. that this is a lexico-semantic selectional requirement. according to our account here). 169 Chomsky (1999: fn. only an indefinite nonspecific nominal does not require a D°. Extracting from the subject permits both forms. nonreferential nominals such as nonspecifics. quantified and predicate nominals need not be assigned D. 168 201 .10) associates D with referentiality. regardless of the word order.The implication is that psych verbs do not allow Experiencers to be existential. 42) Encoding Definiteness and Specificity in Turkish Definite Specific Requires D° Indefinite Specific Requires D° Indefinite Non-Specific No D° We are at a point where we can predict the behavior of an expression in a RC based on whether it is a DP or NP. but not the SR. as shown in (43)a-b.

up-SR school ‘the school where teachers are fed up with naughty students’ 44) a. as in ‘parents’ from the Theme DP [parents-GEN naughty children-AGR] does not effect the outcome: non-subject extraction using the SR form is impossible with a +human clausal subject.up-NSR parents ‘the parents who the teachers are fed up with their naughty children’ b.up-NSR school ‘the school where the teachers are fed up with naughty students’ b. *[ö retmenler [Ø 1 yaramaz ö renciler-in]-den bı k-an] okul1 teachers naughty students-AGR-ABL fed. [[Ø 1 ö retmenler-i-nin] yaramaz ö renciler-den bı k-tı -ı okul1 ] teachers-AGR-GEN naughty students-ABL fed. [ö retmenler-in [Ø 1 yaramaz çocuklar-ı teachers-GEN naughty children-AGR-ABL fed.up-NSR school ‘the school whose teachers are fed up with naughty students’ b. 43) a.up-NSR school ‘the school where the teachers are fed up with its naughty students’ d. *[ö retmenler yaramaz ö renciler-den bı k-an] okul teachers naughty students-ABL fed. *[ ö retmenler [Ø 1 yaramaz çocuklar-ı n]-dan bı k-an] anne-babalar1 teachers naughty children-AGR-ABL fed. *[[Ø 1 yaramaz ö renciler-in]-den ö retmen(-ler) bı k-an] okul1 naughty students-AGR-ABL teacher(s) fed. [ö retmenler-in yaramaz ö renciler-den bı k-tı -ı ] okul teachers-GEN naughty students-ABL fed.‘school’ which moved out of the larger Theme DP [school’ s naughty students]. I include the example in (45) to demonstrate that making the relative head +human. [[Ø 1 ö retmenler-i] yaramaz ö renciler-den bı k-an] okul1 teachers-AGR naughty students-ABL fed.up-SR school Intended: ‘the school where teachers are fed up with (its) naughty students’ e.up-SR parents ‘the parents who teachers are fed up with their naughty children 202 .up-SR school ‘the school whose teachers are fed up with naughty students’ c.up-SR school n]-dan bı k-tı -ı anne-babalar1 ] 45) a. [ö retmenler-in [Ø 1 yaramaz ö renciler-in]-den bı k-tı -ı ] okul1 teachers-GEN naughty students-AGR-ABL fed.

that it be a DP.house fed. *[ [Ø 1 geç kalmalar-ı n]-dan yayı nevi bı k-an] yazarlar1 tardiness-AGR-ABL publishing. 47) a. Just as we saw with k subjects that were +human (which we determined in the Chapter 5 had to be DPs). [ Ø 1 [yazar-ı n geç kalması n]-dan bı k-an] yayı nevi1 author-GEN tardiness-AGR-ABL fed. extraction of a non-subject using the SR form is unacceptable even when the Theme has been scrambled around the subject. This example is repeated as (47).house/publisher search-SR article ‘the article whose author publishing. We saw in Chapter 5. The RC in (47)a is unacceptable with ‘publisher’ as the subject because ‘publisher’ being +human must enter the derivation as a DP. the Experiencer subject cannot remain in situ without case.up-SR author Intended: ‘the authors whose tardiness publishing houses are fed up with’ Note that there is no requirement on the lexical item ‘publishing house’ itself. No such requirement exists on ‘publishing house’ which as an NP can remain in situ.up’ must require its Experiencer subject to be a DP. an example.up-SR publishing.Furthermore. This is demonstrated in (46)b with the inanimate ‘publishing house’ as the Experiencer subject. just as with Dative Theme psych verbs. where the nominal ‘publisher’ led to a derivational crash but not the expression ‘publishing house’ .houses/*publishers are looking for’ 203 . [[Ø 1 yazar-ı ]-nı yayı nevi /*yayı ncı aray-an] makale1 author-POSS-ACC publishing.house ‘the publishing house which is fed up with the author’ s tardiness’ b. This is evidence that the psych verb bı ‘fed. the Experiencer subject with Ablative Theme behaves the same as RCs with +human subjects. 46) a.

[ [Ø 1 çocuklar-ı ]-nı köpek yala-yan] adam1 children-AGR-ACC dog lick-SR man ‘the man1 who [such that] a dog/dogs licked (his1) children’ b.yayı nevi-nin b. as in . and is assigned accusative case. TP]. is an NP and remains in situ. TP]. The object with the +Wh specifier man-GEN. to behave as a DP.house-GEN search-NSR-3S article /publisher-GEN ‘the article whose author the publishing. in . CP TP vP DP+ACC man dog VP DP [ [ man’s ] children] +Wh+GEN V° trust v° C° T° [+EPP] 204 .1 Subject Experiencers with accusative Theme We already know that relativization of an accusative DP is not possible using the SR form. The SR clause in (48)a is one such example. Accusatives cannot move to [Spec. 2. We saw that the possessor of the accusative can be relativized using the SR form as long as the subject is an NP.house/the publisher is/was looking for’ We are forced to conclude that it is the psych verb that is requiring the same lexical item.1. with the derivation in (48)b. moves around the subject to a higher Spec of vP. [[Ø 1 yazar-ı ]-nı /yayı -nı ncı n ara-dı -ı ] makale1 author-POSS-ACC publishing. dog. 48) a. CP]. ‘publishing house’ . in . The genitive possessor then raises from the Spec of the accusative object to [Spec. and then again to [Spec. The subject.

The example in (49)a is unacceptable. The derivation in (49)b is a Case Filter violation. From what we observed in Subject Experiencers. CP TP vP DP+ACC dog VP DP C° T° [+EPP] v° V° trust [man’s children] In conclusion. This will bleed the SR form. psych verbs with Experiencer subjects do not permit such a derivation.2 Experiencer objects Let’ s now turn to psych verbs with Object Experiencers. as shown by the unacceptability of the sentence in (50) without accusative case on the Experiencer object ‘dog(s)’ . Why does this derivation crash? Because the Experiencer subject cannot be an NP. it must raise for case to [Spec. 49) a. will we find that Object Experiencers are also required to be DPs? The evidence certainly points in this direction. The derivation in (49)b should proceed in an identical manner to the one in (48)b. TP] for case. TP]. 2. 205 . none of the subject Experiencer psych verbs will allow non-subject relativization using the SR form.But. The subject Experiencer has to be a DP and must obligatorily raise to [Spec. *[ [Ø 1 çocuklar-ı ]-nı köpek sev-en] adam1 children-AGR-ACC dog love-SR man Intended: ‘the man1 who [such that] dogs love (his1) children’ b.

This is support for the view that psych verbs do not permit an existential Experiencer. As demonstrated in (51)b.Notice from the English equivalent that there is no sense of ‘dogs’ as being definite. this child dog(-pl)*(-ACC) much bother-AOR ‘This child bothers dogs a lot. ‘parents’ to move to. [ [Ø 1 çocuklar-ı ] -nı felaket haberleri etkiley-en] anne-babalar1 children-AGR-ACC disastrous news affect-SR ] parents ‘the parents whose children disastrous news affected’ And. *Felaket haberleri çocuk etkil-er disastrous news child affect-AOR 206 . is a non-specific NP which can (indeed must) remain without case in situ. [felaket haberleri-nin [Ø 1 çocuklar-ı ] -nı etkile-di -i] anne-babalar1 disastrous news-GEN children-AGR-ACC affect-NSR ] parents ‘the parents whose children the disastrous news affected’ b. extraction from the Experiencer object is possible using the SR form. there is no such categorial requirement on the subject. This is because the subject. (52). it must be a DP and bear accusative case. In this phrase. 50) Bu çocuk köpek(-ler)*(-i) çok sı k-ar. but it seems the syntax requires that the Experiencer object be case-marked.’ (In the sense that ‘dogs feel bothered’ ) While the Experiencer object is required to be a DP. the Experiencer object cannot be non-specific. 52) a. ‘disastrous news’ . a requirement on DPs. 51) a. [Spec. TP] is vacant for the +Wh-expression. at least not syntactically. as we had hypothesized. and the SR form is required.

[sihirbaz hileleri-nin büyüle-di -i] çocuklar magician tricks-GEN fascinate-NSR children ‘the children who are/were fascinated by the magic tricks’ c. TP] triggering the SR form. the NSR form is required. ‘magic tricks’ is an NP.Intended: ‘Disastrous news affects children’ b. CP]. This is why (53)a is bad even though the RC subject. Felaket haberleri çocu -u / çocuk-lar-ı etkil-er disastrous news child-ACC/children-ACC affect-AOR ‘Disastrous news affects the child/the children/children’ Another example of an Accusative Object Experiencer psych verb is büyüle ‘to fascinate’ . ‘children’ A-bar moves to [Spec. In the SR in (53)c. [[ Ø 1 çocuklar-ı ]-nı sihirbaz hileleri büyüley-en] anneler1 children-AGR-ACC magician tricks fascinate-SR mothers ‘the mothers1 who [such that] magic tricks fascinated (their1) children’ Note that relativization of a locative is not possible using the SR form (54)a. Recall that an accusative DP cannot A-move to [Spec. The NP subject remains in situ. and cannot raise to Tº. the relative head is embedded in the accusative object. TP]. 53) a. The tree in (54)c demonstrates that both the subject and the accusative object intervene 207 . the +Whaccusative object. When the subject is a DP. *[sihirbaz hileleri büyüley-en] çocuklar magician tricks fascinate-SR children Intended: ‘the children who are fascinated by magic tricks’ b. as in (53)b. while the possessor of the Accusative Experiencer raises to [Spec.

[çiçekleri-nin] çocuklar-ı büyüle-di -i] bahçe flowers-GEN children-ACC fascinate-NSR garden ‘the garden where the flowers fascinate children’ c. In the SR example in (55). is possible. move a minimality violation. 170 208 . TP]. The accusative object is in [Spec. of a locative to [Spec. but A-bar movement.170 A-movement. We saw in Chapters 1 and 2 that these require the SR form as the +Wh-locative is often the only DP in the clause that can satisfy T’ s EPP feature. vP]. as in acceptability of the NSR in (54)b. and the +Wh-teachers has raised from its Spec to [Spec. CP TP vP children+ACC flowers VP C° T° [+EPP] v° garden-LOC children +Wh V° fascinate This confirms what we saw in previous chapters. TP] is . Except in impersonal passive constructions which have no external argument. . extraction from the accusative object using the SR form is possible when the subject is an NP. hence the 54) a. the subject. ‘Iranian films’ is an NP which remains in situ.between the locative and Tº. *[çiçekler çocuklar-ı büyüle-yen] bahçe flowers children-ACC fascinate-SR garden Intended: ‘the garden where flowers fascinate children’ b.

do-SR mothers ‘the mothers whose sons speakers/sexy women provoke/arouse’ 3 Conclusion Let’ s take stock. Relative clauses with Experiencer subject psych verbs exhibit behavior similar to RCs with human subjects: non-subject extraction using the SR form is not possible.do-SR mothers ‘the mothers whose sons speeches sex films provoke/arouse’ b. TP] for case. Just as in other transitive (i. The 209 . Compare the good (56)a with the unacceptable (56)b with a +human subject. TP] of the relative head obeys minimality. [[Ø 1 o ular-ı ]-nı konu malar/seks filmi tahrik ed-en] anneler1 sons-AGR-ACC speeches/sex film arouse. There is the added complication of the accusative object which is frozen for further A-movement. Psych verbs with Experiencer objects allow non-subject extraction using the SR form as long as the subject is an NP and the movement to [Spec. non-subject relativization of the expression in the Spec of the object licenses the SR form when the clausal subject is an NP. 56) a.55) [[Ø 1 ö renciler-i]-ni Iran filmleri büyüle-yen] ö retmenler1 students-AGR-ACC Iranian films fascinate-SR teachers ‘the teachers whose students Iranian films fascinate’ Recall that the SR form is impossible when the subject is +human. The unacceptability of (56)b is predictable because we have already concluded that +human subjects cannot be NPs that remain in situ. *[[Ø 1 o ular-ı ]-nı konu maci(-lar)/seksi kadı n(-lar) tahrik ed-en] anneler1 sons-AGR-ACC speaker(s)/sexy women arouse. they must raise to [Spec. accusative) constructions.e.

we saw evidence that suggested this was true for Accusative objects as well. We saw identical behavior in extraction from clauses with human subjects.171 First. and we determined this was because +human features required a DP layer. WCO. 171 These are the two syntactic diagnostics used throughout this work. they display a selectional restriction.conclusion of the evidence here is that we have not had to assume any new technology to explain relativization with respect to psych verbs. and second. which can never be licensed when relativizing the Dative Experiencer or a lower expression. Although we looked mainly at Subject Experiencers. Neither test can be used to determine the category of a Dative expression. however. datives enter the derivation already case-marked. Having said this. there is no way of assessing whether it must A-move. TP] (regardless of whether the subject is an NP or DP). Recall the two diagnostics used to test DP-hood: overt case and required raising to a functional case-assigning projection. regardless of its theta-role or initial merge position. and semantic interpretation. that the Experiencer be a DP. or whether it blocks A-movement of lower nominals. the subject is base-generated in a position higher than the Dative. The only “ innovation” in this chapter is that psych verbs seem to require that the Experiencer be non-existential. it is reasonable to think that the requirement that Experiencers be DPs probably holds of Dative Experiencers as well. we saw that a c-commanding NP induced intervention effects even though the NP itself could not satisfy the EPP on T°.172 For Datives. and thus intervenes in the movement of the Dative to [Spec. These tests require the availability of the SR form. This requirement is harder to test on Dative Experiencers. prior to overt syntax. Binding. 210 . Other diagnostics may be available. that is. 172 In Chapter 4.

and grammatical processes only refer to structural information which indirectly reflects information . to use B&R’ s terminology) with these features and that the 211 . They do not permit NP Experiencers... In fact. NPs do not need to raise for case..Obviously. but the fact that we were able to account for diverse behavior within a limited theoretical account is interesting. -hierarchies and the like intervene only once. in the formation of D-structures. We were unable to coerce them to behave as NPs.. From there on. this is not the complete story about psych verbs in Turkish. Substantive distinctions between -roles are irrelevant within formal grammar but play a crucial role at the interface between formal grammar and other cognitive systems. reference to such entities is excluded in formal grammar. We have seen that Turkish psych verbs require that the Experiencer enter the derivation as DPs. It must be that these expressions enter into the derivation (at D-structure. [e]verything is mediated through structure. (Belletti and Rizzi 1988:294-295 [bold font mine]) I have highlighted the last lines in bold because this is exactly what the evidence from Turkish psych verbs seem to be indicating. and we saw that Experiencer subjects and direct objects behave in the syntax as if they were DPs. they contribute to determining the initial syntactic representations (D-structures) through a system of mapping principles projecting -structures onto syntactic structures . The facts here are reminiscent of what Belletti and Rizzi (1988) (B&R) conclude in their study of Italian Psych verbs.

we were able to explain Turkish relatives without resorting to additional projections for psych verbs a la Pesetsky (1994). subject Experiencers always need case. TP] will always be occupied by the Experiencer subject. Although Belletti and Rizzi’ s work was under a different framework and for a different language. failing or crashing. otherwise interesting because of their diversity in assigning Experiencer-Theme roles. This is not to say that the discussion in this chapter is the complete story on the structure of psych verbs. But. based on the requirement imposed on the nominal elements prior to Merge. The syntactic consequence is that the Experiencer of psych verbs is always a DP. Indeed. Just like other examples throughout this thesis. If we extend our theory merely to include the lexical-selectional requirement that psych verbs denote states that only humans can exhibit. Obviously other syntactic (and semantic) properties of Turkish psych verbs need to be studied. To be more precise. 212 . the evidence suggests that the constraint is that psych verbs require the Experiencer to be perhaps sentient (as encoded by -features) and non-existential. Thus. offer nothing exciting in terms of relativization. on first pass.derivation proceeds. TP] obeys minimality. we need not posit any other rules to explain the behavior of relativization from psych-verb constructions. the evidence in this chapter seems to support their proposal. psych-verbs. the non-subject SR can only be licensed when the move of the +Wh-DP to [Spec. and [Spec.

*The teacher1 [ who1 [ that the principal would fire t1 was expected by the reporters]] is a crusty old battleax. [kimi ga t1 au koto/no] ga atarimae no hito you meet that matter.) the person whom that you see (him) is matter of fact’ c. . Examples (1). I have added t (for trace) in 2) and (3) for ease of comparison. The teacher1 [who1 the reporters expected [that the principal would fire t1 ]] is a crusty old battleax.fact person1 ‘(Lit.) the person whom that I see (him) is difficult’ b. 1) a.) the article which that he has written (it) is well known’ 173 Japanese examples from Kuno (1973:241). b. 213 . (2) and (3) show extraction from English. respectively. 2) a. The teacher1 [who1 it was expected by the reporters [that the principal would fire t1 ]] is a crusty old battleax. [watakusi ga t1 au koto/no] ga muzukasii hito1 I meet that difficult person ‘(Lit.of. Japanese173 and Turkish174. Turkish does not obey Ross’ (1967) Sentential Subject Constraint. 174 The Turkish examples (3) through (12) are from Sezer (1986).Chapter 7: Relativization from Infinitivals in Turkish 1 Background Sezer (1986) notes that like Japanese. [kare ga t1 kaita koto] ga yoku sirarete-iru bun1 he wrote that well known-is article ‘(Lit. c.

is unacceptable. [[t1 iyile i-ce -i] son derece üpheli ol-an] hasta1 recover-FUT-3s last degree doubtful be-SR patient hastane-den yürü-yerek çı . [[t1 tamir ed-il-me-si repair do-PASS-INF-3s millions-DAT cost-SR stadium-LOC sheep-pl grazing ‘Sheep are grazing in the stadium1 [which [[(its1) being repaired] cost millions]].ı l-an] ekonomik kriz1 yet long last-FUT-3s realize-PASS-SR economic crisis memurlar-ı bunalt-tı . stadium-ACC repair do-INF expensive-DAT cost-PST ‘To repair the stadium was costly.’ 214 . civil.’ milyonlar-a malol-an] stat1-ta koyun-lar otlu-yor.servants-ACC depressed ‘The economic crisis1 [which [it is realized that [(it1) will continue longer]]] depressed the civil servants. problem-ACC solve-INF difficult-be-AOR ‘To solve the problem is difficult.’ b.’ b. [[t1 daha uzun sür-ece -i] anla. Relativizing out of an infinitival subject. Stad-ı tamir et-mek pahalı -ya maloldu. as in the (b) examples in (4) and (5). according to Sezer (among others) infinitival sentential subjects are islands. k-tı hospital-ABL walking. Problem-i çöz-mek zor-dur.’ b. 4) a.by leave-PST ‘The patient1 [who [[that (he1) would recover] was extremely doubtful]] walked out of the hospital. *[[ Ø 1 tamir et-mek] pahaliya malol-an] stad1 repair do-INF expensive-DAT cost-SR stadium-ACC Intended: ‘the stadium that to repair it was costly’ 5) a. However.3) a.’ c. We saw in (3) that sentential subjects are not islands in Turkish. * [[Ø 1 çöz-mek] zor ol-an] problem1 solve-INF difficult be-SR problem Intended: ‘the problem which to solve is difficult.

[Ali-nin [Ø 1 okumak]-tan zevk al-dı -ı ]] kitap1 Ali-gen read-INF-ABL pleasure take-NSR-3s book ‘the book that Ali enjoys reading’ (lit: ‘the book Ali gets pleasure from (to) read’ ) 9) a. Bakan [meclis-te konu -mak] iste-di minister parliament-LOC speak-INF want-PST ‘The minister wanted to speak in the parliament. [Ali-nin [[Ø 1 gir-mek] için] on yı u ra -tı -ı l ] okul1 Ali-GEN enroll-INF for ten year struggle-NSR-3s school ‘the school that Ali tried for ten years to get into’ 215 . [bakan-ı n [Ø 1 konu -mak] iste-di -i] meclis1 minister-GEN speak-INF want-PST-3S parliament ‘the parliament that the minister wanted to speak (in)’ 7) a. enjoys) reading the book. And the example in (9) shows that relativization out of infinitival adjunct clauses is also possible.’ b.’ b. Ali book-ACC read-INF-ABL pleasure take-aor ‘Ali gets pleasure from (i. 6) a. Ali okul-a gir-mek için on yı u ra -tı l Ali school-DAT enroll-INF for ten year struggle-pst ‘Ali tried for ten years to get into that school’ b. yeni idare [kitaplar-ı yasakla-ma]-ya çalı -ı yor new administration books-ACC ban-INF-DAT try-pres ‘The new administration is trying to ban (certain) books.Sezer demonstrates that infinitive clauses in Turkish are not of themselves islands. Ali [kitab-ı oku-mak]-tan zevk al-ıyor.’ b. [yeni idare-nin [Ø 1 yasakla-ma]-ya çalı -tı -ı]] kitaplar1 new administration-GEN ban-INF-DAT try-NSR-3s books ‘the books that the new administration is trying to ban’ 8) a.e. In examples (6) through (8). the infinitive is a complement or verbal argument.

175 11) a. the constraint does not seem to exist at all. if possible. *[[Ø 1 yaz-mak] Ali-ye be bin liraya otur-an] kitap1 write. [[Ø 1 yazı l-ma-sı ] Ali-ye be bin liraya otur-an] kitap1 write-PASS-INF-3s Ali-DAT 5000 lira come-SR book ‘the book that its writing cost Ali 5000 lira’ 12) a. but the infinitive verb in (11)b bears agreement inflection whereas in (12)b. [[Kitab-ı yaz-mak] Ali-ye be bin liraya otur-du book-ACC write-INF Ali-DAT 5000 lira come-PST ‘The writing of (this) book cost Ali 5000 lira.’ b.INF Ali-DAT 5000 lira come-SR book Intended: ‘the book that to write (it) cost Ali 5000 lira’ Unfortunately. Let’ s take another look at the bad examples in (4) and 175 We will look at inflected infinitival constructions a little further on in this chapter. the constraint in (10) is stipulative. [[Kitab-ı yazı n l-ma-sı ] Ali-ye be bin liraya otur-du book-GEN write-PASS-INF-3s Ali-DAT 5000 lira come-PST ‘The writing of (this) book cost Ali 5000 lira. both infinitival clauses are subjects. In these examples.’ b. 216 . In fact. Sezer formulates the constraint in (10). and it would obviously be better to identify a principled account of the phenomena.To explain the unacceptability of extraction from infinitival sentential subjects. This constraint seems to correctly predict the minimal pair in (11)b and (12)b. the verb is an uninflected infinitive. 10) The Unmarked Sentential Subject Constraint Nothing may relativize out of a clause that is unmarked for agreement and is dominated by a subject NP node.

We have already seen that DPs with accusative case are barred from moving to [Spec.(5) repeated as (13) and (15).SR villager-pl ‘the villagers whose houses that to repair (them) was costly’ b. stadium-ACC repair do-INF expensive-DAT cost-PST ‘To repair the stadium was costly. but the relative head inside the infinitival subject has accusative case. the result is a perfectly well-formed relative clause. villager-pl-GEN houses-AGR-ACC repair do-INF expensive-DAT cost. TP]. [[ ehir-in sokaklar-ı ] tamir et-mek] pahalı -nı -ya malol-du city-GEN streets-AGR-ACC repair do-INF expensive-DAT cost. Let’ s see what happens when we tweak the infinitival subject a little such that the relativized head has some other case. Notice that in (13)b the RC is the SR form which means that the relativized expression would have to move through the RC [Spec. Stad-ı tamir et-mek pahalı -ya maloldu. Again changing the relative head to an expression that does not have accusative case yields an acceptable relative clause. 217 . [[ köylü-ler-in evler-i-ni] tamir et-mek] pahalı -ya malol-du. [[[ Ø 1 sokaklar-ı ] tamir et-mek] pahalı -nı -ya malol-an ] ehir1 streets-AGR-ACC repair do-INF expensive-DAT cost.SR street ‘the city whose streets that to repair (them) was costly’ The same can be accomplished by tweaking example (15). as in (16). [[[ Ø 1 evler-i-ni] tamir et-mek] pahalı -ya malol-an ] köylü-ler1 houses-AGR-ACC repair do-INF expensive-DAT cost.’ b. *[[ Ø 1 tamir et-mek] pahalı -ya malol-an] stad1 repair do-INF expensive-DAT cost-SR stadium-ACC Intended: ‘the stadium that to repair it was costly’ 14) a. As shown in (14)b.PST ‘To repair the villagers’ houses was costly’ a’ .PST ‘To repair the streets of the city [lit: city’ s streets] was costly’ b’ . TP]. 13) a.

[[ara tı rmacı -lar-ı makaleler-i-ni] bastı n r-mak] kolay gel-di researcher-pl-GEN papers-AGR-ACC publish-INF easy come-PST ‘To publish the researchers’ papers was easy (lit: came easily)’ b . and 3) the accusative direct object in a vP Spec (higher than the NP subject) was barred from moving to [Spec. In previous chapters. In the constructions above. TP]. of course). Similarly. vP]. TP] because of its case. We saw that embedding in a scrambled expression circumvents the intervention from the subject. [[ Ø 1 motor-u-nu] çalı tı r-mak] zor ol-an] makine1/araba1 engine-AGR-ACC start-INF difficult be-SR machine/car ‘the machine/car which to start (its) engine is/was difficult’ b. TP] because it was an NP.’ n çalı tı r-mak] zor dur /ol-du 16) a. Problem-i çöz-mek zor-dur. * [[Ø 1 çöz-mek] zor ol-an] problem1 solve-INF difficult be-SR problem Intended: ‘the problem which to solve is difficult. by 218 .15) a. [[makine-nin /araba-nı motor-u-nu] machine-GEN/car-GEN engine-AGR-ACC start-INF difficult be-AOR/be-PST ‘To start the (this) machine’ s/car’ s engine is/was difficult’ a .’ b. 2) the NP subject blocked movement of lower expressions from moving to [Spec. problem-ACC solve-INF difficult-be-AOR ‘To solve the problem is difficult. [[ Ø 1 makaleler-i-ni] bastı r-mak] kolay gel-en] ara tı rmacı 1 lar papers-AGR-ACC publish-INF easy come-SR researchers ‘the researchers which to publish (their) papers was easy (lit: came easily)’ Let’ s be clear about exactly how we tweaked the good examples above. We determined that the unacceptability in unergatives was due to intervention effects from the subject whose base position we assumed to be in [Spec. 1) the subject could not move to [Spec. we saw that non-subject SR clauses were acceptable for unaccusative verbs but not for unergatives (the subject must always be non-specific.

TP] without intervention effects from the subject. or genitive expression kitap ‘book’ in the Spec of the accusative object is able to move to [Spec. 219 . [[Ø sayı falar-ı ]-nı kedi parçala-yan] kitab pages-AGR-ACC cat tear.176 17) a. we were able to A-move the specifier of the scrambled expression to [Spec.up-SR book ‘the book whose pages a cat tore up’ b. the “ possessor” . we saw that it was possible to extract the specifier of the object. with possessor-possessee direct object as in (17)b. and 5. See Chapters 2.up v° C° CP T° 176 Recall that we were able to get around the intervention effects of an NP subject in unergative constructions in a similar manner: by scrambling a DP/PP around the subject. accusative object prior to extraction of possessor ‘book-GEN’ : [DP [DP kitab-ı n] sayı falar-ı ]-nı book-GEN pages-AGR-ACC ‘the book’ s pages’ c.cat +Wh-book +GEN pages VP D° DP1 [book’s pages] V° tear.making the accusative object more complex. As demonstrated in the tree in (17)c. TP]. thus triggering the SR form. 4. +Wh-book+GEN TP +Wh-book+GEN vP DP1+ACC NP.

178 I use the term “ argument” in its standard usage here. 18) a. We will look more closely at extraction from inflected infinitivals later in this chapter.NSR city ‘the city whose streets that to repair (them) was costly’ c. The verbs in (19)b-c select for inherently case-marked “ arguments” 178 while the verb iste ‘want’ in (19)a assigns optional accusative case. *[[[ Ø 1 sokaklar-ı ] tamir et-mey]-in pahalı -nı -ya malol-du -u ] ehir1 streets-AGR-ACC repair do-INF-GEN expensive-DAT cost. It is interesting that the NSR form is unacceptable for all these cases. *[[[ Ø 1 evler-i-ni] tamir et-mey]-in pahalı -ya malol-du -u ] köylü-ler1 villager-pl ‘the villagers whose houses that to repair (them) was costly’ houses-AGR-ACC repair do-INF-GEN expensive-DAT cost. * [[ Ø 1 makaleler-i-ni] bastı r-may]-ı n kolay gel-di -i] ara tı rmacı 1 lar papers-AGR-ACC publish-INF-GEN easy come-NSR researchers ‘the researchers which to publish (their) papers was easy (lit: came easily)’ Let’ s return to Sezer’ s examples (6)-(8) of infinitivals in positions other than subjects that did not exhibit “ island” effects. except for the infinitival direct object in (6). In all these examples. *[[[ Ø 1 motor-u-nu] çalı tı r-may]-ı zor n ol-du -u] makine1/araba1 engine-AGR-ACC start-INF-GEN difficult be-NSR machine/car ‘the machine/car which to start (its) engine is/was difficult’ d. Although both inflected and uninflected forms are infinitivals. as their equivalents in (18) demonstrate. Here. I will use the term ‘infinitival’ to refer to a construction containing an infinitive with no agreement inflection.177 Now.NSR b. (This is different from other Chapters of this thesis where I used this term to refer strictly to those expressions merged without inherent case.2 Uninflected Infinitivals In this section we will be mainly looking at uninflected infinitivals. I use the bare term ‘infinitival’ to denote the latter to reduce wordiness. the infinitival phrases bear case. This 177 220 . note that the well-formed infinitival RC examples (14)a -b and (16)a -b above have the SR form. The matrix verbs in (6)-(8) are listed in (19).

19) a. iste: want b. çalı : try/work.on

Ay e ders(-ı ) iste-di Ay e homework(-ACC) want-PST Ay e ders-e çalı -tı Ay e homework-DAT work-PST Ay e ders-ten zevk al-dı Ay e homework-ABL enjoy-PST

c. zevk.al: enjoy (Lit: gain pleasure)

Now note that the examples in (6) are unacceptable when the infinitival complement clause is marked with accusative case, as in (20). 20) a. *Bakan [meclis-te konu -may]-ı iste-di minister parliament-LOC speak-INF-ACC want-PST ‘The minister wanted to speak in the parliament.’ b. *[bakan-ı n [Ø 1 konu -may]-ı iste-di -i] meclis1 minister-GEN speak-INF-ACC want-PST-3S parliament ‘the parliament that the minister wanted to speak (in)’ So, whereas infinitival clauses can be case-marked, for example, (7)b with dative case and (8)b with ablative case, the facts in examples (18) and (20) lead to the conclusion that uninflected infinitival clauses do not permit structural case marking. Because structural case in Turkish is uniformly assigned in a Spec-Head configuration, after raising to a functional projection, we will assume the same for uninflected infinitivals. That is, infinitivals as direct objects in transitives do not raise to [Spec, vP], and infinitivals as sentential subjects do not raise to [Spec, TP]. If they did, they would be marked with genitive case as subject of RCs. This prohibition against raising is what forces the SR form for RCs with infinitival subjects and prohibits the NSR form; [Spec, TP] is left vacant for the +Wh expression to move through. The EPP of T

denotation was for expediency and did not carry any theoretical import except perhaps to illustrate that we lack a term to refer to those expressions that merge into theta positions without lexical case.)

221

must be satisfied,179 and in the absence of the (sentential) subject moving to [Spec, TP], the Wh-expression must move there.180 The above assumptions have a further implication: we are, in essence, assuming that uninflected infinitival clauses are NPs: NPs cannot have structural case and they must remain in their first merge positions (but they allow extraction from within them). We see the same behavior for uninflected infinitivals, so we will consider them syntactic NPs. Let’ s review our conclusions thus far for uninflected infinitivals: Uninflected infinitival clauses are NPs in terms of their syntactic behavior: they do not move, nor can they be structurally case-marked.181 Infinitival clauses are not islands. Now let’ s look at another example from Kornfilt (1997), (21)a from (21)b. 21) a. *[ Ø 1 yüz-mek] güzel ol-an] deniz1 swim-INF nice be-SR sea Intended: ‘the sea which to swim (in) is nice’ b. [ deniz-de yüz-mek] güzel.dir sea-LOC swim-INF nice.be-AOR ‘It’ s nice to swim in the sea’

We said that the reason extraction from the infinitival sentential subject in (4) and (5) was bad was because the Wh-element has accusative case which was barred from
179

I am assuming that T has an EPP feature when it is selected by C. In a sentence with an infinitival sentential subject with no relativization, I assume no CP projection; T in this sentence will have no EPP feature. 180 The reader must already have several questions in mind regarding the nature of these infinitival phrases and case-marking. I ask for the reader’ s patience as I try to present the issues one by one. 181 Note that in example (15)a, the verb is in the aorist tense which must be used for generic subjects. I assume that the aorist either has no T projection or has a defective T, one that neither assigns case nor has an EPP feature. This is consistent with the subject in this example being an NP.

222

moving to [Spec, TP]. But in (21), the relative head has inherent locative case prior to extraction. We have seen many examples where locative extraction using the SR form was fine. The examples in (22)a and (23)a are also bad, even though the relativized expression is marked with inherent case in each.

22) a. *[[ Ø 1 ho lan-mak] zor ol-an] kı 1 z like-INF difficult be-SR girl Intended: ‘the girl who to like is difficult’ b. [(bu) kı z-dan ho lan-mak] zor-dur difficult be-AOR (this) girl-ABL like-INF ‘It’ s difficult to like this girl’ ‘(Lit.) It’ s difficult to feel good from this girl’ 23) a. *[[ Ø 1 git-mek] kolay ol-ma-yan] ehir1 go-INF easy be-neg-SR town Intended: ‘the town that to go to is not easy’ b. [(o) ehir-e git-mek] kolay de il (that) town-DAT go-INF easy be-neg ‘It’ s not easy to go to that town’ Again, the problem seems to be an intervention effect because if we embed the Whexpression in a PP and scramble the PP, we derive an acceptable RC, as in (24) which is minimally different from (21).

24) [[ Ø 1 için-de] yüz-mek] güzel ol-an] deniz1 in-LOC swim-INF nice be-SR sea ‘the sea which to swim (in) is nice’ Let’ s look at this construction a little closer. The structure of the PP deniz-in için-de ‘in the sea’ is shown in (25)a with tree in (25)b.

223

25) a. [DP [DP deniz]-in [NP iç-in] D° ]-de sea-GEN inside-AGR-LOC ‘(Lit.) at the sea’ s inside’ b.
DP+LOC sea+GEN NP DP sea D° N° inside+AGR

Now let’ s take a look at the structure of the infinitival clause, as shown in (26).182

26) [PRO deniz-de yüz-mek] = [PRO [in the sea] to swim]
CP/NP TP PRO VP DP/PP [in the sea] T° V° swim C°/N°

Again, recall that in our story about the SR form, we saw intervention effects from the subject in unergatives but not in unaccusatives.183 We were able to circumvent the intervention effects in unergative constructions by embedding the Wh-expression in a PP and scrambling the PP above the in-situ subject in [Spec, vP]. This is shown in (27). In the bad (27)a, the subject blocks the raising of the locative Wh-expression
182

I have nothing to say just yet as to whether infinitivals have a CP layer or are only TPs. The issue is orthogonal to what is being discussed and will make no difference to the discussion at this point. 183 By subject, I mean the argument merged into the outermost theta position (not assigned lexical case).

224

‘couch’ to [Spec, TP]. However, when ‘couch’ is embedded in a complex DP/PP, this DP/PP can scramble around the subject and ‘couch’ can move to [Spec, TP], and then to [Spec, CP] triggering the SR form.

27) a. *[bayan Ø 1 otur-an] kanepe1 woman sit-SR couch Intended: ‘the couch that a woman is sitting on’ b. [[Ø 1 üst-ün]-de bayan otur-an] kanepe1 top-AGR-LOC woman sit-SR couch ‘the couch that a woman is sitting on (top of)’ [Lit: ‘the couch whose top a woman is sitting on’ ] I have argued elsewhere that scrambled elements may only scramble once after which they are frozen for further movement.184 Let us assume this is correct. The result is that in order to get around an intervening subject, a Wh-expression may not scramble because it will become frozen in the scrambled position. However, a Wh-expression embedded in the specifier of a DP, can move with the DP as it scrambles around the subject, after which the Wh-expression can move to a functional projection. Importantly, the Wh-expression must be in the specifier of the DP so that the DP and the Wh-element will be equidistant from the target.185 Returning to the example in
184

Although scrambling is beyond the scope of this work, I have speculated that the reason for this “ freeze” after one scrambling is due to Recoverability. It seems to me Feature checking leaves a “ trail” of sorts, but scrambling, as far as I can see, does not check any features; certainly, it does not check case or EPP features (see Chapters 2 and 4). So, you’ re allowed “ one free move” . If a move is possible (i.e. does not violate minimality, PF and LF conditions), a DP can take it, and the (LF) interface component can reconstruct back one possible move, but that’ s all. Reasoning through this possibility is not feasible within the confines of this work, but I mention it for future research. But see fn. 26 in Chapter 4 where I offer other possible explanations. 185 Equidistance is computed in terms of Chomksy’ s (2000) Defective Intervention Constraint, (i). (i) Defective Intervention Constraint (DIC): In the structure, > > , where > is c-command, and and match probe , but is inactive, the effects of matching are blocked. I have adopted this for all movement in the sense that if there is nothing that c-commands but not , then and are equidistant from a higher probe. This definition resolves two problems: not only does

225

as shown in (28). it also provides us with a measure of Economy. Finally. I assume. TP]. 226 . though porous for movement from within them. Quite the contrary. it define what is and is not an intervener. are frozen for A-movement. Richards (2005) also uses such a measure of equidistance in explaining movement in Tagalog.(24). for the moment. Whereas movement to the specifier of T is clearly prohibited for accusatives. When two expressions are “ equidistant” in terms of the above description. one can make the case that adjunction to T is not a case assigning position. TP]) and. accusative objects can A-scramble above the subject in [Spec. Here the complex DP/PP scrambles above PRO (which. 186 I make the assumption that this is an adjunction site for three reasons. First my assumptions: 1. and this is the reason accusative objects can scramble/adjoin to T.186 28) [ [[[DP [DP deniz-in] iç-in] D° ]-de] PRO [DP/PP deniz-in iç-in-de] yüz-mek] sea-GEN inside-AGR-LOC swim-INF ‘to swim in the sea’ ‘(lit) in the sea’ s inside PRO to swim’ CP/NP TP DP/PP+LOC PRO VP DP/PP T° V° swim C°/N° [sea’s inside]-LOC 2. adjoins to TP. let’ s revisit what has to happen inside the infinitival phrase. the movement of one expression is not more “ economical” than movement of the other. First. There is evidence that adjuncts. Second. unique specifier positions are often the culprit in moving prohibitions. I have located in [Spec.1 Review of assumptions Let’ s review what we have determined so far. nothing in the data throughout this work suggests that Turkish has multiple specifier positions. this movement freezes the expression.An uninflected infinitival clause is an NP and remains in situ.

to check T’ s EPP feature. [[ [DP Ø 1 akalar-ı n]-dan ho lan-mak] zor ol-an] kı 1 z jokes-AGR-ABL like-INF difficult be-SR girl ‘the girl whose jokes which to like is difficult’ 227 .Inherently case-marked expressions can be the relative head. TP] of the RC. in which case. triggering the SR. as in (29)b. 3. 29) a.An expression may A-scramble around PRO. is as follows: 1. This expression will move out of the scrambled DP in the infinitival phrase to the RC [Spec. the expression in its Spec can be the relative head. TP] of the RC of which the infinitival is the subject remains vacant. *[[ Ø 1 ho lan-mak] zor ol-an] kı 1 z like-INF difficult be-SR girl Intended: ‘the girl who to like is difficult’ b. TP]. Repeating (22)a as (29)a. TP]. the solution to the infinitival sentential subject puzzle then. Let’ s revisit other bad examples to see that we can indeed get around the intervention effects. 3. note that by embedding the relative head kı ‘girl’ z in a larger DP.An accusative direct object of the infinitival sentential subject may not be the relative head of the RC because an accusative expression is barred from moving to [Spec. I can now relativize out of the infinitival subject.(Because of 1) [Spec. except that they are lower than PRO in the infinitival clause. With these assumptions in mind. 2.(Because of 2) the relativized expression must move from within the infinitive clause to [Spec.2.

2 Non-subject infinitivals Let’ s now turn to non-subject infinitivals beginning with item (6) repeated as (31). PRO blocks the A-movement of a lower expression. it evades intervention by PRO and is free to A-move. Bakan [meclis-te konu -mak] iste-di minister parliament-LOC speak-INF want-PST ‘The minister wanted to speak in the parliament. like the unergative subject. [bakan-ı n [Ø 1 konu -mak] iste-di -i] meclis1 minister-GEN speak-INF want-NSR-3S parliament ‘the parliament that the minister wanted to speak (in)’ 228 . we are circumventing intervention effects for A-movement within the infinitival. We must conclude that in such a derivation. and then extract ‘city’ thereby avoiding the intervention from PRO. PRO is also an intervener. 30) a. By making the Wh-expression ehir ‘city’ a possessor of the DP [city’ s neighborhoods]. *[[ Ø 1 git-mek] kolay ol-ma-yan] ehir1 go-INF easy be-neg-SR town Intended: ‘the town that to go to is not easy’ b [[ [DP Ø 1 mahaleler-i]-ne git-mek] kolay ol-ma-yan] ehir1 neighborhoods-AGR-DAT go-INF easy be-neg-SR town ‘the town whose neighborhoods that to go to is not easy’ If this analysis is on the right track. and left behind a remnant. 2. Recall that such a strategy is not needed for A-bar movement which can be long-distance.’ b.Likewise for (23)a repeated as (30)a. When that expression is a constituent of a larger element that scrambles around PRO. it is possible to scramble this complex DP around PRO. Extraction of the locative Wh-DP was not possible unless we embedded it. we must conclude that. 31) a.

we see extraction of a dative from an adjunct infinitival clause. If this is indeed the case. we have obligatory control (OC) PRO. it lends support to the Hornstein (1999) proposal that OC PRO is derived via movement. I assume that the dative is generated lower than PRO.to ten year struggle-pst ‘Ali tried for ten years in order to get into that school’ b. In fact. and yet it can move past the PRO without the intervention effects we saw with non-OC PRO. Why would that be? Notice that in this example. and yet no embedding and scrambling was necessary. In (9) repeated as (32).to ten year struggle-NSR-3s school ‘the school that Ali tried for ten years in order to get into’ This suggests that arbitrary PRO and OC PRO are different animals. then the consequences are such that OC PRO is a residue of movement leaving behind either nothing (except relevant features it has checked on 229 . this is a case of OC PRO. Again. [Ali-nin [[PROOC Ø 1 gir-mek] için] on yı u ra -tı -ı l ] okul1 Ali-GEN enroll-INF in.order.Note in the RC in (31)b that the relative head has inherent locative case prior to movement. Extraction of a locative from a complement infinitival clause is perfectly acceptable. There seem to be no intervention effects from PRO in this construction. 32) a.order. Could it be that arbitrary PRO creates intervention effects and OC PRO does not? Let’ s look at another example. Ali [[ PROOC okul-a gir-mek] için] on yı u ra -tı l Ali school-DAT enroll-INF in.

First. In example (33)a. recall that all RCs require that the EPP of T be checked. note that the infinitival in this sentence is not an argument. i. the antecedent. note that we again have an instance of OC PRO except that it is object control. In the infinitival in (33)a neither the PRO subject nor the accusative direct object can check T’ s EPP. So the question is how did ‘book’ become the relative head? If it’ s just a matter of ‘book’ raising out of the infinitival. the infinitival lacks the verbal morphology of an RC. 187 230 .189 Wh-movement in Turkish. (in spite of the English gloss) there is no relative clause the structure of (33)a. 188 We are at this point still looking at intervention effects with the infinitival itself. then it is on the wrong side of ‘to read’ . 189 First.e. Furthermore. John bought Mary [a book [ to read Ø ] ]. is possible only in relative clauses. does not seem to be an intervener. as specifiers are leftward. movement to the CP domain. 33) a. Ahmet Ay e-DAT read-INF one book buy-PST Intended: ‘Ahmet bought Ay e a book to read’ b. features have been checked or deleted so that there is no crash at the interface levels. Second.functional heads)187 or a trace which does not serve as an intervener for A-movement from its c-commanding domain. *Ahmet Ay e-ye [ Ø 1 oku-mak] bir kitap1] al-dı . the Turkish example is unacceptable. i. Let’ s look closer at the structure of (33)a and perhaps we can account for its unacceptability.e. In stark contrast to its English counterpart in (33)b. things are not so simple. or more specifically.188 However. By this I mean that the history of the derivation is “ accessible” . The matrix direct object is kitap ‘book’ . sluicing structures and for Topic movement. As demonstrated by example (34)/(35) the controller.

again. The structure of (34) then. ‘book’ moves out of the infinitival phrase without intervention from the OC PRO subject. as shown by (34). 35) Ahmet1 Ay e2-ye [pro1/2 [PRO1/2 Ø 3 oku-mak] iste-di -i] kitab3]-ı al-dı ] . no features. i. i. which picks up its referent depending on the referent of the subject of the RC ‘[the book that he/she wants to read]’ . it receives accusative case from the v° of ‘read’ in the infinitival phrase.e.e.190 I come to this conclusion because we saw intervention effects from PRO which would not have been possible if the expression could A-bar move. 231 . The accusative ‘book’ must move 190 This could mean that either there is no CP projection or that the CP layer has no specifiers. is really as in (35) where the RC subject is a null pronoun which can take Ahmet or Ay e as an antecedent. The question is what position does the Wh-book land in in the embedded relative clause? I assume that ‘book’ is marked with accusative case inside the infinitive. Ahmet Ay e-DAT read-INF want-NSR-3s book-ACC buy-PST ‘Ahmet bought Ay e the book he/she wanted to read’ Notice though that in (34) the subject of the RC can be either Ahmet or Ay e. we are back to a case of Subject Control for the PRO of the infinitival. 34) Ahmet Ay e-ye [[Ø 1 oku-mak] iste-di -i] kitab1]-ı al-dı . where the infinitive is an argument (the direct object) of the embedded RC verb ‘want’ . EPP or uWh. Let us assume for the moment (although we will see arguments for it later) that elements moving out of an infinitival do not A-bar move.Extraction out of Object Control PRO infinitives seems to be possible. to check via phrasal movement. So. Ahmet Ay e-DAT read-INF want-NSR-3s book-ACC buy-PST ‘Ahmet bought Ay e the book he/she wanted to read’ In (35).

Returning to the movement of ‘book’ in (35). I suggest that this is what happens with possessor DPs moving to [Spec. From the RC [Spec. The clausal complement of the verb ‘tell’ must obligatorily be a DP. the only way to denote a non-specific. TP] is taken up by the RC subject. We had determined that only inherently case-marked elements may move to a structural case assigning position. the relative head ‘book’ A-bar moves to the RC [Spec. can move to [Spec. TP] of a subordinate clause because the case in that position is morphologically identical. Although this is an instance of an accusative expression moving to an accusative assigning position. But. It seems that even in English. but. Let’ s look at another example. assigned structural case by D°. DP] of the possessee D°. there is no issue of case mismatch. CP]. embedded T assigns “ genitive” case. 191 Perhaps this is because the “ thing” you utter is referential.191 These are the inflected infinitivals that we will look at in more detail in Part 2. existential utterance as the complement of said 232 . vP] of the RC verb ‘want’ . I assume that the genitive assigned by D° is also a structural case. cannot move there. which is unoccupied because the infinitival clause. the unacceptable (36). The answer here lies in the verb. after which it is promoted to the external head position. being an NP. vP]. Because a possessor DP must raise to the [Spec. TP]. crucially. as [Spec. the verb ‘tell’ takes a case-marked infinitival complement. one that must have agreement inflection. The relative clause becomes acceptable. it seems clear that [Spec. What is the difference between (34) and (36) that makes one bad and the other good? Note what happens when we change the RC verb to ‘want’ . it cannot be non-specific. vP] is the only position available. an embedded possessor genitive.directly to [Spec. as in (37). As shown in (38). this move is presumably possible because structural case is evaluated at PF.

as shown in (39). (ben) [PRO öl-mek]-ten kork-uyor-du-m I die-INF ABL fear-PROG-PST-1s ‘I was afraid of dying’ ‘(Lit. (ii) There was something said. the verb kork ‘fear’ allows for all three types of complements.36) *Ahmed-in Ay e-ye [ Ø 1 oku-mak] söyle-di -i kitap1 Ahmet-GEN Ay e-DAT read-INF tell-NSR-3s book Intended: ‘the book John told Mary [ PRO to read]’ [Ahmed-in [ Ø 1 oku-mak] iste-di -i] kitap1 Ahmet-GEN read-INF want-NSR-3s book ‘the book Ahmet wanted [ PRO to read]’ Ahmet Ay e-ye [ kitab-ı oku-ma-*(-sı )]*(-nı söyle-di ) Ahmet Ay e-DAT book-ACC read-INF -3s -ACC tell-PST Ahmet told Ay e to read the/this book’ 37) 38) Kornfilt (1997) points out that the choice of complement type. For example.51) is through the use of passive voice.) I was afraid Ahmet to have died’ (Kornfilt: p. i. (i) John said something. 39) a. 192 The term “ Factive” is used in the literature to refer to NSR –DIK complement clauses as opposed to “ Active” infinitival complement clauses. Compare (i) where something is indefinite but specific. 233 . an inflected infinitive or a “ Factive” 192 NSR clause. is determined by the verb.) I was afraid Ahmet to have died’ c. (ben) [Ahmed-in öl-me-sin]-den kork-uyor-du-m I Ahmet-GEN die-INF-3s-ABL fear-PROG-PST-1s ‘I was afraid that Ahmet had died’ ‘(Lit. with (ii) where something is indefinite and non-specific.e. These labels carry no theoretical implications for the purpose of this work. (ben) [Ahmed-in öl-dü -ün]-den kork-uyor-du-m I Ahmet-GEN die-INF-NSR-3s-ABL fear-PROG-PST-1s ‘I was afraid that Ahmet had died’ ‘(Lit. a bare infinitive.) I was afraid to die’ b.

196. however. [NP PRO öl-mek]-ten die-INF ABL b. then the ablative structures of (39)a and (39)b would be as in (40)a and (40)b. We saw elsewhere that inherently case-marked elements are ambiguously specific or non-specific. in this case an ablative. 40) a. This is expected because we are assuming that D° assigns genitive case to the expression in its specifier.193 41) [DP Ahmed-in öl-me-sin]-den Ahmet-GEN die-INF-3s-ABL DP/TP?-ABL Ahmet-GEN VP Ahmet V° die D°/T°? 193 See fn. and the inflected infinitive is a DP. If we are right in assuming that the bare infinitive is an NP. For us this means. that they may be either DPs or NPs. The approximate structure of 40)b is shown in (41) where what is important is that the subject Ahmet must have raised to a structural case-assigning position if it bears overt genitive case. respectively. That is. let’ s take a moment to remark that the subject of the inflected infinitival in (40)b bears genitive case.Note. 234 . the bare infinitive in (39)a is an NP. that the verb kork ‘fear’ takes an inherently case-marked argument. [DP Ahmed-in öl-me-sin]-den Ahmet-GEN die-INF-3s-ABL While we are looking at these structures.

like “ to read” for example. we can skirt PRO (which I am 235 .The point of these examples is to demonstrate that just as with nominals. In this example.i) iste-mi-yor-um I die-INF want-NEG-PROG-is ‘I don’ t want to die’ b. (44). This also complies with the movement account of OC PRO because N° does not assign case. By providing the infinitive a direct object that must raise to get accusative case. (ben) [PRO öl-mek](*. (ben) [Ahmed-in öl-me-sin]*(-i) iste-mi-yor-um I Ahmet-GEN die-INF-3s-ACC want-NEG-PROG-is ‘I don’ t want Ahmet to die’ In example (43). Structural case can only appear on DPs and direct object DPs of accusative transitives must bear overt accusative case. the infinitive bears inherent dative case. and because it is uninflected. we can conclude that the subject of die in (42)a raises for case reasons. as we saw earlier? That is. but D° does. whereas no such raising is necessary for the subject of the DP infinitival in (42)b. we are assuming it is an NP. the ablative is presumably lower in the structure than PRO. Relativization of the ablative from within the infinitival is not possible. one can use case-marking to determine the category. If we do not assume a special case for PRO. 42) a. we have an instance of object control. Let’ s change the infinitival verb to a transitive. either accusative or the embedded genitive. Interestingly. but it must be structural case. Is this due to intervention effects from PRO. OC PRO direct object clauses are always bare whereas the inflected clauses must have accusative case. as in (42).

TP] to satisfy T’ s EPP feature. we see that an accusative object can indeed be relativized. we saw that an accusative DP cannot move to [Spec. 195 In the sentence in (i). the antecedent of his in the instrumental phrase [his ball] cannot be Ahmet. Note that the non-specific subject. 43) (o) Ahmed-i [sı f-tan kaç-ma ]-a nı zorla-dı he Ahmet-ACC class-ABL run. I ask for the reader’ s patience as this issue will be resolved in Section 3. vP].away-INF-DAT compel-PST ‘He compelled Ahmet to run away from class’ *[ pro Ahmed-i [ Ø 1 kaç-ma ]-a zorla-dı -ı ] sı f1 nı Ahmet-ACC run. The instrumental must be lower than [Spec. a dolphin2 is playing with his*1/2 ball’ 194 236 . vP] in order to allow the binding of the pronoun by the subject. TP]. is in situ in [Spec.balı ı [top-u*1/2]-ile oynu-yor. vP]) using the “ free-rider” strategy we saw in previous chapters. dolphin. as in (45). We must assume this is because. this time trying to relativize an instrumental. and therefore T’ s EPP would not be satisfied in 45).assuming remains in its base position in [Spec. Again. I am for the moment ignoring the issue of T’ s EPP feature in the infinitival. the result in (46)b is unacceptable. (i) [Ahmed-in1 havuz-un1]-da bir yunus.away-INF-DAT compel-NSR-3s class ‘the class he compelled Ahmet to run away from’ 44) With this change in the infinitival. an instrumental is generated in a position lower than PRO. unlike the accusative direct object in a specifier of vP higher than PRO.194 45) [ pro Ahmed-i [ Ø 1 okumay]-a zorla-dı -ı ] kitap1 Ahmet-ACC read-INF-DAT compel-NSR-3s book ‘the book he compelled Ahmet to read’ Let’ s try another structure. 2 Ahmed-GEN pool-3sAGR-LOC one porpoise ball-3sAGR-INST play-PRES ‘In Ahmet’ s1 pool.195 PRO is an intervener. The locative has raised to [Spec.

PRO picks up its theta-role when it merges in [Spec. vP]. for the moment. the direct object merges in V and gets its theta-role. but then 237 . In (47) (where we are. This is demonstrated in the tree in (47) for the ‘[PRO (to) read book-ACC]’ we saw in (45). *[(o-nun) Ahmed-i [PRO Ø 1 ilac al-may]-a zorla-dı -ı ] ka ı 1 k he-GEN Ahmet-ACC medicine take-INF-DAT compel-NSR-3s spoon Intended : ‘the spoon that he compelled Ahmet to take medicine (with)’ Object control PRO seems to pattern with arbitrary PRO in that it induces intervention effects. 47) [ pro Ahmed-i [ Ø 1 okumay]-a zorla-dı -ı ] kitap1 Ahmet-ACC read-INF-DAT compel-NSR-3s book ‘the book he compelled Ahmet to read’ XP?/NP vP DP-book+ACC (+Wh) PRO VP DP-book (+Wh) X?°/N° v° V° read Here’ s an even better way to prove the point about the intervention from PRO: by “ tweaking” the unacceptable (46)b. PRO is no longer an intervener for movement of the direct object. (o) Ahmed-i [ PRO bu ka ı k-la ilac al-may]-a zorla-dı he Ahmet-ACC this spoon-INST medicine take-INF-DAT compel-PST ‘He compelled Ahmet to take medicine with this spoon’ b. and the direct object raises to a higher Spec of vP to be assigned accusative case. Recall that we can scramble over PRO. but only when the expression is structurally higher than PRO.46) a. only concerned with the derivation up to the vP projection). Extraction from these infinitival clauses is possible.

the possessor was a “ free rider” as the direct object raised over PRO for case. [ben-im sen-i [[DP Ø 2 ka ı -ı 1 ]-la PRO t1 ilac al-may]-a I-GEN you-ACC spoon-3sAGR-INST medicine take-INF-DAT zorla-dı -ı m] hastane2 compel-NSR-1s hospital ‘the hospital whose spoon I compelled you to take medicine (with)’ Another non-subject control verb is tavsiye etmek ‘to recommend’ where the controller is the indirect object. But. Extraction of an accusative object is not possible. Let’ s make the instrumental ka ı k-la “ spoon-INST” in (46)b a larger DP by giving it a +Wh possessor. ada ‘island’ in (50)b. The new sentence appears as (48)a. Again. but its specifier is free to move. 238 . I have changed the matrix (RC) subject and direct object to 1st and 2nd person pronouns to avoid any ambiguity as to the possessor of ‘spoon’ which has 3rd person singular agreement. Neither is extraction of the dative expression.the scrambled expression is “ frozen” . as in (49)b. and then move again into the matrix RC clause. Notice that the relative clause is now good! Why? Because the +Whhospital was able to ride parasitically as a constituent of a scrambled DP as it moved around (and higher than) PRO. Ben sen-i [[DP hastane-nin ka ı -ı 1 PRO t1 ilac ]-la al-may]-a I you-ACC hospital-GEN spoon-3s-INST medicine take-INF-DAT zorla-dı -m compel-PST1S ‘I compelled you to take medicine with the hospital’ s spoon’ b. 48) a. as expected. with the corresponding relative clause in (48)b. as in (51)b. the possessor of a direct object of the infinitival can be relativized.

As for (47). I will tentatively suggest that the accusative direct object of the infinitive is too far away for A-bar movement and has no available intervening A-position to move to. but only when the expression is above PRO.49) a. Remember that we are 239 . *[bize [Ø 1 git-me]-ye tavsiye et-ti -i] ada1 us-DAT go-INF-DAT recommend do-NSR-3s island Intended: ‘the island that he recommended to us to visit (go to)’ 51) a. Biz-e [her gün erbet-i iç-me]-ye tavsiye et-ti us-DAT every day syrup-ACC drink-INF-DAT recommend do-PST ‘He recommended to us to drink the syrup every day’ b. [ pro Ahmed-e [[DP Ø 1 makale-si]-ni oku-may]-a tavsiye et-ti -i] (he) Ahmet-DAT paper-AGR-ACC read-INF-ACC recommend do-NSR ara tı rmacı 1 researcher ‘The researcher whose paper he recommended to Ahmet to read’ So what can we conclude? The examples in (48) and (51) demonstrate that movement from within the infinitive is possible. pro Ahmed-e [[DP ara tı rmacı n makale-si]-ni oku-may]-a -nı (he) Ahmet-DAT researcher-GEN paper-AGR-ACC read-INF-ACC tavsiye et-ti recommend do-PST ‘He recommended to Ahmet to read the researcher’ s paper’ b. This is evidence that PRO that is controlled by an indirect object is an intervener. *[biz-e [her gün Ø 1 iç-me]-ye tavsiye et-ti -i] erbet1 us-DAT every day drink-INF-DAT recommend do-NSR-3s syrup Intended: ‘ the syrup that he recommended to us to drink every day’ git-me]-ye tavsiye et-ti 50) a. Biz-e [o ada-ya us-DAT that island-DAT go-INF-DAT recommend do-PST ‘He recommended to us to visit (go to) that island’ b. Extraction is possible only when the RC head is a “ free rider” on an expression that moves around the PRO subject.

vP] to the matrix (RC) [Spec. Whereas the T° of recommend does have an EPP feature. must move to [Spec. the subject is overt and has genitive case. In the RCs in (52)b and (53)b.’ b. CP] of the infinitive clause doesn’ t seem to be an option.assuming that movement (excluding scrambling) is driven by the EPP. the subject. we would not have seen the contrast between (46)b and (48)b. [yeni idare-nin [Ø 1 yasakla-ma]-ya çalı -tı -ı]] kitaplar1 new administration-GEN ban-INF-DAT try-NSR-3s books ‘the books that the new administration is trying to ban’ 240 . CP]. 52) a. As noted above. This means that [Spec. By the time matrix (RC) C° merges with the structure. and then to [Spec. vP] of the RC. The example in (47)b is an argument for phases. the most deeply embedded vP. What is remarkable though is that successive cyclic A-bar movement does not seem to be available to save this derivation. Movement to a CP projection of the infinitival clause is not an option. TP] to be assigned case. movement out of the infinitival is good. A-bar movement through a [Spec. yeni idare [kitaplar-ı yasakla-ma]-ya çalı -ı yor new administration books-ACC ban-INF-DAT try-pres ‘The new administration is trying to ban (certain) books. that of the infinitival has been spelt-out and is no longer accessible for even long distance A-bar movement. In these examples. the accusative object of the infinitive must have moved first to [Spec. CP]. Again. he. Let’ s return to the examples of subject OC PRO in (7) and (8) repeated as (52) and (53). and it is not possible for the direct object to move in one fell swoop from the infinitive [Spec. TP] is occupied. The verb recommend takes a dative argument which means that its v° does not assign case and does not have an EPP feature. otherwise.

Kornfilt (1997) also points this out. does not hold for inflected infinitival complements. Ali [kitab-ı oku-mak]-tan zevk al-ı yor. as in (54)b. Hasan1 [Ø 1 yarı -ı kazan-mak] isti-yor Hasan race-ACC win-INF want-PRES. (56)b. the bare infinitival complement does not permit any element between itself and the verb. Though Turkish complement clauses can generally be separated from the matrix verb by a number of elements. however. Hasan Ali-ye [Ay e-nin yarı -ı kazan-dı -ı n]-ı söyle-di Hasan Ali-DAT Ay e-GEN race-ACC win-NSR-3s-ACC tell-PST ‘Hasan told Ali that Ay e won the race’ b.’ b.PROG. [Ay e-nin yarı -ı kazan-dı -ı n]-ı Ali-ye HASAN söyle-di Ay e-GEN race-ACC win-NSR-3s-ACC Ali-DAT HASAN tell-PST ‘HASAN (focus) told Ali that Ay e won the race’ 55) a. ‘Hasan wants to win the race’ 241 . enjoys) reading the book.e. which are case-marked and permit elements between them and the verb.53) a. [Ali-nin [Ø 1 okumak]-tan zevk al-dı -ı ]] kitap1 Ali-gen read-INF-ABL pleasure take-NSR-3s book ‘the book that Ali enjoys reading’ (lit: ‘the book Ali gets pleasure from (to) read’ ) 3 Inflected infinitivals It was argued above that the bare uninflected infinitival –mak behaves like an NP. Kornfilt notes that “ the infinitive not marked for case behaves like incorporated nouns which are non-specific and are not marked for case” (p. as in (55)b. 54) a. 407). Ali book-ACC read-INF-ABL pleasure take-aor ‘Ali gets pleasure (i. This prohibition.

e. with the uninflected infinitival being an NP and the inflected infinitival being a DP. behaves syntactically as a DP).PROG. and completely in line with what we have seen for non-clausal arguments. We have concluded that the structure of these infinitival subjects is different. these examples aren’ t minimal pairs at all.b. It is this difference which explains the contrast in extraction from sentential subjects. [[DP Kitab-ı yazı n l-ma-sı ] Ali-ye be bin liraya otur-du book-GEN write-PASS-INF-3s Ali-DAT 5000 lira come-PST ‘The writing of (this) book cost Ali 5000 lira. as (57) and (58). ‘HASAN (focus) wants to win the race’ 56) a. Actually. [Ay e-nin proje-yi bitir-me-sin]-ı Ali-ye HASAN söyle-di Ay e-GEN project-ACC finish-INF-3s-ACC Ali-DAT HASAN tell-PST ‘HASAN (focus) told Ali that Ay e should finish the project’ In our framework. such as those in (59) and (60). *[Ø 1 yarı -ı kazan-mak] HASAN1 isti-yor race-acc win-INF HASAN want-PRES. 57) a. Hasan Ali-ye [Ay e-nin proje-yi bitir-me-sin]-i söyle-di Hasan Ali-DAT Ay e-GEN project-ACC finish-INF-3s-ACC tell-PST ‘Hasan told Ali that Ay e should finish the project’ b. Let’ s look at a more evenly matched pair. as (57) is a passivized infinitive. Let’ s look at Sezer’ s examples again repeating the minimal pairs in (11) and (12).’ b. the non-case-marked infinitival is an NP and must remain in-situ. and must raise for case. whereas the inflected infinitival is a DP (i. [[DP Ø 1 yazı l-ma-sı ] Ali-ye be bin liraya otur-an] kitap1 write-PASS-INF-3s Ali-DAT 5000 lira come-SR book ‘the book that its writing cost Ali 5000 lira’ 242 .

196 Now. a TP projection.-GEN streets-ACC clean-INF-3s municipality-DAT 5000 dollar-DAT come-PST b. This is not so far a stretch seeing as how the generic PRO in these sentential subject environments is non-specific. [[DP Ali-nin Ø 1 temizle-me-si] belediye-ye Ali-GEN clean-INF-3s municipality-DAT be bin dolar-a ol-an ] sokaklar1 5000 dollar-DAT come-SR streets ‘the streets which (for) Ali to clean cost the municipality 5000 dollars. (60).INF Ali-DAT 5000 lira come-SR book In our new examples. *[[ NP Ø 1 yaz-mak] Ali-ye be bin liraya otur-an] kitap1 write. [[NP Kitab-ı yaz-mak] Ali-ye be bin liraya otur-du book-ACC write-INF Ali-DAT 5000 lira come-PST ‘The writing of (this) book cost Ali 5000 lira. That is. the label on these phrases is for ease of exposition only. PRO in (60) remains in its base-generated theta position. let’ s assume that PRO in (60) does not need case. (59) and (60). With these assumptions. My point is that the infinitive in (60)/(61) lacks a case-assigning functional projection above the vP. [[DP Ali-nin sokaklar-ı temizle-me-si] belediye-ye be bin dolar-a ol-du ‘[(For) Ali to clean the streets] cost the municipality 5000 dollars.58) a. let’ s assume that the bare infinitival also lacks a structural caseassigning projection.’ b. I have nothing to say about the category of the head of such a projection and have referred to it as D°/T°. let’ s make comparable assumptions about the internal structure of these phrases. the structure of the infinitival in (60) is as in (61). and non-referential anyway. the verbs are comparable except that one is an inflected infinitive with a genitive subject (59). i.’ To be more precise. and the other is a bare infinitive. Since we have committed ourselves to treating these infinitival phrases as nominals. just as we had assumed that a bare NP does not have a case-assigning DP projection above it (although there may be other projections that would normally be between DP and NP).’ A.e. 196 243 . 59) a. Thus.

244 . TP] and is assigned genitive case.60) a. I have nothing to say at this point. as with all transitive subjects. [[NP Sokaklar-ı temizle-mek] belediye-ye be bin dolar-a otur-du streets-ACC clean-INF municipality-DAT 5000 dollar-DAT come-PST ‘[To clean the streets] cost the municipality 5000 dollars. As we saw. the infinitive in (59) is DP-like. after which the direct object streets raises to a higher vP Spec to receive accusative case. I do assume that.’ b. In contrast to (60). extraction of a genitive of the direct object was possible because there was no intervener between the relative head and the matrix clause. about the CP projection. except to say that it does not seem to have an available Spec position. In this structure. the subject raises to [Spec. The structure of the infinitive subject in (59) is shown in (62). and based on our new assumptions. this means that there is an external structural case assigning head in this clause. *[[ NP Ø 1 temizle-mek] belediye-ye be bin dolar-a otur-an] sokaklar1 clean-INF municipality-DAT 5000 dollar-DAT come-SR streets 61) [NP sokaklar-ı temizle-mek] streets-ACC clean-INF CP(?) vP C° streets+ACC (NP)-PRO VP streets clean V° v° Again. vP]. PRO first-merges in [Spec.

One option would be to postulate that the bare infinitive. we would have to stipulate that the TP projection is such that no DP may raise above it unless its EPP feature has been checked. Let’ s readdress our assumptions about the structure of these two clauses. only the NP-like bare infinitive exhibits A-movement intervention effects from PRO. when there is a non-PRO subject in [Spec. also has a TP projection. other expressions may A-bar move to the infinitival [Spec. like its inflected counterpart. CP]. of course.62) [DP Ali-nin sokaklar-ı temizle-me-si] Ali-GEN streets-ACC cl