PRONOUN-NOUN CONSTRUCTIONS AND THE SYNTAX OF DP
by
Jaehoon Choi
____________________________
A Dissertation Submitted to the Faculty of the
DEPARTMENT OF LINGUISTICS
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
In the Graduate College
THE UNIVERSITY OF ARIZONA
2014
2
THE UNIVERSITY OF ARIZONA
GRADUATE COLLEGE
As members of the Dissertation Committee, we certify that we have read the dissertation
prepared by Jaehoon Choi, entitled Pronoun-Noun Constructions and the Syntax of DP and
recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor
of Philosophy.
_______________________________________________________________________
Date: 09/20/2013
Heidi Harley
_______________________________________________________________________
Date: 09/20/2013
Andrew Carnie
_______________________________________________________________________
Date: 09/20/2013
Simin Karimi
Final approval and acceptance of this dissertation is contingent upon the candidate’s submission
of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and recommend
that it be accepted as fulfilling the dissertation requirement.
________________________________________________ Date: 09/20/2013
Dissertation Director: Heidi Harley
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of the requirements for an
advanced degree at the University of Arizona and is deposited in the University Library to be
made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided
that an accurate acknowledgement of the source is made. Requests for permission for extended
quotation from or reproduction of this manuscript in whole or in part may be granted by the head
of the major department or the Dean of the Graduate College when in his or her judgment the
proposed use of the material is in the interests of scholarship. In all other instances, however,
permission must be obtained from the author.
SIGNED: Jaehoon Choi
4
ACKNOWLEDGEMENTS
I have countless people to express my deepest gratitude to for supporting me in various
ways throughout my life in Tucson. They have either directly or indirectly aided in the
development and completion of this thesis.
First and foremost, I wish to thank the members of my dissertation committee:
Heidi Harley, Andrew Carnie, and Simin Karimi. This dissertation has immensely
benefited from helpful suggestions, feedback, guidance and mentorship provided by my
advisor, Heidi Harley. Heidi has been extremely generous with her time and endlessly
patient in her willingness to discuss many of my half-baked ideas. Her everlasting
enthusiasm for learning new things about linguistics and her passion for teaching and
helping students have enormously influenced my scholarship for the past years. I am truly
honored to be her advisee. I am also sincerely grateful to my other committee members.
Andrew Carnie has influenced me in many good ways. Taking his course, ‘Analysis &
Argumentation’, was one of the best experiences ever, and helped to shape and improve
this dissertation significantly. Also, Andrew’s expertise on predication and his insightful
comments put me on the right track in improving my dissertation. I also thank Simin
Karimi, who has been supportive and interested in my work, for her perceptive comments
and suggestions. Her feedback brought my attention to many details that I neglected.
Addressing them helped me patch bugs.
I am much indebted to my primary language consultants. Above all, Anthi Zafeiri
deserves a huge thank-you. Had I not met her, this dissertation would not have been
initiated in the first place. Anthi, as a native speaker of Greek as well as a linguist, has
been an extremely knowledgeable, and reliable consultant. She not only provided her
own judgments, but also double-checked her judgments by asking other Greek speakers
including among others Evagelia Zafeiri and Gregory Tsetsos. Greg Key (English and
Turkish) and Mercedes Tubino-Blanco (Spanish) have also been extremely helpful
consultants throughout the process of writing this dissertation. Whenever I asked a
question, they always replied to me with answers plus handful of additional information.
Their sharp and crispy judgments regarding the construction investigated in this thesis
along with their linguistic insights have been valuable.
Though it turned out that I did not deal with all the examples I collected, thanks
also go to my other friends in and outside Tucson who shared their knowledge with me:
(languages alaphabetized) Chen-chun Er, Hui-Yu Huang and Yan Chen for Chinese,
Alex Trueman, Amy Fountain, Jessamyn Schertz, Megan Stone, Ryan Nelson and Sylvia
Reed for English, Lio Mathieu for French, Christopher Bock for German, Leila
Lomashvili for Georgian, Bryan James Gordon and Yaron Hadad for Hebrew, Ronan
Havelin for Irish, Andrea Borlizzi for Italian, Hiromi Onishi, Kaori Furuya, Rie
Maruyama, Shiho Yamamoto, Shuhei Abe, Tatsuya Isono, Yosuke Sato, Yuri Piskula and
Yuta Sakamoto for Japanese, Hyeoxik Shin, Hyun Kyoung Jung and Sunghun Moon for
Korean, Roman Nikolaev and Tatyana Slobodchikoff for Russian, Muriel Fisher for
Scottish Gaelic, Jaime Parchment and Rolando Coto for Spanish, and Deniz Tat for
Turkish.
I would like to thank the current and past faculty at the University of Arizona who I
5
took classes from and/or worked for as TA or RA. This includes Adam Ussishkin, Alina
Twist, Amy Fountain, Andrew Carnie, Andy Barss, Andy Wedel, Dalila Ayoun, Diana
Archangeli, Diane Ohala, Janet Nicol, Massimo Piattelli-Palmarini, Mike Hammond,
Natasha Warner, Sheila Dooley, and Tyler Peterson. Thanks are also due to the current
and past staff in the Department of Linguistics, who covered my back so that I could
focus on my work without worrying about administrative stuff: Jennifer Columbus,
Marian Wiseley, Kimberley Young, Shayna Walker, and Jennie Bradley.
For years of financial support through teaching and research assistantships and
fellowships, I am immensely grateful to the Department of Linguistics, the Graduate
College, and the Confluence Center. Thanks also go to the Graduate and Professional
Student Council for supporting my trips to conferences.
I would like to thank Hyun Kyoung Jung on different levels. She has been an
awesome friend and great colleague throughout my life in Tucson. The emotional support
from and the fruitful discussion with her greatly helped me overcome many obstacles. I
would like to thank my current and former colleagues at the University of Arizona for the
friendship and support: Alex Trueman, Anthi Zafeiri, Chen-chun Er, Colin Gorrie, Dave
Medeiros, Deniz Tat, Eunjeong Ahn, Hyunsuk Sung, Jae-Hyun Sung, Jeff Punske,
Jessamyn Schertz, Jorge Muriel, Kara Hawthorne, Leila Lomashvili, Lindsay Butler, Lio
Mathieu, Mercedes Tubino-Blanco, Priscilla Shin, Ryan Nelson, Shannon Bischoff,
Sylvia Reed, and Yan Chen. Also thanks go to my other friends in and outside Tucson:
Minryung Song, Ryeojin Park, Soomin Jwa, Young-Gie Min, and Youngkyoon Suh.
Parts of this dissertation have been presented at conferences including among
others NELS 42, 43, and LSA 86, 87. I would like to thank the audience of the
conferences. Special thanks go to Marcel den Dikken, who was willing to spare his time
for a meeting when I visited CUNY. His comments were extremely helpful for my
dissertation.
My first step into the field of linguistics was thanks to the faculty of Ajou
University, Ho Han, Jai-Hyoung Cho, and Seung-Jae Moon. Special thanks are due to Ho
Han, who advised me to pursue a career in linguistics and supported my decision.
Finally, I appreciate the constant support overseas from my family in Korea. Had it
not been for their everlasting love, I would not have been able to finish this chapter of my
life. I dedicate my dissertation to my mother, Boo Nam Ko, my little brother, Jae Up
Choi, and to the memory of my father, Young Taek Choi.
6
DEDICATION
To my mother, little brother, and to the memory of my father.
7
TABLE OF CONTENTS
LIST OF ABBREVIATIONS............................................................................................11
ABSTRACT.......................................................................................................................12
CHAPTER 1 INTRODUCTION....................................................................................14
CHAPTER 2 PRONOUN-NOUN CONSTRUCTIONS ≒ DEMONSTRATIVENOUN CONSTRUCTIONS............................................................................................24
2.1 Introduction................................................................................................................24
2.2 Syntactic and Semantic Evidence.............................................................................25
2.2.1 Syntactic Evidence................................................................................................25
2.2.1.1 Complementary Distribution.........................................................................29
2.2.1.2 Collocation with a Reinforcer........................................................................30
2.2.1.2.1 Demonstratives and Reinforcers.............................................................30
2.2.1.2.2 Pronouns and Reinforcers......................................................................37
2.2.1.2.3 Feature Match Condition........................................................................41
2.2.2 Semantic Evidence................................................................................................44
2.2.2.1 Deictic Contrastive Interpretation.................................................................44
2.2.2.1.1 Deictic Contrastive Demonstrative-Noun Constructions and PronounNoun Constructions...........................................................................................44
2.2.2.1.2 Deictic Contrastive Nature of Demonstratives and Pronouns...............46
2.2.2.2 Generic Interpretation...................................................................................50
2.2.2.3 The Role of Person of Demonstratives and Pronouns...................................59
2.3 Some Asymmetries.....................................................................................................66
2.3.1 Anaphoric Interpretation of Demonstrative-Noun Constructions........................66
2.3.2 The Universal Availability of Deictic Contrastive Interpretation and the Lack of
Generic Interpretation...........................................................................................67
2.4 Summary.....................................................................................................................71
8
TABLE OF CONTENTS – Continued
CHAPTER 3 THE SYNTAX OF DEMONSTRATIVE-NOUN CONSTRUCTIONS...73
3.1 Introduction................................................................................................................73
3.2 The Syntactic Derivation of Demonstrative-Noun Constructions.........................75
3.2.1 The First Merge Position of Demonstratives........................................................78
3.2.1.1 Demonstratives as the Complement of the Noun...........................................80
3.2.1.2 Demonstratives as the Highest Modifier in DP.............................................82
3.2.1.3 Demonstratives as the Subject of the Nominal Predicate..............................92
3.2.1.4 Demonstratives in the Specifier of the Extended Nominal Projection.........106
3.2.2 Movement and Agreement within Demonstrative-Noun Constructions..............108
3.2.2.1 Movement.....................................................................................................108
3.2.2.2 Agreement....................................................................................................119
3.2.3 Interim Summary.................................................................................................128
3.3 Disambiguating the Meanings of Demonstrative-Noun Constructions..............129
3.4
3.3.1
Deictic Contrastive and Generic Interpretations....................................130
3.3.2
Deictic Contrastive and Anaphoric Interpretations................................133
Summary.............................................................................................................136
CHAPTER 4 THE SYNTAX OF PRONOUN-NOUN CONSTRUCTIONS.............138
4.1 Introduction..............................................................................................................138
4.2 The Syntactic Derivation of Pronoun-Noun Constructions.................................140
4.2.1 The Base Structure of Pronoun-Noun Constructions.........................................140
4.2.2 Movement............................................................................................................141
4.2.3 Mediated Pronoun-Predicate Agreement...........................................................142
4.2.3.1 Agreement inside the Pronoun-Noun Construction.....................................142
4.2.3.2 Agreement outside the Pronoun-Noun Construction...................................146
4.2.4 Motivating the Low First Merge Position of the Pronoun..................................149
4.2.5 Disambiguating the Two Meanings of Pronoun-Noun Constructions................155
9
TABLE OF CONTENTS – Continued
4.2.6 Phrasal Status of Pronouns................................................................................156
4.3 The Locus of Person and Two Types of Pronouns...............................................158
4.4 Parallelism between DP and CP.............................................................................165
4.5 Previous Analyses of Pronoun-Noun Constructions.............................................172
4.5.1 Predication Analysis...........................................................................................173
4.5.2 Head Analysis.....................................................................................................178
4.6 Revisiting English....................................................................................................182
4.7 Summary...................................................................................................................184
CHAPTER 5 PRO-DROP IN PRONOUN-NOUN CONSTRUCTIONS..................186
5.1 Introduction..............................................................................................................186
5.2 The Optionality of the Pronoun in Pronoun-Noun Constructions as Pro-Drop..187
5.3 Defining Conditions on PNC Pro-Drop.................................................................193
5.3.1 The Dependency of PNC Pro-Drop on the Clausal Domain Pro-Drop.............193
5.3.2 Not All Pro-Drop Languages Allow PNC Pro-Drop..........................................197
5.3.2.1 Typology of Pro-Drop Languages...............................................................197
5.3.2.2 Redefining the First Condition.....................................................................202
5.3.2.2.1 PNC Pro-Drop in Partial Pro-Drop Languages..................................203
5.3.2.2.2 PNC Pro-Drop in Expletive Pro-Drop Languages...............................205
5.3.2.2.3 PNC Pro-Drop in Radical Pro-Drop Languages.................................206
5.3.3 The Dependency of PNC Pro-Drop on the Definite Article...............................208
5.3.4 Interim Summary.................................................................................................214
5.4 Analysis: Mediated Pro-Drop.................................................................................216
5.4.1 Two Cornerstones for an Analysis of PNC Pro-Drop........................................217
5.4.1.1 Pro-Drop......................................................................................................217
5.4.1.2 Mediated Agreement....................................................................................220
5.4.3 Mediated Pro-Drop.............................................................................................221
10
TABLE OF CONTENTS – Continued
5.4.3 PNC Pro-Drop and Theories of Pro-Drop.........................................................223
5.4.4 When PNC Pro-Drop Is Blocked........................................................................227
5.4.4.1 Reinforcers Block PNC Pro-Drop...............................................................228
5.4.4.2 Demonstratives Are Never Dropped............................................................231
5.5 Summary...................................................................................................................234
CHAPTER 6 CONCLUDING REMARKS.................................................................235
6.1 Summary...................................................................................................................235
6.2 Questions for Future Research...............................................................................239
REFERENCES...............................................................................................................242
11
LIST OF ABBREVIATIONS
ACC
COP
DECL
DNC
EPIS
F
1
FUT
GEN
INT
M
NOM
PST
PER
PL
PRED
PRES
PNC
2
SG
SUBJ
3
TOP
Accusative
Copular
Declarative
Demonstrative-Noun Construction
Epistemological
Feminine
First person
Future
Genitive
Interrogative
Masculine
Nominative
Past
Perfective
Plural
Predicate
Present
Pronoun-Noun Construction
Second person
Singular
Subject
Third person
Topic
12
ABSTRACT
This dissertation is a study of the syntactic structure of noun phrases. In particular, this
study focuses on the Pronoun-Noun Construction (PNC) which is composed of a nonpossessive pronoun and a common noun as in We Tucsonans love rain. The core theme of
this thesis lies in the idea that the PNC forms a natural class with the DemonstrativeNoun Construction (DNC). Though this idea is not radical (Giusti 1997, 2002), neither
this claim nor its consequences has been adequately recognized or explored.
This study advances this idea by demonstrating the existence of syntactic and
semantic parallels between the PNC and the DNC. This hypothesis leads to a unified
analysis of the two constructions: the pronoun merges in the specifier of an extended
nominal projection and moves to [Spec, DP], on analogy with previous analyses of the
structure of the DNC (Giusti 1997, 2002; Panagiotidis 2000; Rosen 2003).
This proposed analysis necessitates reconsideration of important theoretical issues
in syntax. In particular, the current analysis of the PNC implies a novel view of the DPinternal locus of person, which demarcates pronominal DPs from non-pronominal DPs.
That is, the source of the valued person feature is the pronoun embedded in the DP, rather
than the D head of the DP. This view of the locus of person leads in turn to a proposal of
the agreement between PNC subject and predicate in which DP-internal agreement feeds
DP-external agreement. Third, the proposed analysis of agreement provides a
straightforward account for the optionality of the pronoun in the PNC across languages, if
coupled with a pro-drop theory in which an empty category is postulated (e.g., Rizzi
13
1986). I justify the particular choice of a pro-drop theory by showing that the competing
head-movement-based approaches to pro-drop (e.g., Alexiadou and Anagnostopoulou
1998) not extendable to pro-drop in the PNC. Lastly, I show that the dislocation of
demonstratives and pronouns to the left periphery of DP patterns with the wh-movement
to the left periphery of CP in a given language. This constitutes a new piece of evidence
for the parallelism between DP and CP.
Evidence used in this thesis is primarily drawn from Modern Greek and English,
with additional data from Chinese, German, Italian, Japanese, Korean, Modern Hebrew
Hebrew, Russian, Spanish, and Turkish.
14
CHAPTER 1
INTRODUCTION
Some languages allow a noun phrase which include a non-possessive strong1 pronoun
and a noun. A representative example of the construction in English is given in (1), which
I will refer to as the Pronoun-Noun Construction (PNC)2 throughout this study.
(1)
We Tucsonans love rain.
English
Although there has been extensive research on NPs and DPs in many languages from
diverse perspectives since Abney’s (1987) seminal work, the PNC has received relatively
less attention from syntacticians and semanticists. Besides, the empirical coverage of the
previous literature the PNC has been rather narrow, and the primary subject language has
been English.
If we turn our attention to the PNC in other languages, however, we immediately
find that the PNC in many languages behaves differently compared to the English
counterpart. One such language is Modern Greek (Greek). An example from this
language is given in (2).
(2)
1
2
Emis
*(i) glossologi imaste
we
the lingusits be.1PL.PRES
‘We linguists are smart.’
In the sense of Cardinaletti and Starke (1999).
This terminology is used in Furuya (2009).
exypni.
smart
Greek
15
A salient difference compared to the English PNC is immediately brought to our attention:
The definite article is an obligatory component in forming PNCs in Greek. This
dissertation takes this non-trivial empirical fact as a starting point, and extensively
investigates the syntactic and semantic properties of the PNC in Greek first. The resulting
analysis is then extended to other languages.
The central goal of the present study is two-fold: (i) to develop an analysis of the
PNC that can reconcile the differences between English and Greek, on the one hand, and
(ii) to explore the consequences of that analysis, on the other.
With respect to (i), there are mainly three types of analyses previously proposed in
the literature. First, it has been almost universally assumed in generative linguistics that
the pronoun which occurs as a part of PNC is a sort of definite article and thus is a
member of D°(Postal 1966; inter alia). As such, the pronoun is taken to select for an
extended nominal projection (or a simple noun phrase, depending on one’s assumptions)
as its complement and to eventually project DP. I will refer to this as the head analysis of
PNCs. This type of analysis is schematically represented in (3).
(3)
Head analysis:
DP
D°
Pronoun
…
NP
Other approaches treat PNCs as involving a predication relationship between the pronoun
and the noun (as in a small clause) (Panagiotidis and Marinis 2011; among others).
16
According to this view, the pronoun is a subject and the noun a predicate, as illustrated in
(4). I will call this as the predication analysis of PNCs.
(4)
Predication analysis:
XP
X’
Subject
Pronoun
X°
Predicate
Noun
In contrast to the more standard head-analysis and the predication analysis, I argue
in this dissertation in favor of a third view, according to which the pronoun surfaces in
[Spec, DP] as the result of movement from a lower specifier in the extended nominal
projection, as illustrated in (5), where X°stands for an intermediate functional head
between DP and NP (Giusti 1997).
(5)
DP
Pronoun
D°
D’
…
XP
Pronoun
X°
X’
...
Though this idea is not novel, it is no exaggeration to say that this claim has not been
explored to the extent that its validity and consequences can be properly addressed.
17
The present study advances this less-recognized approach to the PNC by arguing
for one main hypothesis: The PNC forms a natural class with and the DemonstrativeNoun Construction (DNC). Chapter 2 illustrates the syntactic and semantic parallels
between the DNC and the PNC (whilst admitting some differences between them). The
syntactic evidence includes the fact that the pronoun is in complementary distribution
with the demonstrative in the PNC, and the fact that both pronouns and demonstratives
can be collocated with a reinforcer and obey similar proximal/distal feature-match
restrictions in that context. The semantic evidence comes from the fact that the PNC and
the DNC receive the same interpretations: either a deictic contrastive interpretation or a
generic interpretation. The other semantic similarity is concerned with the subtle
semantic difference between DNC/PNC, on the one hand, and nominal expressions that
lack demonstratives/pronouns, on the other hand. The presence of a
demonstrative/pronoun embedded in DP imposes what I call a ‘membership restriction’:
depending on the person feature of the demonstrative or pronoun—namely, first, second
or third person—the speaker or addressee are included in, or excluded from, the set
denoted by the DNC/PNC.
Chapter 3 develops an analysis of how the DNC is built in syntax, and of the
membership restriction effect of the demonstrative. The DNC has thus far been one of the
main empirical domains of research on the syntax and semantics of DP both cross- and
intra-linguistically, and as a consequence, several analyses have been suggested. I make
extensive use of this resource and selectively adopt sub-components of the previous
analyses with modifications where necessary in the course of developing an analysis of
18
the DNC. Partially adopting the existing analyses of the DNC (Giusti 1997, 2002;
Panagiotidis 2000; Rosen 2003), I argue in favor of the view that the word order variation
in the DNC in Greek can be accounted for by demonstrative and N°movement. I propose
that the demonstrative merges in the specifier of dx°, which I define as the functional
head whose role is to introduce a deictic element (i.e., demonstratives or pronouns). The
demonstrative raises to [Spec, DP], and N°raises either as high as dx°(when there is an
adjective), or as high as Num°(when there is no adjective). The chain of demonstrative
copies created by movement undergoes chain reduction at the interfaces with PF and LF.
At PF, either the head (when demonstrative precedes the definite article) or the tail (when
it follows the article) of the chain is pronounced; at LF, the head of the chain is
interpreted regardless. Interpreting the head of the demonstrative chain at LF is required
in order for the membership restriction to be properly established. I then propose a
semantic denotation for the demonstrative that treats this membership restriction effect as
presuppositional: this type of demonstrative serves as a partial identity function of the <e,
e> type.
I further argue that the three interpretations of the Greek DNC can be distinguished
as follows. On the one hand, the deictic contrastive and generic interpretations associated
with the pre-article demonstrative are distinguished by appealing to the semantic duality
of the definite article. On the other hand, the deictic contrastive and anaphoric
interpretations of the DNC are distinguished by associating a distinct syntactic structure
with each interpretation. I propose that the anaphoric interpretation is linked to the TopP
by adopting the Split-DP hypothesis.
19
Chapter 4 borrows the analysis of the DNC outlined in Chapter 3, and applies it to
the PNC given the conclusion of Chapter 2 that the two constructions form a natural class.
That is, the pronoun enters the derivation of the PNC by merging in the specifier of the
designated functional head, dx°, which can introduce a demonstrative as well. The
surface word order—the left-most position of the pronoun within the DP—is derived by
moving the pronoun to [Spec, DP] and pronouncing the higher copy of the pronoun.
Unlike the DNC, in Greek, the PNC has a fixed word order in which the pronoun always
appears as the left-most element which precedes D°. Despite this fact, I independently
motivate the low merge position of the pronoun based on the relative word order facts
between adjective and pronoun in the PNC in Korean. Unlike Greek, Korean allows the
pronoun to surface either high or low in the structure. I show that the low position of the
pronoun is the base-position while the high one is a derived focus position based on two
position-dependent interpretations of the pronoun. The membership restriction effect
added by the presence of a pronoun is naturally captured as well. The difference between
the demonstrative and the pronoun lies in the value of the person feature, and thus
switching the person feature value yields the correct denotation of the pronoun in the
PNC. As for the demonstrative, I propose that the pronoun in question also introduces a
partial identity function of the <e, e> type, adding the presuppositional membership
restriction effect to the PNC interpretation.
Again, in parallel to the analysis of DNCs, I appeal to the semantic ambiguity of
the definite article to distinguish the deictic contrastive and generic readings which the
PNC receives.
20
This chapter proceeds further to discuss the inevitable consequences of the unified
analysis of the PNC and the DNC proposed in this study. The first consequence is
concerned with how we view the locus of person within DP. While the proposed analysis
fares well with the agreed upon view of the locus of other phi-features such as number
and gender, it conflicts with the prevailing hypothesis that the syntactic head of the PNC
(i.e., the D head) is the source of the valued person feature. If the proposed analysis is on
the right track, what follows is that the valued person feature actually originates from the
pronoun or the demonstrative present in the DP. With this distribution of phi-features in
mind, all the valued phi-features scattered across on DP-internal elements are collected
on the D head of the PNC via syntactic agreement within DP. This new outlook
concerning the locus of person in turn suggests that the agreement established between a
PNC subject and a verb involves two processes of agreement, which I call ‘mediated
agreement’: one within the subject PNC in which D°obtains the value of each phifeatures, and the other between T°and the subject PNC. I propose that mediated
agreement is implemented in terms of feature-sharing approach to agreement, as
suggested by Pesetsky and Torrego (2007).
Furthermore, parallels between the movement patterns of the
demonstrative/pronoun and wh-elements in the clausal domain are shown to constitute a
new piece of evidence for parallelism between DP and CP, which is supported by an
increasing number of authors (Szabolcsi 1983, 1987, 1994; Abney 1987; Ritter 1991;
Giusti 1996, 2005, 2006; Cardinaletti and Starke 1999; Pesetsky and Torrego 2001;
Haegeman and Ürögdi 2010; among others). The main point is that the discourse-related
21
interpretation of the moved element is correlated with its surface position: The in-situ
position (when the lower copy is pronounced) indicates an anaphoric interpretation
whereas the dislocated position (when the higher copy is pronounced) indicates a nonanaphoric interpretation. Interestingly, this interpretational asymmetry holds in the CP as
well as in the DP. In the case of Greek, in-situ demonstratives and wh-phrases, when
anaphorically interpreted, stay in-situ (see Vlachos 2012 for anaphoric in-situ wh-phrases
in Greek) while dislocated demonstratives and wh-phrases receive non-anaphoric
interpretations. This predicts a wh ex-situ language to disallow in-situ demonstratives and
a wh in-situ scrambling language to allow both in-situ and ex-situ demonstratives (as a
result of scrambling). The two predictions are borne out by English and Korean,
respectively.
Finally, I demonstrate that the proposed analysis provides a unified way of
accounting for the PNC in both Greek and English.
Chapter 5 delves into the optionality of the pronoun in the PNC. Languages differ
with respect to whether they allow the pronoun in the PNC to remain silent, as illustrated
in (6). For example, English does not allow it to be suppressed, while Greek does.
(6)
a.
b.
*(We) Tucsonans love rain.
(Emis)
*(i) glossologi imaste
we
the lingusits be.1PL.PRES
‘We linguists are smart.’
exypni.
smart
English
Greek
I argue that the optional pronoun in the PNC resembles the well-known pro-drop
phenomenon. I then inspect the environments in which pro-drop can be licensed in PNCs.
22
This task is conducted by investigating the (un)availability of such pro-drop not only in
Greek and English but also in Chinese, German, Italian, Japanese, Korean, Modern
Hebrew (Hebrew), Russian, Spanish, and Turkish. I include at least one language from
five distinct types of languages (consistent pro-drop, partial pro-drop, expletive pro-drop,
radical pro-drop, and non-pro-drop) so that we can reveal the connection between the
type of pro-drop and the availability of pro-drop in PNCs.
A straightforward account of pro-drop in PNCs emerges, if we assume the analysis
of the PNC laid out in Chapter 4 along with a pro-drop theory which assumes a null
pronominal element, pro (Rizzi 1986; inter alia). I defend the choice of this particular
type of pro-drop theories by demonstrating that another type of pro-drop theory
(Alexiadou and Anagnostopoulou 1998; inter alia), according to which pro can be
licensed by head-movement, cannot be extended to account for the phenomenon in
question. While the debate is still on-going as to which one of the two theories is superior
to the other, I argue that pro-based theories fare better in account for the possibility of a
suppressed pronoun in the PNC. Thus, the optionality of the pronoun in PNCs constitutes
a testing ground for the two major types of pro-drop theory. Finally, I discuss why the
demonstrative cannot be suppressed in the DNC, which might come as a surprise in the
present context given that the DNC is treated on par with the PNC. I conclude that the
non-recoverability of the deleted information is key to explaining this difference between
DNCs and PNCs.
This dissertation is not concerned with loose appositives, which are different from
PNCs under discussion. Appositives differ from PNCs in (at least) two respects. First, in
23
English, the definite article the cannot be inserted into PNCs (e.g., *we the linguists) but
the ungrammatical example can be saved in an apposition structure, indicated by an
intonation pause represented orthographically by a comma (e.g., we, the linguists).
Second, PNCs differs from appositives in that in English, only first and second person
plural pronouns can be a part of PNCs, whereas appositives are free from such
restrictions. See Pesetsky (1978) for more discussion (see Delorme and Dougherty 1972
for an appositive analysis). I also exclude a particular instance of an appositive lacking an
intonation break in English, we the people, that is found in the Constitution of the United
States, since it is not productively used, but rather a fixed expression. All the other
languages discussed in this thesis, such as Chinese, German, Greek, Italian, Japanese,
Korean, Modern Hebrew, Russian, Spanish, and Turkish, require an intonation pause for
appositives, and also allow only some pronouns to be a part of PNCs.
24
CHAPTER 2
PRONOUN-NOUN CONSTRUCTIONS ≒
DEMONSTRATIVE-NOUN CONSTRUCTIONS
2.1
Introduction
In this chapter, I aim to show that PNCs and DNCs share some syntactic and semantic
similarities, in order to support the main hypothesis that PNCs must be treated on par
with DNCs. On the syntactic side, the evidence that I consider includes their
complementary distribution and the possibility of being combined with a reinforcer
(section 2.2.1). On the semantic side, I consider the two interpretations which PNCs and
DNCs have in common—deictic contrastive and generic interpretations—and the role of
the person feature which pronouns and demonstratives bear in relation to the
interpretation of PNCs and DNCs (section 2.2.2). As it turns out below, however, it is not
the case that PNCs are equivalent to DNCs in every single respect: some differences
between the two constructions under investigation—such as the lack of the generic
interpretation of DNCs with post-nominal demonstratives or of singular PNCs—will be
discussed (section 2.3).
25
2.2
Syntactic and Semantic Evidence
In this section, I present evidence for the similarities between PNCs and DNCs. I first
discuss the syntactic evidence such as the facts that pronouns and demonstratives are in
complementary distribution within the same DP and can be collocated with a reinforcer
(section 2.2.1). I then discuss the semantic evidence such as the fact that both
constructions can receive either deictic contrastive or generic interpretations (section
2.2.2).
2.2.1
Syntactic Evidence
Before presenting the syntactic evidence for the main hypothesis of this work, a brief
discussion of some basic facts about DNCs and PNCs is necessary to move forward. It
has been reported in ample literature that the demonstrative can surface in different
positions within DP in Greek (Brugè1996, 2002; Brugèand Giusti 1996; Giusti 1997,
2002; Grohmann and Panagiotidis 2004, 2005; Guardiano 2009, 2011; Horrocks and
Stavrou 1987; Panagiotidis 2000, among others; see Alexiadou et al. 2007 for an
overview). Consider (1) (from Panagiotidis (2000:718)).
(1)
a.
b.
Aftos
o
neos
this
the young
‘This young man’
O
andras
aftos
the man
this
‘This man’
andras
man
Greek
26
c.
O
neos
aftos
the young
this
‘This young man’
andras
man
(1a), (1b), and (1c) represent the pre-article position, the post-nominal position, and the
post-adjectival position of demonstratives within DP, respectively. What is of importance
to the discussion that follows is the fact that a specific interpretation is associated with
each specific position of the demonstrative: a deictic contrastive interpretation obtains
with the pre-article demonstrative, on the one hand, while an anaphoric interpretation
obtains with the non-pre-article demonstrative, on the other hand.1, 2 Another crucial
property of DPs containing a demonstrative in Greek is that the demonstrative, regardless
of its position within DP, must co-occur with a definite article. The absence of a definite
article incurs ungrammaticality, as in (2).
(2)
a.
b.
c.
1
*Aftos
neos
this
young
‘This young man’
*Andras aftos
man
this
‘This man’
*Neos
aftos
young
this
‘This young man’
andras
man
Greek
andras
man
A third interpretation of DNCs, which has not been discussed in the literature cited above, is generic
interpretation. We will discuss this in section 2.2.2.2.
2
See the discussion in section 2.3.1 for discussion of the anaphoric interpretation.
27
The above facts regarding demonstratives in Greek are quite different than the facts
concerning demonstratives in English. As is well-known, in English, the distribution of
demonstratives is consistent (limiting our attention to the combination of a demonstrative,
an adjective, and a noun), as shown in (3).
(3)
a.
b.
c.
This young man
*Young this man
*Young man this
Also, the occurrence of a definite article is banned when there is a demonstrative
irrespective of the relative word order between the two, as illustrated in (4).
(4)
a.
b.
*This the young man
*The this young man
Unlike demonstratives, whose DP-internal position can vary depending on the
interpretation, the position of pronouns in PNCs is fixed to the pre-article position in
Greek. Consider (5) and (6).
(5)
Emis/Esis i
(exypni)
we/you
the smart
‘We/You (smart) linguists’
(6)
a.
b.
*I
the
*I
the
emis/esis
we/you
(exypni)
smart
glossologi
linguists
(exypni)
smart
emis/esis
we/you
glossologi
linguists
glossologi
linguists
Greek
Greek
28
c.
*I
the
(exypni)
smart
glossologi emis/esis
linguists we/you
As shown in (5) and (6), the only possible word order is Pronoun > DefiniteArticle >
(Adjective) > Noun (in a monadic DP). All the other variants are ungrammatical. Once
again, and as will become crucial to our discussion in the following chapters, the definite
article is an obligatory element in forming PNCs in Greek:
(7)
*Emis/Esis
we/you
(exypni)
smart
glossologi
linguists
Greek
The absence of the definite article renders (5) ungrammatical, as illustrated in (7). As
discussed above, the same fact also holds in DNCs (see (2)).
English PNCs allow for the same word order as Greek PNCs, differing only with
respect to the presence of a definite article. A definite article cannot be included in the
formation of PNCs in English. Compare (8a) and (8b).
(8)
a.
b.
c.
We/You (smart) linguists
*We/You the (smart) linguists
*The we/you (smart) linguists
With these basic facts about DNCs and PNCs in Greek and English in place, I will
consider two main pieces of syntactic evidence to support the claim that PNCs must be
treated on par with DNCs.
29
2.2.1.1 Complementary Distribution
If PNCs and DNCs are parallel constructions, it is naturally expected to see
complementary distribution between pronouns and demonstratives within DP. Consider
the simplest manifestation of DNCs and PNCs in (9).
(9)
Pronoun and pre-article demonstrative:
a.
*Emis/Esis
afti/eki
i
we/you
these/those the
b.
*Afti/Eki
emis/esis i
these/those
we/you
the
glossologi
linguists
glossologi
linguists
Greek
As shown above in (1), demonstratives can appear either in the pre-nominal
position or in the post-nominal position, depending on the presence or absence of an
adjective. One might reasonably wonder if the ungrammaticality of (9) is merely due to
the pre-article position of the demonstrative, and if (9) can improve with a demonstrative
in the non-pre-article positions. Interestingly, it is not only the pre-article demonstrative
but also the pre- and post-nominal demonstrative that is prohibited from co-occurring
with a pronoun within the same DP, as in (10).
(10) Pronoun and post-nominal demonstrative:
a.
*Emis/esis o
andras
aftos/ekini
we/you
the man
these/those
Pronoun and pre-nominal demonstrative:
b.
*Emis/esis o
neos
aftos/ekini
we/you
the young
these/those
Greek
andras
man
30
The fact that the co-occurrence of a pronoun and a demonstrative in any position within
the same DP is forbidden as demonstrated in (9) and (10) suggests that pronouns and
demonstratives compete for the same syntactic position.
We have seen that the presence of a definite article is obligatory in forming both
DNCs and PNCs in Greek, as shown in (2) and (6). (Recall that omission of the definite
article in any of (9) or (10) does not improve the grammaticality of the DP.) Given this
fact, we can safely conclude that the ungrammaticality shown in (9) and (10) is due to the
co-occurrence of a pronoun and a demonstrative within the same DP. This suggests that
pronouns and demonstratives compete for the same syntactic position.3
2.2.1.2 Collocation with a Reinforcer
2.2.1.2.1 Demonstratives and Reinforcers
Demonstratives in some languages are known to be combinable with a reinforcer.
Reinforcers are morphemes that are added to DPs containing a demonstrative with the
3
Giusti (2002) also argues that the position of the pronoun within PNC DP is the left periphery—namely,
[Spec, DP] (SpecFPmax in Giusti’s terminology), rather than D°(Fmin in Giusti’s terminology). She claims
that only definite articles belong to the D°category and all referential elements (e.g., demonstratives) must
check their referential feature in [Spec, DP]. Given this assumption, Giusti predicts that pronouns, which
are referential, cannot co-occur with demonstratives in Italian and Romanian, as shown in (i) (from Giusti
2002:27).
(i)
a.
b.
*Noi
we
*Noi
we
questi
these
aceşti
these
ragazzi
boys
băieti / *Noi băieti aceşti
boys we
boys these
Italian
Romanian
Even though Giusti’s view idea is correct, her account is solely based on the assumption that it is only
definite articles that occupy the head D.
31
purpose of strengthening the deictic property4 of the demonstrative or clarifying
proximity or distance (Bernstein 1997, 2001; Alexiadou et al. 2007). Consider (11) for
non-Standard English5.
(11) a.
b.
These here linguists
Those there linguists
Before entering into a further discussion of reinforcers, let us first discuss the three
criteria by which we can properly identify true reinforcerhood. This discussion is
important since we will be able to rule out adverbials, which without a close inspection,
can be mistaken for reinforcers.
According to Bernstein (1997, 2001), non-Standard English allows both
[Demonstrative-here/there-Noun], as in (11), and [Demonstrative-Noun-here/there], as in
(12).
(12) a.
b.
These linguists here
Those linguists there
The examples in (12), at first glance, seem to be a variant of the reinforcer examples in
(11). Bernstein argues, however, that this is not the case; only the cases in (11) are
instances of a true reinforcer while the cases in (12) are instances of an adverbial.
See section 2.2.2.1.2 for a discussion of deixis.
Bernstein does not clearly define what she means by ‘non-Standard English’. According to Choi (2013), it
includes, among others, northeastern working-class dialects.
4
5
32
Bernstein’s argument is based on the dependent relationship that holds between
demonstratives and here/there that is adjacent to them, but is absent with the nonadjacent here/there. The demonstratives in (12) can be replaced by a definite article (or
by an indefinite article when the noun is singular), as in (13), without affecting
grammaticality. In contrast, replacing the demonstratives in (11) with a definite article
results in ungrammaticality, as in (14).
(13) a.
b.
The linguists here
The linguists there
(14) a.
b.
*The here linguists
*The there linguists
This suggests that the presence of the here/there adjacent to demonstratives is contingent
on the presence of a demonstrative. Put differently, a reinforcer can be present if and only
if there is a demonstrative with which it can be associated. Based on this fact, Bernstein
concludes that only here/there immediately following a demonstrative constitutes true
reinforcers. As we shall see below, Bernstein’s argument can be carried over in order to
distinguish reinforcers from adverbials in Greek.
In Choi (2013), dubbing the reinforcerhood test suggested by Bernstein a
‘dependency diagnostics’, I suggest two additional diagnostics for reinforcers—namely, a
‘replacement diagnostics’ and a ‘modification diagnostics’. I show that both diagnostics
correctly filter out true reinforcers both in non-Standard English and Greek, corroborating
33
the result of Bernstein’s dependency diagnostics. The reasoning that lies behind both
diagnostics is as follows. When here and there are simply adverbials that refer to a
location, their exact content is vague. In order to clarify what one means by the adverbial
here and there, a speaker can use an alternative linguistic expression to convey more
specific information about the location referred by adverbial here and there. One can
either replace or modify here and there with a locative expression (e.g., an adpositional
phrase). In contrast, when here and there are reinforcers, such replacement and
modification should not be allowed, since reinforcers, whose function is limited to
reinforcing the deictic property of demonstratives, do not refer to a location. In a nutshell,
the application of replacement and modification diagnostics yields the same result as
Bernstein’s dependency diagnostics. Compare (15) and (16)-(17) for the result of the two
reinforcerhood tests.
(15) a.
b.
These here linguists
Those there linguists
(16) Replacement diagnostics:
a.
*These in Greece linguists
b.
*Those in Greece linguists
(17) Modification diagnostics:
a.
*These here in Greece linguists
b.
*Those there in Greece linguists
34
Replacing and modifying here and there in (15) with a prepositional phrase is banned, as
in (16) and (17), which indicates that here and there in this case are reinforcers. On the
contrary, such replacement and modification applied to the examples in (18) is allowed,
as shown in (19) and (20).
(18) a.
b.
These linguists here
Those linguists there
(19) Replacement diagnostics:
a.
These linguists in Greece
b.
Those linguists in Greece
(20) Modification diagnostics:
a.
These linguists here in Greece
b.
Those linguists there in Greece
Let us now examine reinforcers in Greek. Greek is known to allow demonstrativereinforcer constructions (Campos and Stavrou 2004; Alexiadou et al. 2007). I observe
that there are two possible word orders like English when here/there is added to a DNC.
In the simplest case, without any other modifiers such as adjectives, the two word orders
are (i) [Demonstrative-here/there-DefiniteArticle-Noun], as in (21), and (ii)
[Demonstrative-DefiniteArticle-Noun-here/there], as in (22).
(21) a.
Afti
edho i
glossologi
these
here the linguists
‘These here linguists.’
Greek
35
b.
(22) a.
b.
Ekini
eki i
glossologi
those
there the linguists
‘Those there linguists.’
Afti
i
glossologi
these
the linguists
‘These linguists here.’
Ekini
i
glossologi
those
the linguists
‘Those linguists there.’
edho
here
Greek
eki
there
This is the point at which the reinforcerhood disagnostics become important to our
discussion. That is, in order to know what to investigate, we need to set the stage properly
by determining which class of examples—(21) and/or (22)—include true reinforcers.
Choi (2013) shows that the reinforcerhood diagnostics bring us the same outcome as
English, suggesting that edho ‘here’ and eki ‘there’ in (21) are true reinforcers,6 as
illustrated in (23)-(25)7, but those in (22) are not, as illustrated in (26)-(28).
As reported by Campos and Stavrou (2004), demonstrative-reinforcer constructions always receive a
deictic contrastive interpretation. For this reason, the demonstrative-reinforcer construction must appear in
the left-most position, which is the locus of the deictic interpretation. Otherwise, the grammaticality is
degraded, as shown in (i) (adapted from Campos and Stavrou (2004:159)).
6
(i)
a.
Afto edho to
vivlio
Greek
this here the
book
‘this here book’
b.
??To vivlio afto edho
the
book this here
7
The examples in (24) and (25) can in fact be grammatical when the prepositional phrase in (24) and the
adverbial combined with the prepositional phrase in (25) modify the following noun phrase, rather than the
preceding demonstrative, as indicated by speakers reporting a restrictive interpretation in this instance. (i)
and (ii) are such cases in point.
(i)
a.
b.
Afti
these
Ekini
those
[[stin
in
[stin
in
Ellada]
Greece
Ellada]
Greece
i
the
i
the
glossologi]
linguists
glossologi
linguists
Greek
36
(23) Dependency diagnostics:
a.
*Edho
i
glossologi
here
the linguists
b.
*Eki
i
glossologi
there
the linguists
(24) Replacement diagnostics:
a.
*Afti
stin Ellada
these
in
Greece
b.
*Ekini
stin Ellada
those
in
Greece
(25) Modification diagnostics:
a.
*Afti
edho stin
these
here in
b.
*Ekini
eki stin
those
there in
Greek
i
the
i
the
Ellada
Greece
Ellada
Greece
glossologi
linguists
glossologi
linguists
Greek
i
the
i
the
Greek
glossologi
linguists
glossologi
linguists
(26) Dependency diagnostics:
a.
I
glossologi edho
the linguists here
‘The linguists here.’
b.
I
glossologi eki
the linguists there
‘The linguists there.’
(ii)
a.
b.
Afti
these
Ekini
those
[[edho stin
here in
[[eki stin
there in
Greek
Ellada]
Greece
Ellada]
Greece
i
the
i
the
glossologi]
linguists
glossologi]
linguists
Greek
This fact indicates that the grammatical versions of (24) and (25) are actually instances of an adverbial, and
thus are respectively equivalent to (27) and (28), which are examples of adverbials. The fact that the
prepositional phrase can modify the noun that it precedes may be related to the fact that adjectives with a
complement can modify the following noun in Greek. Note that the word order manifested in (27) and (28)
are preferred to the word order in (i) and (ii), however.
37
(27) Replacement diagnostics:
a.
Afti
i
glossologi
these
the linguists
‘These linguists in Greece.’
b.
Ekini
i
glossologi
those
the linguists
‘Those linguists in Greece.’
stin Ellada
in
Greece
Greek
stin Ellada
in
Greece
(28) Modification diagnostics:
a.
Afti
i
glossologi edho
these
the linguists here
‘These linguists here in Greece.’
b.
Ekini
i
glossologi eki
those
the linguists there
‘Those linguists there in Greece.’
stin Ellada
in
Greece
Greek
stin Ellada
in
Greece
The results of the application of the three diagnostics lead to the conclusion that (i)
[Demonstrative-here/there-DefiniteArticle-Noun] is a correct schematization of DNCs
containing a reinforcer whereas (ii) [Demonstrative-DefiniteArticle-Noun-here/there]
summarizes the word order of DNCs containing an adverbial.
2.2.1.2.2 Pronouns and Reinforcers
If pronouns in PNCs are to be treated on par with demonstratives in DNCs, they are
expected to have deictic properties (see section 2.2.2.1.2), and thus to be able to collocate
with a reinforcer as well. Not surprisingly, I observe two possible word orders which
exactly parallel the possibilities with demonstratives in such collocations both in English
and Greek. Both [Pronoun-here/there-DefiniteArticle-Noun], as in (29) for English
38
(without a definite article) and (30) for Greek, and [Pronoun-DefiniteArticle-Nounhere/there], as in (31) for English (without a definite article) and (32) for Greek, are
allowed.
(29) a.
b.
We here linguists
You there linguists
(30) a.
Emis edho i
glossologi
we here the linguists
‘We here linguists.’
Esis eki i
glossologi
you there the linguists
‘You there linguists.’
b.
(31) a.
b.
We linguists here
You linguists there
(32) a.
Emis i
glossologi
we the linguists
‘We linguists here.’
Esis i
glossologi
you the linguists
‘You linguists there.’
b.
edho
here
Greek
Greek
eki
there
The exact same result with respect to the reinforcerhood diagnostics for DNCs is
replicated when we apply the three diagnostics to the above cases that involve PNCs.
That is, it is only when here/there and edho/eki ‘here/there’ immediately follow the
pronoun that they function as true reinforcers in both languages. Removing the pronoun,
39
as shown in (33) and (34), replacing here/there and edho/eki ‘here/there’ with a
prepositional phrase, as shown in (35) and (36), and modifying here/there and edho/eki
‘here/there’ with a prepositional phrase, as shown in (37) and (38), yield
ungrammaticality.8
Dependency diagnostics:
(33) *Here/There linguists
(34) a.
b.
*Edho
here
*Eki
there
i
the
i
the
glossologi
linguists
glossologi
linguists
Greek
Replacement diagnostics:
(35) *We/You in Tucson linguists
(36) a.
b.
*Emis
we
*Esis
you
stin
in
stin
in
Ellada
Greece
Ellada
Greece
i
the
i
the
glossologi
linguists
glossologi
linguists
Greek
i
the
i
the
Greek
Modification diagnostics:
(37) a.
*We here in Tucson linguists
b.
*You there in Tucson linguists
(38) a.
b.
*Emis
we
*Esis
you
edho
here
eki
there
stin
in
stin
in
Ellada
Greece
Ellada
Greece
glossologi
linguists
glossologi
linguists
The examples in (36) and (38) are all grammatical (but less preferred compared to (42) and (44) when the
the modification relationship is established between the noun and PP in a parallel way to the examples in
fn. 7.
8
40
By contrast, the absence of a pronoun, and replacement and modification of here/there
and edho/eki ‘here/there’ with a prepositional phrase have no effect on grammaticality
when they follow the noun, as shown in (39), (41) and (43) for English and (40), (42) and
(44) for Greek.
Dependency diagnostics:
(39) Linguists here/there
(40) a.
b.
I
glossologi edho
the linguists here
‘The linguists here.’
I
glossologi eki
the linguists there
‘The linguists there.’
Greek
Replacement diagnostics:
(41) We/You linguists in Tucson
(42) a.
b.
Emis
i
glossologi
we
the linguists
‘We linguists in Greece.’
Esis
i
glossologi
you
the linguists
‘You linguists in Greece.’
stin Ellada
in
Greece
Greek
stin Ellada
in
Greece
Modification diagnostics:
(43) a.
We linguists here in Tucson
b.
You linguists there in Tucson
(44) a.
b.
Emis
i
glossologi edho
we
the linguists here
‘We linguists here in Greece.’
Esis
i
glossologi eki
you
the linguists there
‘You linguists there in Greece.’
stin Ellada
in
Greece
stin Ellada
in
Greece
Greek
41
The above results of the application of the three diagnostics suggest that only when
immediately preceding the noun do here/there and edho/eki ‘here/there’ function as a true
reinforcer in combination with pronouns, but not when following the noun. Hence, our
discussion of reinforcers does not have to be concerned with the latter.
2.2.1.2.3 Feature Match Condition
In the previous two sub-sections, we have seen that both demonstratives and pronouns of
DNCs and PNCs can be collocated with a reinforcer. One issue that has not been
addressed is that in both constructions the addition of a reinforcer is restricted by a
certain condition, which I dub the ‘feature match condition’. This restriction was
originally noted for demonstratives by Brugè(1996) and subsequently mentioned by
Alexiadou et al. (2007). Demonstratives can be followed by a reinforcer provided that
there is no mismatch between them with respect to the type of deictic interpretation (i.e.,
proximal/distal) which each receives. That is, proximal and distal demonstratives are only
compatible with a proximal and distal reinforcer, respectively. Consider (45) and (46) for
demonstrative-reinforcer constructions, first.
(45) a.
b.
These here/*there linguists
Those there/*here linguists
(46) a.
Afti
edho/eki i
these
here/there the
‘These/Those linguists.’
glossologi
linguists
Greek
42
b.
Ekini
eki/*edho i
glossologi
those
there/here the linguists
‘Those there/*here linguists.’
In the case of English, neither this/these nor that/those is ambiguous with respect to their
deictic property; the former are proximal and only compatible with here, and the latter
are distal and only compatible with there, as illustrated in (45). In Greek, on the contrary,
afti ‘these’ and its case/number variants are neutral with respect to this feature, and thus
can be reinforced by either edho ‘here’ or eki ‘there’, as in (46a) (Alexiadou et al., 2007).
Ekini ‘those’ and its variants, in contrast, are always distal, and thus adding edho ‘here’
results in ungrammaticality, whereas the addition of eki ‘there’ is fine, as shown in (46b).
I presume that the ungrammatical example in (46b) is an instance of a semantic clash
between proximal and distal feature of the demonstrative and the reinforcer.
Pronoun-reinforcer constructions are also subject to the feature match condition in
both English and Greek, as illustrated below:
(47) a.
b.
We here/*there linguists
You here/there linguists
(48) a.
Emis
edho/*eki i
we
here/there the
‘We here/*there linguists.’
Esis
edho/eki i
you
here/there the
‘You here/there linguists.’
b.
glossologi
linguists
glossologi
linguists
Greek
43
In both languages, the first person plural pronouns are always proximal by definition and
thus can be combined with a proximal reinforcer (here and edho ‘here’) but not with a
distal one (there and eki ‘there’), as in (47a) and (48a). The second person plural
pronouns are ambiguous like afti ‘this’ and its variants, and thus can be combined with
either a proximal or distal reinforcer, as in (47b) and (48b).
One might wonder why second person plural pronouns are neutral with respect to
proximity/distalness.9 The key to understanding their deictically neutral status comes
from consideration of how the concepts of proximity and distalness is defined. Proximity
and distalness are determined in relation to the speaker (Lyons 1977; Lyons 1999;
Stirling and Huddleston 2002). More specifically, in the case of English, this, referring to
something that is close to the speaker, “is roughly equivalent to ‘the one near [the
speaker]’” (Lyons 1977:648). By the same token, that in English can be paraphrased as
‘the one distant from the speaker’. Such a view of proximity provides a clue to the reason
why second person pronouns, but not first person pronouns, are deictically neutral. In the
real world, the addressee can be either close to or distant from the speaker. When the
addressee is close to the speaker, the addressee is proximal in relation to the speaker;
when the addressee is distant from the speaker, the addressee is distal in relation to the
speaker. Hence, second person pronouns are compatible with either a proximal or a distal
reinforcer, depending on the situation. As expected, the combination of a second person
pronoun and a proximal reinforcer is felicitous when the addressee is close to the speaker,
It is not clear why seemingly proximal demonstratives are of a neutral status with respect to the deictic
property in Greek. Maybe, there are two sets of such demonstratives in the lexicon: ones that are proximal
and the others that are neutral, given the fact that the demonstratives can be interpreted proximally without
the aid of the proximal reinforcer.
9
44
while the combination of a second person pronoun and a distal reinforcer is felicitous
when the addressee is distant from the speaker. The deictic properties of demonstratives
and pronouns will be discussed in more detail in section 2.2.2.1.2.
2.2.2
Semantic Evidence
This section presents semantic evidence in support of the hypothesis that PNCs should be
treated on par with DNCs. I first show that both constructions can receive a deictic
contrastive interpretation (section 2.2.2.1). I then discuss how both constructions can also
receive a generic or kind-denoting interpretation (section 2.2.2.2). I then demonstrate that
the presence of the demonstrative in DNCs and the pronoun in PNCs makes a semantic
contribution to the interpretation of the two constructions (section 2.2.2.3). These facts
are construed to lend further support to the main hypothesis of this thesis.
2.2.2.1 Deictic Contrastive Interpretation
2.2.2.1.1 Deictic Contrastive Demonstrative-Noun Constructions and Pronoun-Noun
Constructions
One of the primary interpretations of DNCs is a deictic contrastive one.10 In Greek,
demonstratives, when occurring in the pre-article position, receive a deictic contrastive
10
Another primary usage is anaphoric, which will be discussed in section 2.3.1.
45
interpretation. That is, such demonstratives can pick out referents in the real world. The
utterance of the deictic contrastive demonstrative is able to accompany a pointing gesture,
which is a primary diagnostics of the deictic contrastive usage of demonstratives (see
section 2.2.2.1.2 for more discussion). Consider (49) (from Alexiadou et al. (2007:120)).
(49) Context: At the butcher’s, pointing to a pork joint.
a.
Thelo
[afto to
apaho
butaki]DP.
want.1SG this the lean
joint
‘I want this lean joint.’
b.
#Thelo
[to butaki
afto]DP.
want.1SG the joint
this
c.
#Thelo
[to apaho
afto butaki]DP.
want.1SG the lean
this joint
Greek
Under the circumstance given in (49), it is only (49a) that is felicitous. By contrast, (49b)
and (49c), in which demonstratives occur in the non-pre-article positions, are infelicitous.
This deictic contrastive usage of DNCs can deliver a contrastive feel in the sense that the
speaker contrasts the referent of a DNC with something else. In (49), for instance, it
could be either simply the case that the speaker is picking out a lean joint which he or she
wants to purchase, or the case that the speaker is contrasting ‘this lean joint’ with another
‘lean joint’.
PNCs can also receive a deictic contrastive interpretation. Consider the two
alternative continuations of the conversation in (50).
46
(50) Context: At an award ceremony.
a.
Speaker A: Pios kerdise
to
vravio?
who won
the award
‘Who won the award?’
b.
Speaker B: Emis i
glossologi!
we the linguists
‘We linguists!’
c.
Speaker B: Esis i
glossologi!
you the linguists
‘You linguists!’
Greek
By the utterance in (50b) and (50c), Speaker B is referring to a specific group of linguists.
The utterance can naturally accompany a pointing gesture; Speaker B can point either at
himself or herself in the case of (50b) or at the addressee in the case of (50c). Once again,
the possibility of using a pointing gesture is the hallmark for deictic expressions such as
DNCs. The fact that the utterances in (50a) and (50b) are compatible with a pointing
gesture supports the existence of a deictic interpretation for PNCs. Again, the use of
PNCs in this context simultaneously makes another other group of linguists who did not
win the award salient in the conversation (as has been pointed out by Elbourne 2005), and
this interpretation is parallel to the deictic contrastive interpretation of DNCs. The fact
that both DNCs and PNCs can receive deictic contrastive interpretations demonstrates
that the two constructions are similar.
2.2.2.1.2 Deictic Contrastive Nature of Demonstratives and Pronouns
The fact that both DNCs and PNCs can receive a deictic contrastive interpretations is
expected if we consider the fact that both demonstratives and pronouns are on their own
47
deictic in nature and they can carry a contrastive feel. Their deictic nature has been
discussed in ample literature (Lyons 1977; Lyons 1999; Diessel 1999; Stirling and
Huddleston 2002; among many others). Deixis is defined as a phenomenon in which:
“the reference of certain kinds of expression is determined in relation to features of
the utterance-act: the time, the place, and the participants, i.e., those with the role
of speaker or addressee” (Stirling and Huddleston (2002:1451)). [Emphasis added]
The following extract from Lyons (1977:637) basically says the same thing:
“By deixis is meant the location and identification of persons, objects, events,
processes and activities being talked about, or referred to, in relation to the
spatiotemporal context created and sustained by the act of utterance and the
participation in it, typically, of a single speaker and at least one addressee.”
[Emphasis added]
Though deixis is related to both the speaker and the addressee, the specific deictic
property—proximal or distal—is determined in relation to the speaker, as discussed in
section 2.2.1.2.3. For this reason, the speaker is called the “deictic centre” (Stirling and
Huddleston (2002:1453); Lyons (1999:107)).
48
Given the above definition of deixis, we can take demonstratives to be deictic since
the referent of demonstratives and DNCs is typically determined by relative closeness to
the speaker (i.e., the deictic centre). Consider the English examples in (51).
(51) a.
b.
This is my book, and that is your book.
This book is mine, and that book is yours.
The referent of the deictic expressions in (51) is determined based on the relative distance
of the object from the speaker. This and this book refer to the book which is closer to the
speaker compared to a book which is farther from the speaker, referred to by that and that
book. Additionally, it is possible to contrast the referent of this (book) and the referent of
that (book), when these phrases are stressed in an appropriate context.
Since the referent of a deictic expression is not fixed, but rather varies depending
on the situation, in order to clarify the intended referent, certain ‘paralinguistic features’11
can accompany the utterance, such as “pointing with the fingers, head, or other body
parts, touching or brandishing the referent, or merely eye-movements” (Stirling and
Huddleston 2002:1452). For instance, if there are more than two books in the case of (51),
a pointing gesture can serve to identify the intended referent. The use of such
paralinguistic features as an aid for referent identification is reflected in the origin of
deixis which in Greek means ‘pointing/indicating/showing’ (Lyons 1977, Stirling and
Huddleston 2002).
Following Lyons’ (1977:637) terminology; ‘indexing acts’ in Striling and Huddleston’s (2002:1452)
terminology.
11
49
In this connection, personal pronouns are also a type of deictic (contrastive)
expression. Personal pronouns do not have a fixed referent; their referent is determined in
relation to the discourse participants, namely, the speaker and/or the addressee. Consider
(52) (adapted from Harley and Ritter (2002:487)).
(52) a.
b.
A:
B:
IA think TomC wants yourB advice.
IB think youA’re nuts. HeC wants hisD advice.
In (52), the referent of the personal pronouns varies depending on who is speaking or
listening. For instance, in (52a), I and your refer to speaker A and speaker B, respectively,
while in (52b), I and you refer to speaker B and speaker A, respectively. The referent of
he in (52b) is anaphorically linked to the linguistic antecedent Tom in (52a) mentioned by
speaker A. The referent of his in (52b) could be co-referential with a referring expression
in a preceding dialogue, as is he. It could also refer to a contextually salient entity—for
instance, John who is just entering the room where A and B are having a conversation, in
which case speaker B, knowing already that Tom wants advice from John, could indicate
the referent of his with the aid of a paralinguistic feature.12 Furthermore, the contrastive
meaning can be delivered if these pronouns are stressed.
As shown above, both demonstratives and pronouns are deictic expressions by
nature and can deliver a contrastive feel, even outside the context of DNCs and PNCs. As
According to Lyons (1977), anaphoric third person pronouns tend to be unstressed while deictic ones
tend to be stressed.
12
50
such the two expressions, when encased in DNCs and PNCs, are expected to make a
contribution to the deictic contrastive interpretation of DNCs and PNCs.
2.2.2.2 Generic Interpretation
The other meaning shared by both DNCs and PNCs is a generic or kind-denoting
interpretation. The generic interpretation of DNCs and PNCs in Greek and English, to the
best of my knowledge, has received almost no attention. Alexiadou et al. (2007: section
4.1.2) explicitly states that DNCs in Greek cannot be interpreted generically; the
literature on generic noun phrases in English does not pay attention to DNCs.
There exists, however, one work, Bowdle and Ward (1995), that discusses generic
DNCs in English. Consider the English example in (53) (from Bowdle and Ward
(1995:33)).
(53) a.
b.
A:
B:
My roommate owns an IBM ThinkPad.
Those IBM ThinkPads are quite popular.
The utterance by speaker B in the given context is concerned with the property of being
quite popular not of a specific group of IBM ThinkPads, but rather of the kind IBM
ThinkPad. Hence, B’s utterance is roughly equivalent to the statement that IBM
ThinkPads are in general quite popular.
51
The same holds for DNCs whose noun component is [+HUMAN], as in (54). (For
reasons to be discussed in section 2.3.2, we are mainly concerned with DNCs whose
referent is [+HUMAN].)
(54) These/Those linguists (sure) are smart.
The sentence in (54) is a statement about the kind linguist, rather than a specific group of
linguists. Put differently, (54) amounts to stating that linguists are in general or as a class
are smart.13 Note that though the sentence is ambiguous between the two available
readings, deictic contrastive and generic, the addition of sure facilitates the generic
interpretation. The fact that DNCs can denote a kind is further corroborated by examples
like those in (55).
(55) a.
b.
c.
These/Those generative syntacticians appeared in the late 1950s.
These/Those linguists come in many subtypes such as syntacticians,
phonologists, semanticists, etc.
These/Those linguists study the structure of sentences in a scientific way.
All the sentences in (55) are generic and their subject DNCs may denote a kind,
patterning with a well-known type of generic sentence whose subjects are bare plural
noun phrases, as presented in (56) (from Zamparelli (2002:4)).
This way of paraphrasing (54) (as well as (55)) is missing some information delivered by the sentences—
to pinpoint, information delivered by the demonstrative. It is the missing information that distinguishes
generic sentences containing DNCs (as well as PNCs to be discussed below) from the conventional generic
sentences in (56). For now, however, it suffices to paraphrase sentences in question in such a way. The
difference will be discussed in section 2.2.2.3 in more detail.
13
52
(56) Kind predicate:
a.
Domestic dogs appeared 100,000 years ago.
Taxonomy:
b.
Dogs come in many sizes.
Characterizing sentence:
c.
Dogs have four legs.
Additionally, I observe that generic sentences containing a generic DNC allow for
an exception, which is another characteristic of generic sentences (to be precise,
characterizing sentences). In garden-variety characterizing expressions, the property
denoted by the predicate of a characterizing sentence may be false of at least one of the
members that belongs to the set denoted by the subject of the sentence, as shown in (57).
(57) Linguists are smart… but I don’t think I am (smart).
(where the speaker referred to by I is also a linguist.)
In (57), the first part of the utterance characterizes the linguists as in general smart, while
the second part asserts that the property of being smart is not a characteristic of the
speaker, who is a linguist. The same pattern obtains for characterizing sentences with a
DNC subject. Denying the characterizing property of a generic sentence with a DNC
subject for at least one of the members of the set denoted by the subject is also allowed,
as in (58).
(58) These/Those linguists (sure) are smart… but I would say John is not (smart).
(where John is also a linguist that belongs to the set denoted by the subject DNC.)
53
The fact that DNCs in characterizing sentences behave in the same way as the generic
noun phrases in (56) and (57) speaks to the availability of a kind interpretation for DNCs.
As is well-known, genericity is generally assumed to be determined by the type of
the predicate. It is then expected that the generic interpretation becomes unavailable with
a predicate of non-generic type. For instance, the predicate in the sentence in (59) cannot
be generic and the DNC cannot either, when combined with such a stage-level predicate.
(59) These/Those linguists won the award yesterday.
DNCs in Greek behave exactly the same way as the English case. Consider the
examples in (60).
(60) Afti i
these the
iPhones
iPhones
einai
be.3PL.PRES
ekpliktika,
amazing
Greek
thelo
ki
ego na agoraso
ena.14
want.1SG and I
to
buy.1SG.PRES
one
‘These iPhones are amazing, I want to buy one, too.’
14
In this example, ki goes with ego to mean something like me too in English, rather than connecting the
two clauses. The fact that it is not a conjunction in this case is supported by the contrasting grammaticality
between (i) and (ii).
(i)
a.
…,
*ki
(ii)
a.
b.
c.
…,
…,
…,
[ki
ego]
thelo na
thelo na
thelo ego
na
thelo na
agoraso
agoraso
agoraso
ena.
agoraso
ena
[ki
[ki
ego]
ena.
ego].
ena.
As shown in (ii), as long as ki and ego are adjacent to each other, the sentences are grammatical.
Greek
54
In the context of (60), the subject DNC refers to the kind iPhone, but not to specific
individual iPhones. Such a generic interpretation is still available when the noun is
[+HUMAN]. Consider the examples in (61), each of which corresponds to the English
examples in (54) and (55).
(61) Individual-level predicate:
a.
Afti/Ekini i
glossologi einai
exypni.
these/those the lingusits be.3PL.PRES
smart
‘These/Those linguists are smart.’
Kind predicate:
b.
Afti/Ekini i
genetistes glossologi emfanistikan
these/those the generative linguists appear.3PL.PST
dekaetias tou 50.
decade
the 50
‘These generative linguists appeared in the late 1950s.’
Taxonomy:
c.
Afti/Ekini i
glossologi ehoun
polles
these/those the linguists have.3PL.PRES many
Greek
sta
in
teli tis
end the
eksidikefsis
specializations
opos
fonologos,
simasiologos,
klp.
such.as
phonologist
semanticist
etc
‘These linguists come in many subtypes such as phonologists, semanticists,
etc.’
Characterizing sentence:
d.
Afti/Ekini i
glossologi meletoun
glosses
me enan
these/those the lingusits study.3PL.PRES languages in
a
epistimoniko
tropo.
scientific
way
‘These/Those linguists study languages in a scientific way.’
The sentences in (61) with a DNC subject are all interpreted generically, like their
English counterparts.
55
Interestingly, a generic interpretation of the DNC is available if and only if the
demonstrative is in the pre-article position (with one exception). Recall that
demonstratives in Greek can appear in non-pre-article positions, as has been shown in
section 2.2.1.1. In its simplest manifestation without other modifiers such as adjectives,
the only available non-pre-article position is the post-nominal position, as in (62).
(62) a.
b.
c.
d.
*I
the
*I
the
glossologi
lingusits
genetistes
generative
dekaetias tou
decade
the
*I
glossologi
the linguists
afti/ekini
einai
exypni.
these/those
be.3PL
smart
glossologi afti/ekini emfanistikan
linguists these/those appear.3PL.PST
sta
in
50.
50
afti/ekini ehoun
these/those have.3PL.PRES
eksidikefsis
specializations
opos
fonologos,
such.as
phonologist
*I
glossologi afti/ekini
the lingusits these/those
epistimoniko
scientific
simasiologos,
semanticist
meletoun
study.3PL.PRES
polles
many
Greek
klp.
etc
glosses
me
languages in
teli tis
end the
enan
a
tropo.
way
All of these DNC subjects contain a post-nominal demonstrative and cannot receive a
generic interpretation, regardless of the type of the predicate or the context. The one
exception is that DNCs with a post-nominal demonstrative whose referent is non-human
can be generic, as will be discussed in section 2.3.2. The sentences in (62) all contain
human-denoting head nouns and are all ungrammatical when intended to be generically
interpreted.
56
As expected, generic sentences in Greek with a DNC subject allow for an exception
to the characterization, as we have seen above for English.
(63) Afti/Ekini i
these/those the
glossologi einai
lingusits be.3PL.PRES
exypni,
smart
Greek
alla ohi i
John.
but not the John
‘These/Those linguists are smart, but John is not.’
(where John is also a linguist that belongs to the set denoted by the subject DNC.)
Let us next turn our attention to the generic interpretation of PNCs in English and
Greek. Let us begin with the discussion with the English case. Consider the sentences in
(64):
(64) Individual-level predicate:
We/You linguists (sure) are smart.
The sentence can be a statement about the linguists as a class or in general. Hence, a
rough paraphrase of (64) would be that linguists as a class or in general are smart.15 Note
once again that the addition of sure facilitates the generic interpretation. Without it, the
sentence in (64) is ambiguous between deictic contrastive and generic interpretation.
Interestingly, but not surprisingly, PNCs behave in a parallel way to the ordinary
generic noun phrases shown in (56) above. PNCs are compatible with a kind-level
This rough paraphrase does not suffice to fully convey the meaning of (64), like the paraphrases given for
generic DNCs above, as briefly discussed in fn. 13. For now, the paraphrase is adequate for our current
discussion. The information which is missing in the paraphrase will be taken up in section 2.2.2.3.
15
57
predicate, as in (65a), PNCs can be subdivided into sub-kinds, as in (65b), and PNCs can
partake in forming characterizing sentences, as in (65c).
(65) Kind predicate:
a.
We generative syntacticians appeared in the late 1950s.
Taxonomy:
b.
We linguists come in many types such as syntacticians, phonologists, etc.
Characterizing sentence:
c.
We syntacticians study the structure of sentences in a scientific way.
The situation is identical for PNCs in Greek. The sentences in (66) constitute a case
in point.
(66) Individual-level predicate:
a.
Emis i
glossologi imaste
we the lingusits be.1PL.PRES
‘We linguists are smart.’
b.
Esis i
glossologi isaste
you the lingusits be.2PL.PRES
‘You linguists are smart.’
exypni.
smart
Greek
exypni.
smart
The sentences in (66) are ambiguous. One available reading is the deictic contrastive
interpretation (as discussed in 2.2.2.1), in which a specific group of linguists are asserted
to be smart. However, they can also mean that linguists as a class or in general smart.
Such a generic reading is available when PNCs are used in the other contexts considered
above: as subjects of a kind predicate, as in (67a-b), subjects of a taxonomic predicate, as
in (67c-d), and in a characterizing sentence, as in (67e-f).
58
(67) Kind predicate:
a.
Emis i
genetistes glossologi emfanistikame
we the generative linguists appear.3PL.PST
b.
Greek
sta teli tis dekaetias tou 50.
in
end the decade
the 50
‘We generative linguists appeared in the late 1950s.’
Esis i
genetistes glossologi emfanistikate
sta
you the generative linguists appear.3PL.PST in
teli tis
end the
dekaetias tou 50.
decade
the 50
‘You generative linguists appeared in the late 1950s.’
Taxonomy:
c.
Emis i
glossologi ehoume
polles
eksidikefsis
we the linguists have.1PL.PRES many
specializations
opos
fonologos,
simasiologos,
klp.
such.as
phonologist
semanticist
etc
‘We linguists come in many subtypes such as phonologists, semanticists, etc.’
[generic]
d.
Esis i
you the
glossologi ehete
linguists have.2PL.PRES
polles
many
eksidikefsis
specializations
opos
fonologos,
simasiologos,
klp.
such.as
phonologist
semanticist
etc
‘You linguists come in many subtypes such as phonologists, semanticists, etc.’
Characterizing sentence:
e.
Emis i
glossologi meletame
glosses
me enan
we the lingusits study.1PL.PRES languages in
a
f.
epistimoniko
tropo.
scientific
way
‘We linguists study languages in a scientific way.’
Esis i
glossologi meletate
glosses
me
you the lingusits study.2PL.PRES languages in
epistimoniko
tropo.
scientific
way
‘You linguists study languages in a scientific way.’
enan
a
59
I have shown so far that both DNCs and PNCs have something in common with
respect to their interpretations: the two constructions can be interpreted deictic
contrastively (section 2.2.2.1) or generically (section 2.2.2.2) depending on the context.
In the next sub-section, I will address the issue of the interpretational impact of the
demonstrative or pronoun, which was mentioned in fn. 13 and 15.
2.2.2.3 The Role of Person of Demonstratives and Pronouns
Another common property of DNCs and PNCs can be observed in a closer inspection of
the semantics of DNCs and PNCs with attention to the role played by the person feature
borne by demonstratives and pronouns. If we further look into the constructions under
consideration, rather than merely saying that they can be interpreted either deictic
contrastively or generically, it turns out that their person features play a crucial role in
precisely defining the meaning of the sentences that contain DNCs and PNCs.
We begin our discussion focusing on the generic interpretations of PNCs and
DNCs, in which the effect of the person feature is more readily detected compared to the
deictic contrastive case. In order to discern the subtle difference in meaning, let us
compare the generic sentences that contain a PNC or DNC-type subject the conventional
generic sentences which do not contain a DNC or PNC as their subject. Consider (68) and
(69):
60
(68) Generic PNC:
a.
Emis
i
glossologi
we
the linguists
‘We linguists are smart.’
b.
Esis
i
glossologi
you
the linguists
‘You linguists are smart.’
imaste
be.1PL.PRES
exypni.
smart
isaste
be.2PL.PRES
exypni.
smart
(69) Generic DNC:
Afti/Ekini i
glossologi einai
these/those the linguists be.3PL.PRES
‘These/Those linguists are smart.’
exypni.
smart
Greek
Greek
At first glance, it seems plausible to assume that the generic expressions in (68) and (69)
are not much different than conventional generic sentences such as (70), in the sense that
both cases are stating that being smart is a general property of linguists.
(70) I
glossologi einai
the linguists be.3PL.PRES
‘Linguists are smart.’
exypni.
smart
Greek
However, the presence or absence of a demonstrative or pronoun causes a subtle
difference in meaning. The pronoun and the demonstrative define the membership of the
discourse participants with respect to the referent of PNCs and DNCs. That is, depending
on the person feature carried by the pronoun and the demonstrative, the
speaker/addressee either must or cannot be a member of the set picked out by DNCs and
PNCs. I will dub this a ‘membership restriction’. With this in mind, let us reconsider (68)
and (69).
61
In (68), in addition to the generic assertion that linguists are generally smart, it is
also asserted that the speaker is a linguist, as in (68a), the addressee is a linguist, as in
(68b), or that neither the speaker nor the addressee is a linguist, as in (68c). In other
words, in (68a), the speaker is identifying himself/herself as a member of the set of
linguists referred to by the DP, and asserting that being smart is a general property of the
members of this set. As a result, (68a) is rendered infelicitous when uttered by a nonlinguist, as in (71a); (68b) is rendered infelicitous when uttered to a non-linguist, as in
(71b); and (69) is rendered infelicitous when uttered either by or to a linguist, as in (72):
(71) Generic PNC:
a.
A chemist:
b.
To a chemist:
(72) Generic DNC:
(To) a linguist:
#Emis
i
glossologi imaste
exypni. Greek
we
the linguists be.1PL.PRES smart
‘We linguists are smart.’
speaker ∈ {x | x is a linguist}
#Esis
i
glossologi isaste
exypni.
you
the linguists be.2PL.PRES
smart
‘You linguists are smart.’
addressee ∈ {x | x is a linguist}
#Afti/Ekinii
glossologi einai
these/those the linguists be.3PL.PRES
‘These/Those linguists are smart.’
speaker/addressee ∉ {x | x is a linguist}
exypni.
smart
Greek
The observation of a membership restriction imposed on the interpretation of DNCs
and PNCs disappears when it comes to the case of the ordinary generic expressions in
which the kind-denoting nominal is not a DNC or PNC. In the absence of a demonstrative
62
or a pronoun embedded in the generic nominal, the membership restrictions we just
observed above do not hold, as illustrated in (73).
(73) Non-DNC/PNC generic subject:
(To) anyone:
I
glossologi einai
the linguists be.3PL.PRES
‘Linguists are smart.’
exypni.
smart
Greek
The sentence in (73) merely asserts that the property of being smart holds for linguists in
general, without making any assertion about the relationship between the
speaker/addressee and the set picked out by the subject DP. Accordingly, the sentence in
(73) can be uttered by or to anyone.
Unlike generic DNCs and PNCs, their deictic contrastive counterparts seem to
behave in a slightly different way. It seems that the membership restriction still obtains
with PNCs, while it does not with DNCs. As shown in (74), the presence of a pronoun—
which entails the effect of its person feature—imposes the same membership restriction
on the speaker or the addressee.
(74) Deictic contrastive PNC:
a.
A chemist:
#Emis
i
glosologi imaste
exypni. Greek
we
the linguists be.3PL.PRES smart
‘We linguists are smart.’
speaker ∈ {x | x is a linguist that belongs to the the set
picked out by the subject DP}
63
b.
To a chemist:
#Esis
i
glosologi isaste
exypni.
you
the linguists be.3PL.PRES
smart
‘You linguists are smart.’
addressee ∈ {x | x is a linguist that belongs to the the set
picked out by the subject DP}
Noteworthy is the fact that the membership restriction that is imposed by the person
feature of the demonstrative of the generic DNC, as in (72), seems to be rendered
ineffective in the case of deictic contrastive DNCs:
(75) Deictic contrastive DNC:
(To) a linguist: Afti/Ekini i
glosologi einai
these/those the linguists be.3PL.PRES
‘These/Those linguists are smart.’
exypni.
smart
Greek
The sentence in (75) can be uttered by or to a linguist, differing from the pattern we
observed with the generic DNC in (72). The membership restriction hypothesized for the
interpretational effect of the person feature in relation to DNCs and PNCs then seems to
be challenged. Where does this difference originate?
I argue that the seeming difference is not a result of the absence of the claimed
membership restriction, but rather a result of the different nature of the set picked out by
the generic DNCs. In (75), the speaker, whoever (s)he is, is asserting that a specific group
of linguists has the property of being smart. Let us suppose then that the membership
restriction imposed by the presence of a demonstrative or a pronoun is established with
respect to the specific group of linguists, which is a proper subset of the whole group of
64
linguists. If the role of the third person—or, the lack of person—of the demonstrative of
DNCs is to exclude the speaker and the addressee from the referent of the rest of the part
of the DNC, then the speaker of (75) is excluding the addressee as well as himself or
herself from the specific group of linguists. To put it differently, as long as neither the
speaker nor the addressee is a member of the specific group of linguists, the sentence is
semantically/pragmatically well-formed. That is, the membership restriction does hold—
but with respect to the specific group, not with respect to the kind ‘linguist’—and thus the
speaker or the addressee can still be a linguist. The reason then becomes clear why the
membership restriction is superficially effective in the case of generic DNCs, as
illustrated in (72). The membership of the speaker and the addressee is set to be in
relation to linguists as a class. More specifically, in the generic DNC, the third person of
demonstrative excludes both the speaker and the addressee from the kind linguist, and
thus neither can be a linguist.
This line of reasoning can be made explicit in a certain context. Let us suppose that
in a given world, there are seven linguists in total, all of them are smart, and these/those
linguists, being interpreted deictic contrastively, refers to three of them:
•
Linguists = {A, B, C, D, E, F, G}
•
These/Those linguists = {A, B, C}
Given this situation, consider the examples in (76):
65
(76) Deictic contrastive DNC:
a.
(To) A:
#Afti/Ekinii
glossologi einai
these/those the linguists be.3PL.PRES
‘These/Those linguists are smart.’
b.
(To) G:
Afti/Ekini i
glossologi einai
these/those the linguists be.3PL.PRES
‘These/Those linguists are smart.’
exypni.
smart
Greek
exypni.
smart
As shown in (76a), if the speaker and/or the addressee is a member of the linguists
referred to by the DNC, the sentence is rendered infelicitous. By contrast, if the speaker
and/or the addressee are not in the set of linguists referred to by the DNC, the sentence is
acceptable.
It then is apparently the difference between the first and second person on the one
hand and the third person on the other hand that makes the membership restriction look
always valid on the surface for any type of PNCs. That is, what the first person and
second person on pronouns do is to establish the membership relationship to the effect
that the speaker and/or the addressee are included in whatever set is denoted by the PNC,
unlike the third person on demonstratives which exclude both the speaker and the
addressee from whatever set is denoted by the DNC. For this reason, it appears that the
membership restriction holds for PNCs whereas its effect is hidden for DNCs. However,
if I am on the right track, the line of reasoning leads to the conclusion that the
membership restriction is in fact always effective for any type of PNCs and DNCs.
66
2.3
Some Asymmetries
Before closing the discussion of the common properties shared by DNCs and PNCs, it is
worthwhile briefly discussing some differences between DNCs and PNCs. I will address
two issues: the anaphoric interpretation of DNCs, which PNCs lack (section 2.3.1), and
the universal availability of the deictic contrastive interpretation in contrast with the lack
of generic interpretation (section 2.3.2).
2.3.1
Anaphoric Interpretation of Demonstrative-Noun Constructions
As noted in section 2.2.1.1, Greek DNCs are interpreted anaphorically with a
demonstrative in a non-pre-article position. Anaphoric demonstratives serve to “refer
back to an entity that has been previously mentioned” (Alexiadou et al. (2007:121)). This
usage is illustrated in the example below (from Alexiadou et al. (2007:120-121)):
(77) Context: A paragraph from a guide book about a Greek town.
a.
I
poli eci [pola istorika
ktiria]
pu xronologhunte
the town has many historical buildings that date
b.
Greek
apo ti
vizantini epoci.
back to
Byzantine period
‘The town has many historical buildings that date back to Byzantine period.’
[Ta ktiria
afta] episceptonte
kathe xrono
ekatondadhes
the buildings these visit.3SG.PRES
every year
hundreds
turistes.
tourists
‘These buildings are visited every year by hundreds of tourists.’
67
c.
???[Afta
these
ta
the
ekatondadhes
hundreds
ktiria]
buildings
episceptonte
visit.3SG.PRES
kathe xrono
every year
turistes.
tourists
The intention of the utterances in (77b) and (77c) is to refer to the linguistic antecedent
pola istorika ktiria ‘many historical buildings’ in (77a). In this case, the DP with a postnominal demonstrative is felicitous, as in (77b), whereas the DP with a pre-article
demonstrative is infelicitous, as in (77c).16 Given that PNCs lack an anaphoric usage,
anaphoric uses of DNCs stand out as an interpretive difference between DNCs and PNCs.
It suffices to simply note this difference, for the purposes of this section. This issue will
be taken up in Chapter 4, however.
2.3.2
The Universal Availability of Deictic Contrastive Interpretations and the Lack of
Generic Interpretations
We have seen that both DNCs and PNCs can refer to a kind and be a part of a generic
sentence in English and Greek. Our observation was mainly concerned with plural DNCs
and PNCs, in which case the humanness feature has no effect on the availability of
16
The positional meaning difference of demonstratives in Greek is supported by corpus data as well:
“in spoken material, including plays, the post-nominal use of the demonstrative is statistically very
low, while the pre-article use gets a very high percentage. On the contrary, the pre-article use is very
rare in written language, scientific-theoretical composition included, in which the post-nominal use
is by far preferred. In particular, in scientific texts we find 96.47% of demonstratives in postnominal position and only 3.53% of demonstratives are prenominal. Conversely, in plays only 5.89%
of demonstratives are post-nominal and 94.1% are prenominal.” (from Alexiadou et al. (2007:121)
citing Manolessou (2000)).
68
generic interpretation. However, it is not the case that all DNCs and PNCs can be generic.
Two features seem to play a crucial role in licensing generic interpretations: [NUMBER:
SG/PL]
and [HUMAN: ±]. More specifically, among the entirety of available DNCs and
PNCs, those bearing [SG] and [+HUMAN] can never be generic in English, as in (78), or
Greek, as in (79).
(78) Intended as generic:
a.
#This/That linguist is smart.
b.
#This/That generative syntactician appeared in the late 1950s.
c.
*This/That linguist comes in many subtypes.
d.
#This/That linguist studies the structure of sentences in a scientific way.
(79) Intended as generic:
a.
#Aftos/Ekinos
this/that
b.
#Aftos/Ekinos
this/that
c.
d.
o
the
i
the
glossologos
einai
exypnos. Greek
linguist
be.3SG.PRES
smart
genetistis glossologos
emfanistike
generative linguist
appear.3SG.PST
sta teli tis
in
end the
#Aftos/Ekinos
this/that
dekaetias tou 50.
decade
the 50
i
glossologos
the linguist
eksidikefsis
specializations
#Aftos/Ekinos
this/that
opos
fonologos,
simasiologos,
klp.
such.as
phonologist
semanticist
etc
i
glossologos
emfanistike
glosses
the lingusit
study.3SG.PRES languages
me
in
enan epistimoniko
a
scientific
tropo.
way
emfanistike
have.3SG.PRES
polles
many
69
All the sentences in (78) and (79) are ungrammatical or infelicitous when the intended
reading is generic. On the contrary, singular and non-human DNCs can be generic, as
shown in (80) for English and (81) for Greek.17
(80) Intended as generic:
a.
This/That iPhone is amazing.
b.
This/That iPhone appeared in 2007.
c.
This/That iPhone comes in two colors: black and white.
d.
This/That iPhone is equipped with convenient user-interface.
(81) Intended as generic:
a.
Aftos/Ekino
to
iPhone
einai
ekpliktiko. Greek
this/that
the iPhone
be.3SG.PRES
amazing
‘This/That iPhone is amazing.’
b.
Afto/Ekino
to
iPhone
emfanistike to
2007.
this/that
the iPhone
appeared in
2007
‘This/That iPhone appeared in 2007.’
c.
Afto/Ekino
to
iPhone
vgeni
se
dio xromata:
this/that
the iPhone
comes
in
two colors:
d.
mavro
kai aspro.
black
and white
‘This/That iPhone comes in two colors: black and white.’
Afto/Ekino
to
iPhone
einai
eksoplismeno
this/that
the iPhone
be.3SG.PRES
equipped
me
with
mia kamera
ipsilis
texnologias.
a
camera
high
technology.GEN
‘This/That iPhone is equipped with a high-end camera.’
Heidi Harley (personal communication) pointed out that this generic singular reading is fine even with a
reinforcer present, as in I don’t know about your iPads, but this here iPhone is an amazing piece of
equiptment’.
17
70
Unlike the number-related asymmetry that consistently holds in both English and
Greek, another asymmetry has to do with distributions of the demonstrative within DP,
and thus is specific to Greek. As briefly mentioned in section 2.2.2.2, though DNCs with
a post-nominal demonstrative cannot receive a generic interpretation when the referent is
human, the interpretation becomes available when the referent is non-human. Recall that
the DNCs in (62), which cannot be generically interpreted, are all human. When it comes
to non-human DNCs, as shown in (82), the interpretational restriction related to
humanness of the reference ceases to hold:
(82) a.
b.
I
iPhone
afti/ekini
einai
the iPhones
these/those
be.3PL
‘These/Those iPhones are amazing,’
To iPhone
aftos/ekino
einai
the iPhone
this/that
be.3SG
‘This/That iPhone is amazing,’
ekpliktika.
amazing
Greek
ekpliktiko.
amazing
Although the immediate and salient reading of (82) is not generic but anaphoric, it is not
impossible for it to be read generically in an appropriate context.
At this point, I do not know what to make of this feature-related asymmetry; I will
leave this issue for future research, however, since this asymmetry is only observed with
non-human DNCs. For the purpose of our discussion, I will focus on those DNCs that are
human, as these are parallel to PNCs, which must always be human.
71
2.4
Summary
In this chapter, we have shown that DNCs and PNCs pattern together both syntactically
and semantically. Below is the summary of the syntactic and semantic characteristics the
two constructions commonly feature:
Syntactic similarities:
• Demonstratives and pronouns are in complementary distribution within the same
DP.
• Demonstratives and pronouns can be combined with a reinforcer, and the
reinforcer must observe the feature match condition in both cases.
Semantic similarities:
• DNCs and PNCs can be interpreted either deictic contrastively or generically.
• The person feature specifies the membership of the discourse participants in
relation to the set picked out by the DP.
The conclusion I draw from the above facts is that DNCs and PNCs form a natural class
and thus should receive the same (or at least an extremely similar) analysis.
In spite of the above shared properties, the two constructions show some
asymmetrical behaviors, as summarized below.
Dissimilarities:
72
• Pronouns can only appear in the pre-article position in Greek, whereas anaphoric
demonstratives can occur in other positions within the DP.
• The availability of the generic interpretation of DNCs and PNCs is affected by
human and number features.
Though this thesis work will not be concerned with the second difference for the reasons
mentioned in section 2.3.2, the first difference will receive an adequate discussion in
section 4.4.
Now that we have shown that DNCs and PNCs behave in a parallel way in many
regards, I will focus on developing a syntactic analysis that can capture this fact. In so
doing, I will first propose a syntactic structure of DNCs in Chapter 3. Given the
conclusion of Chapter 2, the analysis developed in Chapter 3 will be the basis of the
syntactic analysis of PNCs in Chapter 4.
73
CHAPTER 3
THE SYNTAX OF DEMONSTRATIVE-NOUN CONSTRUCTIONS
3.1
Introduction
In Chapter 2, I have argued that PNCs and DNCs must be treated on par with each other,
by showing that the two constructions have much in common from both syntactic and
semantic perspectives. In this chapter and the next, I examine how such a hypothesis can
be implemented syntactically. The first step in achieving this goal begins with
establishing the syntax of DNCs, which will be the main object of investigation in the
current chapter. DNCs have received much attention in the literature on DP structure
compared to PNCs. Thus, I review previous analyses on the syntax of DNCs mainly
focusing on literature written on Greek and other languages, and adopt some ingredients
from them with modifications where necessary. Therefore, this chapter is significant in
that it sets the stage of laying out the theoretical tools for the syntactic analysis of PNCs,
as is expected from the main hypothesis that the two constructions—DNCs and PNCs—
form a natural class. This chapter is also important on its own: in adopting (with some
modification) some components of the previous analyses of DNCs, I will provide some
arguments in favor of a particular analysis of certain aspects of the syntax of DNCs over
others.
The discussion in this chapter will lead us to the conclusion that the demonstrative
is base-generated in the specifier position of a low functional head which I dub dx°. It
74
undergoes syntactic movement to [Spec, DP] creating a chain of copies of the
demonstrative. Its pre-article surface position, associated with the deictic contrastive and
generic interpretations, emerges by pronouncing the head of the demonstrative chain, or
its pre- and post-nominal surface positions, associated with the anaphoric interpretation
of DNCs, emerge by pronouncing the tail of the demonstrative chain. In both cases, N°
undergoes head-movement as high as dx°, or further to Num°depending on whether an
adjective is present. For interpretational reasons, in the case of anaphoric DNCs, I
elaborate the left-periphery and assume that the copy of the demonstrative is a topical
element sitting in [Spec, DP/TopP], and as such is co-referential with the necessary
discourse antecedent. I further make use of the semantics of the definite article in order to
distinguish the two interpretations (deictic contrastive versus generic) available for DNCs
containing a pre-article demonstrative in Greek. I follow Giannakidou and Stavrou (1999)
in assuming the semantic duality of the definite article in Greek, and argue that the deictic
contrastive interpretation arises when the definite article functions as the extensional ιoperator, while the generic interpretation arises when the definite article is interpreted as
an intensional ι-operator, namely ∩.
This chapter is organized as follows. I begin section 3.2 with a discussion of some
relevant issues in the syntax of DNCs in order to have a solid foundation to carry over to
the syntax of PNCs. More specifically, I present the assumptions I take as given
regarding the structure of DNCs. I then discuss the low first-merge position of the
demonstrative (section 3.2.1); movement and agreement within DNCs (section 3.2.2);
75
and how to dismantle the ambiguity of DNCs (section 3.3). I summarize the main points
of this chapter (section 3.4).
3.2
The Syntactic Derivation of Demonstrative-Noun Constructions
The structure of DNCs both across and within languages has received much attention in
the literature (Horrocks and Stavrou 1987; Brugè1996, 2002; Brugèand Giusti 1996;
Campbell 1996; Giusti 1997, 2002; Vangsnes 1999; Panagiotidis 2000; Rosen 2003;
Grohmann and Panagiotidis 2004, 2005; Cinque 2005; Abels and Neeleman 2009, 2012;
Roberts 2011; Steddy and Samek-Lodovici 2011; Guardiano 2012; Medeiros 2012;
among others). It was initially argued that the demonstrative competes with the definite
article for the same syntactic position, on the basis of languages such as English. Such an
analysis is suggested because the two elements are in complementary distribution in
English and similar languages. Later on, however, other languages such as Greek,
Spanish, and others were taken into consideration, and it turns out that the account
developed based on English encounters empirical challenges. There are at least two
issues that arise from these languages. On the one hand, unlike English, in the latter
group of languages the definite article and the demonstrative can co-occur (even
obligatorily in some cases or languages; e.g., Greek, as has been discussed in Chapter 2).
On the other hand, while in English-like languages the demonstrative surfaces in a fixed
position (the left most position in the DP), in languages like Greek the position of the
76
demonstrative within DP varies depending on the intended meaning of the entire DP (as
discussed in detail above in section 2.2).
When it comes to the DNC in Greek, there is one property that is to a certain degree
agreed upon (see Alexiadou et al. 2007: section 1.4 for an overview): the low baseposition of the demonstrative. That is, the demonstrative first merges in a position lower
than D°, and word order variations stem from DP-internal movement of the
demonstrative and/or the noun. For present purposes, I mostly adopt the structure
proposed in Giusti (1997, 2002), Panagiotidis (2000), and Rosen (2003) with some
modification. Putting aside details for the moment, the base structure of DNCs that I
suppose is as illustrated in (1a). From this, all variations of word order are derived: the
pre-article position of the demonstrative is derived by moving the demonstrative to [Spec,
DP] in order to satisfy the TH-Criterion1 and pronouncing the higher copy of the
demonstrative, and by raising N°to dx°for checking the [N] category feature on dx°
when an adjective is present or further to also check the [N] feature on Num°in the
absence of an adjective, as in (1b); the post-nominal position of the demonstrative is
derived by moving the demonstrative to [Spec, DP] but pronouncing the lower copy, and
by raising the head noun through dx°to Num°, as in (1c); the pre-nominal position is
derived by moving the demonstrative to [Spec, DP] but pronouncing the lower copy, and
by raising the noun one single step, to dx°, as in (1d). The unpronounced copy is
indicated by strikethrough.
1
Note that the TH-Criterion has nothing to do with the Θ-Criteron. See section 3.2.2.1 for more details.
77
(1)
Base structure:
a.
DP
D°
Pre-article demonstrative:
b.
DP
NumP
(AP)
Num’
Num°
D’
Dem
D°
dxP
NumP
Num
Dem(onstrative) dx’
dx°
dxP
Dem
NP
dx’
dx°
NP
N°
Post-nominal demonstrative:
c.
DP
Dem
Pre-nominal demonstrative:
d.
DP
D’
D°[+TH]
Dem
NumP
Num°
N°
D°[+TH]
dxP
Dem
AP
dx’
dx°
D’
NumP
Num’
Num°
NP
N°
dxP
Dem
dx’
dx°
NP
N°
I take as given that NumP exists and is located between D°and NP; NumP has been
adopted in much of the literature on the syntax of DP since it was proposed by Ritter
(1991). NumP provides information about the morphological number of the DP—
singular or plural. Also, I assume that adjectives are base-generated in [Spec, NumP] in
78
Greek, following Stavrou (1999). Given these assumptions, in sections to follow, I will
discuss in more detail the first merge position of the demonstrative in the DP in section
3.2.1, and the movement and agreement within the DP in section 3.2.2.
3.2.1
The First Merge Position of Demonstratives
As discussed in Chapter 2, it is well-known that DNCs in Greek receive two different
interpretations—a deictic contrastive interpretation and an anaphoric interpretation
(putting aside the issue of the generic interpretation identified in Chapter 2 for the
moment). The examples are repeated below in (2):
(2)
Deictic contrastive DNC with pre-article demonstrative:
a.
Aftos
o
neos
andras
this
the young
man
‘This young man’
Anaphoric DNC with pre-article demonstrative:
b.
O
andras
aftos
the man
this
‘This man’
Anaphoric DNC with pre-article demonstrative:
c.
O
neos
aftos
andras
the young
this
man
‘This young man’
Greek
Also, it has been shown that each interpretation is associated with a different surface
position for the demonstrative within the DP. The deictic contrastive interpretation is
available when the demonstrative is in the pre-article position while the anaphoric
79
interpretation emerges when the demonstrative is in either the pre- or the post-nominal
position.
In order to account for the DP-internal word order and associated interpretations,
several analyses have been proposed in the ample literature on the Greek DP. Despite
slight differences in details, the general assumption is that the two word orders are in a
derivational relationship such that the pre-article position is a result of syntactic
movement applied to demonstratives generated low in the DP (see Horrocks and Stavrou
1987; Brugè1996, 2002; Brugèand Giusti 1996; Campbell 1996; Giusti 1997, 2002;
Panagiotidis 2000; Grohmann and Panagiotidis 2004, 2005; Roberts 2011; Guardiano
2012). The authors cited above do not, however, agree with the issue regarding where
demonstratives are introduced into the derivation of the DP. They suggest several
different ideas regarding the base-position of demonstratives; see (i), (iii), and (iv) below.
In addition to the idea that demonstratives are merged low in the structure, other
researchers (Cinque 2005; Abels and Neeleman 2009, 2012; Steddy and Samek-Lodovici
2011; Medeiros 2012) propose different analyses in order to capture all the attested DPinternal linear orders between demonstratives, numerals, adjectives and nouns and to rule
out all the unattested orders, based on the assumption that demonstratives are merged in
the structure as the highest element compared to the other three. To recap, we have by
and large four possibilities regarding the first-merge position of demonstratives:
i. The demonstrative is a complement of the noun (Horrocks and Stavrou 1987).
80
ii. The demonstrative merges as the highest modifier (Cinque 2005; Abels and
Neeleman 2009, 2012; Steddy and Samek-Lodovici 2011; Medeiros 2012).
iii. The demonstrative is the subject of the noun (Campbell 1996; Panagiotidis 2000;
Roberts 2011).
iv. The demonstrative is a low-base-generated adnominal element (Brugè1996, 2002;
Brugèand Giusti 1996; Giusti 1997, 2002; Vangsnes 1999; Rosen 2003;
Grohmann and Panagiotidis 2004, 2005; Guardiano 2012).
The rest of this sub-section will be devoted to a review of each standpoint on the first
merge position of demonstratives. I will argue against (i), (ii), and (iii) based on both
existing arguments and/or my own novel arguments. More specifically, I reject (i) based
on Panagiotidis’ (2000) argument, (ii) based on Roberts’ (2011) argument and also one of
my own, and (iii) based on my own arguments. Demonstrating the invalidity of the first
three options will leave us to adopt (iv) as a working hypothesis.
3.2.1.1 Demonstratives as the Complement of the Noun
The first possibility is Horrocks and Stavrou’s (1987) idea that the demonstrative is basegenerated as a complement of the noun; the demonstrative can either stay in situ as the
complement of the noun for the anaphoric interpretation, as in (3), or move to [Spec, DP]
for the deictic contrastive interpretation, as in (4).
81
(3)
(4)
Post-nominal demonstrative:
a.
O
andras
aftos
the man
this
‘This man’
b.
[DP [D’ o [NP [N’ andras [XP aftos]]]]]
Pre-article demonstrative:
a.
Aftos
o
andras
this
the man
‘This man’
b.
[DP [XP aftos]i [D’ o [NP [N’ andras ti]]]]
Greek
Greek
However, as pointed out by Panagiotidis (2000), the N°-complement base-position for the
demonstrative is problematic. It is not clear how the demonstrative can occupy the
position when the position is considered to be reserved for argumental genitive phrase
following the noun, as in (5) (from Panagiotidis 2000:720).
(5)
I
katiki
afti tis
the inhabitants these the.GEN
‘These inhabitants of the city’
polis
city.GEN
Greek
Also, if the demonstrative were indeed a complement of the noun, it ought to receive a
theta-role from the noun. However, what theta-role would be assigned by the noun to the
demonstrative remains mysterious.
82
3.2.1.2 Demonstratives as the Highest Modifier in DP
Cinque (2005) attempts to derive the pattern of attested and unattested relative orders
among four elements within DP—demonstratives, numerals, adjectives, and nouns—
across languages, assuming the correctness of Kayne’s (1994) Linear Correspondence
Axiom (LCA). His main idea is as follows. The base-generated position of the four
elements is universally consistent, as shown in (6) (Cinque 2005:321).2
(6)
[ … [WP Dem(onstrative) … [XP Num(ber) … [YP A(djective) [NP N]]]]]
Taking (6) as the universal first external merge order between the four elements, Cinque
proposes a set of movement parameters, as given in (7) (adapted from Cinque 2005:321),
to derive—by means of NP-movement or pied-piped NP-movement (i.e., moving a
phrase containing the NP) around the other elements—all the other attested and
unattested word order variations provided in (8) (adapted from Cinque 2005:319).
Since Cinque’s LCA-based derivational account for the (un)attested DP-internal word order facts, other
researchers (Abels and Neeleman 2009, 2012; Steddy and Samek-Lodovici 2011; Medeiros 2012), as
mentioned above, have taken different approaches to the same facts and proposed alternative analyses.
Though I will not further discuss the details of these alternatives, let me briefly discuss their main ideas.
(The interested reader is referred to Medeiros (2012: Ch. 5.1) for more discussion.) Abels and Neeleman
propose an account of the facts by appealing to head-directionality parameter rather than LCA. Steddy and
Samek-Lodovici propose within the framework of Optimality Theory that the word orders in question can
be ruled in or out by Align-Left constraint imposed on each nominal element under discussion. Medeiros
proposes that movement is motivated to reduce the number of c-commanding relations (i.e., to achieve
better balanced tree) and shows that his hypothesis correctly predicts the attested word order facts.
One crucial thing to note, putting aside the details of their analyses, is that all these alternatives as
well as Cinque’s (2005) initiative analysis neglect one empirical fact that the word order Num-A-Dem-N in
(8q), which is reported to be unattested, is in fact attested in Greek, as in (11b), to be discussed below in
this sub-section. Since all these analyses are designed to rule out this attested word order, they are all
problematic as they stand. Also see Gorrie (2014) for a critique of this type of parametric interpretation of
attested and unattested typological distributions.
2
83
(7)
Parameters of movement:
a.
No movement, or
b.
Movement of [NP [XP]], or
c.
Movement of NP without pied-piping, or
d.
Movement of [XP [NP]].
e.
Movement of NP all the way up, or just partially.
f.
Neither N°-movement nor phrasal movement not containing the NP is
possible.
(8)
Attested and unattested word orders:
a.
√
Dem Num A
N
b.
√
Dem Num N
A
c.
√
Dem N
Num A
d.
√
N
Dem Num A
e.
*
Num Dem A
N
f.
*
Num Dem N
A
g.
*
Num N
Dem A
h.
*
N
Num Dem A
i.
*
A
Dem Num N
j.
*
A
Dem N
Num
k.
√
A
N
Dem Num
l.
√
N
A
Dem Num
m.
n.
o.
p.
q.
r.
s.
t.
u.
v.
w.
x.
*
√
√
√
*
√
√
√
*
*
√
√
Dem
Dem
Dem
N
Num
Num
Num
N
A
A
A
N
A
A
N
Dem
A
A
N
Num
Num
Num
N
A
Num
N
A
A
Dem
N
A
A
Dem
N
Num
Num
N
Num
Num
Num
N
Dem
Dem
Dem
N
Dem
Dem
Dem
Let us take a look at how the various word orders are derived on Cinque’s account.
I will not delve into the derivation of the whole pattern given in (8), though. Instead, only
a few cases will be discussed, which will serve to introduce Cinque’s analysis. The
interested reader is referred to Cinque (2005:321-323) for a description of the entire
pattern (including the unattested word orders). (8a) is derived by moving nothing (cf.
84
(7a)), (8b) by moving the NP around A (cf. (7c)), (8r) by moving [Num A N] around
Dem (cf. (7d)), (8s) involves first moving the NP around A (cf. (7c)) and further moving
[Num N A] around Dem (cf. (7d)), and etc.
A potential problem for Cinque’s account is the word order in (8g). Though marked
as unattested, Cinque discusses in his fn. 26 that (8g) is in fact reported to be the general
word order in Kilivila, citing Senft (1986). Roberts (2011: fn. 6) also provides a list of
other languages that allow (8g) such as Iai, citing Tryon (1971b), Nengone, citing Tryon
(1967), Sundanese, citing Hardjadibrata 1985), Urak-Lawoi’, citing Hogan and Pattemore
(1988), and all Austronesian languages. Some of such examples are given in (9).
(9)
Kilivila (adapted from Senft 1986:69)
a.
Na-yu
vivila
mi-na-si-na
F-two
girls
this-F-PL-this
‘These two beautiful girls.’
Iai (adapted from Tryon 1971b:80)
b.
Kun ta
kuli aŋ aeso
three the dogs here good
‘These three good dogs.’
Nengone (adapted from Tryon 1967:57)
c.
Sa kore wa'i ɔm me wa'am
one the fish here
small
‘This one small fish.’
Dehu (adapted from Tryon 1971a:60)
d.
La ča
tusi čelæ m̥itøt
the one book there sacred
‘This one sacred book’
na-manabweta
F-beautiful
Cinque’s account based upon the assumption of the universal first-merge order in (6),
coupled with the movement parameters in (7), in fact fails to derive the word order in (8g)
(Num N Dem A). (8g) is made available only if we modify either the first-merge order or
85
the movement parameter (see Cinque 2005:322, (6g)). The latter is not a desirable
direction to take in Cinque’s system since such a move will have impacts on other
derivations as well. In order to account for this particular word order, Cinque (2005: fn.
26) makes a change to his assumption concerned with the universal first-merge order and
allows an exception for this type of language. The adjective in such languages is assumed
to be derived from a relative clause, which sits higher than Num. (8g) then involves
[Num N] movement around the adjective (or reduced relative clause) and Dem. Roberts
(2011:8) suggests (10) as his interpretation of Cinque’s (2005: fn. 26) derivation of (8g).
(10)
[DemP [NumP Num NP]i [Dem AP/RelativeClause ti]]
Even if we gloss over this exception that Cinque makes, his account brings to the
fore other problems. First of all, his account is immediately challenged by the DP-internal
word order facts of Greek. In this language, there are two possible combinations of
demonstrative, adjective, numeral and noun:
(11) Dem-Num-A-N:
a.
Afti i
tris nei katiki
these the three new inhabitants
‘These three new inhabitants’
Num-A-Dem-N:
b.
I
tris nei afti katiki
the three new these inhabitants
‘These three new inhabitants’
Greek
86
Recall that although the demonstrative can appear in the pre-article position, the postnominal position, or the pre-nominal position within DP, a post-nominal position for the
demonstrative is not possible when an adjective is present (as discussed in section 2.2).
Among the two word orders in (11), (11a) does not cause any problem for Cinque. The
word order in (11a) is one of the attested word orders (cf. (8a)). Crucial to our discussion
is the fact exemplified in (11b). Greek does allow Num-A-Dem-N word order in (11b),
which is reported to be unattested by Cinque (cf. (8q)) (and others including Greenberg
1963 and Hawkins 1983). Cinque’s account described above is designed to prevent all the
unattested word order from being derived, and thus (8q) (i.e., (11b)) cannot be derived on
his account. On Cinque’s analysis, the fact that the word order in (8q) is different from
the first-merge order he assumes in (6) suggests the involvement of a movement
operation in its derivation. Since all the movement must involve the movement of NP, as
proposed in (7), it is in principle impossible to tweak the word order of the pre-nominal
elements only, without moving NP.
Modifying the movement parameters in (7) and/or the universal first-merge order in
(6) in order cannot save us from the problem since such an approach would have effects
on the other derivations, resulting in unattested ones being predicted to be attested and
vice versa. We could nonetheless avoid this problem by simply assuming some special
rules for Greek. However, such an approach is far from desirable, considering the the
whole purpose of Cinque’s theory of DP-internal word order.
Second, Roberts (2011) argues that Cinque’s analysis is problematic on
independent grounds. Roberts (2011:8) points out that the derivation in (10) questionable
87
in light of the Final-over-Final Constraint (FOFC). Specifically, the structure in (10), he
argues, violates the FOFC, which is argued to be a universal constraint imposed upon
syntactic hierarchical structure, as formulated in (12) (from Biberauer et al. 2010:3).3
(12) The Final-over-Final Constraint:
If α is a head-initial phrase and β is a phrase immediately dominating α, then β
must be head-initial. If α is a head-final phrase, and β is a phrase immediately
dominating α, then β can be head-initial or head-final.
The FOFC, as defined in (12), allows for the structures represented in (13a), (13b), and
(13c), though disallows that in (13d).
(13) a.
Consistent head-final:
βP
αP
γP
β
α
b.
Consistent head-initial:
βP
β
αP
α
γP
The FOFC was originally suggested by Holmberg (2000:124). The FOFC was motivated based upon the
fact that the structure in (13d) is not attested—whilst those in (13a-c) are found—in various syntactic
domains in various languages. Here is the list of the the FOFC violating cases and the relevant references:
3
(i)
a.
b.
c.
d.
e.
f.
*[[V O] Aux]
*[[Pol TP] C]
*[[V O] C]
*[[C TP] V]
*[[Asp VP] T]
*[[N O] P]
(see Biberauer et al. 2007, 2008)
(see Biberauer et al. 2010)
(see Biberauer et al. 2007, 2008; Hawkins 1994; Kayne 1994)
(see Hawkins 1994; Dryer 2009)
(see Julien 2007)
(see Holmberg 2000)
Since the full discussion of the FOFC is beyond the scope of this thesis, I will not enter into a detailed
discussion. The interested reader is referred to the above literature.
88
c.
Initial-over-Final:
βP
β
d.
*Final-over-Initial:
*
βP
αP
γP
αP
α
α
β
γP
Roberts points out that the structure in (10), reproduced as a shorthand version in (14)
below, resembles the banned structure in (13d), violating the FOFC. I abstract away from
the adjective/relative clause and the movement of NumP in order to make the point clear.
Between (13d) and (14), α corresponds to Num, β to Dem, and γ to N.
(14) [DemP [NumP Num NP] Dem]
(13d) *[βP [αP
α
γP]
β]
Assuming the FOFC to hold, (14) must be ruled out. This suggests that Cinque’s
derivation of (10), as formulated by Roberts (2011), cannot be well-formed. In order to
get around this issue, Roberts proposes to treat the demonstrative as the subject of the
nominal predicate and derive the other word order variations by counting on N°or NP
movement (though he does not explore the implications of such N(P)-movement). In the
following section, however, Roberts’ treatment of the demonstrative as the subject will be
argued to be problematic as well.
Additionally, I would like to remark on the status of what is generally taken to be
the complement of a noun on Cinque’s system. His account encounters an empirical
challenge when it comes to the surface order of DNCs whose noun head takes a PP or DP
89
complement. Let us consider some relevant Greek examples in (15) (adapted from
Panagiotidis 2000:720).
(15) a.
b.
I
katiki
tis
polis
the inhabitants the.GEN
city.GEN
‘The inhabitants of the city.’
I
katiki
afti tu
polis
the inhabitants these the.GEN
city.GEN
‘These inhabitants of the city.’
Greek
When there is a post-nominal demonstrative present, its distribution is limited to the
immediate right of the noun, splitting the noun and its complement. The demonstrative
cannot appear to the right of the complement DP, as confirmed in (16).
(16) *I
the
katiki
tu
inhabitants the.GEN
polis
city.GEN
afti
these
Greek
The facts illustrated above are problematic for Cinque’s account. All the movements are
phrasal and involve the NP in his system. This makes a strong prediction that the noun
and its complement should be adjacent to each other at all stages of the derivation if we
make the orthodox assumption that the complement of the noun merges within NP as a
sister. As already shown above, however, the prediction is not fulfilled; the demonstrative
intervenes between the two elements (cf. (15b) and (16)).
In order to get around this problem, Cinque adopts Kayne’s analysis of Noun-ofNoun constructions (Kayne 2000b, 2002, 2004). Kayne (2000a) proposes (18b) as the
90
derivation of (17a); (17a) is derived by moving friends to the specifier of of (from Kayne
2000b:314).
(17) a.
b.
friends of John’s
friendsi [of [John’s ti]]
He extends the above analysis to another instance of Noun-of-Noun. On his account, Bill
was admiring a picture of John is derived, as shown in (18) (from Kayne 2000b:316).
(18) … admiring [John a picture] → merger of of
… of admiring [John a picture] → attraction of Spec, of
… Johni of admiring [ti a picture] → merger of W and raising of of
… ofj+W Johni tj [ti a picture] → movement to Spec, W
… [admiring [ti a picture]]k ofj+W Johni tj tk
Cinque (2005: fn. 34) briefly discusses how Kayne’s analysis can account for the
stranded complement of the noun. If my understanding is correct, the Italian example of
the PP complement stranding construction in (19) (from Cinque 1994:86) would be
derived as illustrated in (20).
(19) La invasione italiana
dell’Albania
the invasion Italian
of.the Albania
‘The Italian invasion of Albania.’
(20) a.
b.
c.
[la italiana invasione] [l’Albania] → movement of NP
[la invasionei italiana ti] [l’Albania] → merger of di
di [la invasionei italiana ti] [l’Albania] → attraction of Spec, di
Italian
Italian
91
d.
e.
f.
l’Albaniaj di [la invasionei italiana ti] tj → merger of W and raising of di
dik+W l’Albaniaj tk [la invasionei italiana ti] tj → movement to Spec, W
[la invasionei italiana ti]l dik+W l’Albaniaj tk tl tj
The crucial point in (20) is that the preposition is base-generated outside the DP and that
the head noun and the prepositional phrase never form a single constituent (see (20f)).
Although adopting Kayne’s idea can derive the correct word order (i.e., the
separation of the noun and its complement), as shown in (20), it seems that such an
approach is not free from difficulty. One obvious problem comes from the constituenthood of the phrase, a picture of John. On Kayne’s account, a picture of John does not
form a constituent. This is at odds with the common assumption that the phrase in
question is indeed a constituent. The results of typical constituency tests (Carnie 2007:8891) show that the phrase in question passes each: replacement, stand alone, movement,
and coordination test. A picture of John can be pronominalized, as in (21a); it can stand
alone, as in (21b); it can be clefted, preposed, and passivized, as in (21c)-(21e); it can be
coordinated, as in (21f).
(21) a.
b.
c.
d.
e.
f.
Bill was admiring [it].
[A picture of John]. (as an answer to What was Bill admiring?)
It was [a picture of John] that Bill was admiring.
[A picture of John] was what Bill was admiring.
[A picture of John] was being admired by Bill.
Bill was admiring [a picture of John] and [the painter].
Given the fact that the claimed non-constituent is in fact a constituent, Kayne’s analysis
described above seems to be compromised. If this is so, Cinque’s justification for the
92
stranding of the complement of the noun, which is based on the untenable analysis
discussed above, seems to be invalid.
3.2.1.3 Demonstratives as the Subject of the Nominal Predicate
Another possible analysis for the base position of the demonstrative is to treat the
demonstrative as the subject of the nominal predicate, and accordingly to base-generate
the demonstrative in [Spec, NP] (Panagiotidis 2000) or [Spec, nP] (Roberts 2011) (see
also Campbell 1996, 1998). I concentrate on Roberts (2011) and show that the
demonstrative cannot be treated as the external argument of the noun.4
Roberts (2011) argues that the demonstrative is the external argument of the
nominal predicate and thus must be merged in [Spec, nP], assuming nP to be the nominal
counterpart of vP whose specifier position is generally assumed to introduce the subject
of the clause. His argument is based on Williams’ (1980) notion of the external argument,
according to which the external argument determines the reference of the predicate. For
instance, a nominal predicate dog denotes a property of being a dog and takes one
argument x, as illustrated in (22).
Panagiotidis (2000), as mentioned above, argues against Horrocks and Stavrou’s (1987) idea that the
demonstrative is the complement of the noun, and also criticizes the postulation of a special functional
projection whose sole function is to introduce the demonstrative, the idea pursued by Brugè (1996, 2002)
and Giusti (1997, 2002). Panagiotidis, for the sake of presentation, takes the assumption as a working
hypothesis that the demonstrative is introduced in the derivation by merging in [Spec, NP]. As Panagiotidis
(2000:728 and fn. 12) admits, it is not clear if the nominal predicate assigns a theta-role to the
demonstrative subject, and even if so, what kind of theta-role is assigned to the demonstrative.
4
93
(22) dog (x)
This predicate can be predicated of, say, Fido, which fills in the argument position x, as
illustrated in (23).
(23) a.
b.
dog (Fido)
Fido is a dog.
On Roberts’ account, it is then the demonstrative that can “directly establish the reference
of NP without the intermediary of a propositional function” (Roberts 2011:21). In this
connection, it is concluded that in DNCs the demonstrative functions as the external
argument of the nominal predicate, as illustrated in (24).
(24) a.
b.
dog (this)
This dog
Roberts argues that treating the demonstrative as the external argument of a
nominal predicate leads, in an interesting way, to a parallelism between DP and CP in the
sense that the distribution of the subject in the clausal domain patterns together with the
DP-internal distribution of the demonstrative—the subject in the nominal domain.5 For
instance, the demonstrative appears before the noun in English, which is a SV(O)
language, while the demonstrative appear after the noun in Celtic languages, which are a
5
As Roberts himself notes, there are exceptions to this pattern.
94
VS(O). Roberts assumes parallelism between DP and CP, as illustrated in (25) (from
Roberts 2011:11).
(25) Clause structure:
a.
[CP C° [TP T° [vP
Nominal structure:
b.
[DP D° [NumP Num°[nP
v°
[VP
V°]]]]
n°
[NP
N°]]]]
Given his argument that the demonstrative is the external argument of the nominal
predicate, he identifies [Spec, nP] as the base-position of the demonstrative, as
schematized in (26). Recall that [Spec, vP] is generally assumed to introduce the external
argument of the verb in the clause.
(26) [DP D°[NumP Num°[nP Dem [n°[NP N°]]]]]
I argue, however, that the predication analysis of DNCs—that is, the idea that the
demonstrative is the subject of the nominal predicate—cannot be maintained. The
reasoning goes as follows. If the demonstrative is an external argument of the nominal
predicate in the sense that it determines the reference of the nominal predicate, the
prediction is that we ought to be able to construct DNCs out of the nominal predicate and
the subject of any expression, as long as the latter can determine the reference of the
former. This is indeed the case in the clausal copular construction, which is, whatever
syntactic analysis one adopts for copular constructions, uncontroversially an instance of
predication, as shown in (27).
95
(27) a.
b.
This is a prize.
This book is a prize.
Both the simplex demonstrative (e.g., this) and the DNC (e.g,. this book) can serve
properly as the external argument of the nominal predicate prize in the context of copular
constructions, as schematized in (28).
(28) a.
b.
prize (this)
prize (this book)
Such a possibility does not emerge in the purportedly parallel predication within
the nominal domain, i.e., in the DNC. A nominal predicate can only occur with a simplex
demonstrative, but not with a phrasal DP subject. In other words, the nominal predicate
of DNCs does not tolerate a DNC (or any other complex DP) as its subject. This point is
illustrated in English, as in (29) and (30)—the simplex demonstrative serve as the
purported external argument of the nominal predicate while its complex counterpart—a
DNC—cannot.
(29) a.
b.
[[This]SUBJ [prize]PRED] is for the winner.
*[[This book]SUBJ [prize]PRED] is for the winner.6
(30) a.
b.
prize (this)
*prize (this book)
In contrast with [this [book-prize]], in which book and prize form some sort of a compound (e.g., cashprize).
6
96
On Roberts’ account, there should be in principle no reason why the referencedetermination function carried out by a simplex demonstrative such as this in (29a)
cannot be accomplished by a DNC such as this book in (29b). Note that in the copular
sentence counterparts in (27), both are allowed as the external argument. Roberts’
account as it stands cannot rule out the possibility that the nominal predicate could take a
DNC external argument, and therefore his account is untenable.
When it comes to Greek, however, a complication arises: the language does appear
to allow for the construction banned in English. That is, on the assumption that DNCs are
instances of DP-internal predication, a nominal predicate can be predicated of a DNC in
Greek. As in English, the phrasal or simplex nature of the subject does not matter in
copular constructions, illustrated in (31). Unlike in English, the nominal predicate within
a DNC can itself take a DNC as its subject, as in (32b)7.
One might take (32b) to be an instance of close apposition such as (i) (Lekakou and Szendrői 2012:108),
in the sense that two DPs are contained within a larger DP in (32b).
7
(i)
a.
b.
O
aetos
the.M
eagle.M
‘The eagle that is a bird.’
To
puli
the.NEUT
bird.NEUT
to
the.NEUT
puli
bird.NEUT
o
the.M
aetos
eagle.M
Greek
It seems, however, that (32b) and (i) are two different constructions for two reasons. First, as shown in (i),
close apposition allows for free word order between the two sub-DPs. Yet, switching the order between the
DNC and the rest of the DP yields ungrammaticality, as in (ii).
(ii)
*I
the
daskala
teacher
afti
this
i
the
kyria
lady
Greek
Second, close apposition does not necessarily require a gender match between the two DPs it
features. Note that in (i), one DP o aetos ‘the eagle’ is masculine while the other to puli ‘the bird’ is neutral
in gender. This is not the case in DNCs. As exemplified in (iii), gender mismatch is not tolerated in DNCs.
(iii)
a.
Afto/*Aftos to
agori
this.N/this.M the.N boy.N
‘This boy’
Greek
97
(31) a.
b.
(32) a.
b.
Afto einai
this be.3SG.PRES
‘This is a waiter.’
Afto to
agori
this the boy
‘This boy is a waiter.
garsoni.
waiter
Greek
einai
be.3SG.PRES
[[Afto]SUBJ to
[garsoni]PRED]
this
the waiter
‘This waiter is careful.’
[[Afto
to
agori]SUBJ to
this
the boy
the
garsoni.
waiter
einai
be.3SG.PRES
prosektiko.
careful
[garsoni]PRED]
waiter
einai
be.3SG.PRES
Greek
prosektiko.
careful
‘The waiter who is this boy is careful.’
The varying pattern with respect to whether or not DNCs can embed another DNC
in English and Greek does not lead to a consistent conclusion that the demonstrative is
the subject of the nominal predicate within the DNC as in Roberts’ predication analysis.
If the demonstrative is, as argued by Roberts (2011), the subject of the nominal predicate
within the DNC, we would not be able to understand why a DNC cannot saturate the
open argument slot of the nominal predicate in English. Since Roberts’ analysis predicts
the [[Demonstrative-Noun]-Noun] construction to be grammatical in English, contrary to
facts (cf. (29b) above), I still conclude that his analysis cannot be maintained as is; he
b.
c.
Afto/*Aftos
this.N/this.M
‘This waiter’
*Afti i
this.F the.F
to
garsoni
the.N waiter.N
kyria to
garsoni
lady.F the.N waiter.N
If this is so, the modifying DNC embedded within the larger DNC contains two definite articles only
because DNCs in Greek obligatorily need a definite article.
98
needs to adjust it to rule out the ungrammatical [[Demonstrative-Noun]-Noun] case in
English.
A question then arises: whence stems this difference between English and Greek
with respect to the occurrence of a DNC within a DP? My tentative answer to this
question is that languages may differ with respect to whether or not they allow complex
adnominal modifiers. Specifically, the restriction is concerned with whether or not the
head of an adnominal modifier can take a complement. The clear-cut examples that show
such a difference between Greek and English come from the realm of adjectives. As
illustrated by the contrasting grammaticality in (33) ((33b) is adapted from Panagiotidis
and Marinis 2011:291), it is well known that prenominal adjectives cannot take a
complement in English8, while they can in Greek.
(33) a.
b.
*A [proud of her son] mother
I
[perifani gia ta
pedia
the proud
of
the children
‘The mother who is proud her son’
tis]
her
mitera
mother
Greek
I argue that the pattern observed in (33) obtains for another adnominal modifier—
DNCs embedded in a larger DP. Whether or not a DNC can function as an adnominal
modifier in the two languages is thus subject to the configuration of the modifier itself in
analogy to the varying size of the pre-nominal adjective in (33). If we assume that all
demonstratives are phrasal in both languages (following Campbell 1996, among many
others), the DNC is headed by D°(obligatorily overt in Greek and obligatorily covert in
8
Thanks to Andrew Carnie (personal communication) for pointing out this English fact to me.
99
English).9 In this case the head of the DNC takes a functional phrase (e.g., NumP) as a
complement, which is overt by virtue of the fact that NumP encases an overt noun. Hence,
if we assume that the size of the adnominal modifiers is subject to a certain restriction,
which is presumably parameterized to the effect that the difference seen in (33) between
English and Greek falls out, it seems reasonable to assume the same restriction for the
size of a DNC that can be embedded within a DP (at least) in the two languages under
consideration. And I tentatively suggest that the restriction governs the overt size of
adnominal modifiers.10 11
Before delving into my next argument, which is based on another size restriction
suggested by Den Dikken (1998, 2006), it is worthwhile discussing demonstrativereinforcer constructions. As has been shown in Chapter 2, demonstratives can be
collocated with a reinforcer in Greek and non-Standard English. The examples are
repeated below for convenience:
(34) a.
b.
[These here] linguists
[Those there] linguists
This argument does not hinge on one’s perspective regarding the status of the demonstrative in languages
like Standard English. That is, even if the demonstrative is considered to be the head of the DNC, this fact
does not change the point that the DNC takes its own complement, whether it is a NumP or some other
phrase. Hence, it does not affect the present argument.
10
It seems that this size restriction must be limited to the pre-nominal adnominal modifier. For instance, a
PP modifier, in which it is obviously the case that the P head takes an overt complement, can modify the
head noun, appearing to the right to the noun.
11
The reason why the condition is concerned with the “overt” size of an adnominal modifier is that simplex
demonstratives and pronouns are treated as syntactically complex. That is, they are argued to be fullyfledged DPs with a null noun (see Chapter 4).
9
100
(35) a.
b.
[Afti edho]
i
these here
the
‘These here linguists.’
[Ekini
eki] i
those
there the
‘Those there linguists.’
glossologi
linguists
Greek
glossologi
linguists
It seems that the fact that DNCs can contain a demonstrative-reinforcer construction in
English may be taken as an exception to the tentative generalization that the head of
adnominal modifiers cannot take an overt complement in English. Herein, I argue that if
we adopt Choi’s (2013) analysis of demonstrative-reinforcer constructions, this problem
goes away and these seemingly complex constructions involving a demonstrative do not
violate the above generalization. Choi (2013) argues that reinforcers involve DPadjunction both in English and Greek. If this is right, then the reinforcer is not an overt
complement of the D head, and thus its presence does not violate the size restriction
imposed on adnominal modifiers in English. This line of reasoning is bolstered by the
fact that adjectives can be modified by some adverbs such as very as in a very* proud
mother12. This fact clearly shows that the size restriction is not sensitive to the mere size
of the surface string, but instead to a specific syntactic configuration, i.e., whether or not
the head of the adnominal modifier takes an overt complement. For this reason, the
presence of demonstrative-reinforcer constructions is not a problem for us.
What follows below is my last argument against treating the demonstrative as the
subject of the nominal predicate, under the assumption of another size restriction
suggested by Den Dikken (1998, 2006). He shows, on the basis of the behavior of
12
* indicates that the adverb can be stacked as many as necessary.
101
qualitative binominal noun phrases (cf. (36a)), that the subject and the predicate of a
small clause within a DP cannot be larger than NumP. On his account, what he calls the
attributive qualitative binominal noun phrase in (36a) is headed by the nominal copula of,
which hosts the subject and the predicate in its complement and specifier, respectively.13
(36a) accords with the size restriction since the size of the subject (a doctor) and the
predicate (an idiot) does not exceed NumP, and thus is well-formed. On the contrary,
(36b) is ill-formed since the subject (the/that/my doctor) is larger than NumP, violating
the size restriction.
(36) a.
b.
An idiot of a doctor
[DP D°[RELATOR PHRASE [NumP an idiot] RELATOR=of [NumP a doctor]]
*That idiot of the/that/my doctor
[DP that [RELATOR PHRASE [NumP idiot] [RELATOR=of [DP the/that/my doctor]]]]
Now we are in a position to investigate the effect of the size restriction described
above on Roberts’ predication analysis. In a nutshell, the size restriction
straightforwardly rules out DNCs as cases of DP-internal predication. The reasoning is as
follows. Let us suppose that the simplex demonstrative is a DP, as has been assumed thus
In Den Dikken’s (2006) theory of predication, the function of the RELATOR (i.e., the head of the
predication) is to accommodate its arguments locally and non-directionally. As for locality, the arguments
of the RELATOR sit in the specifier or the complement position of the RELATOR. As for directionality, there is
no association between a certain argument and a certain position; the subject can occupy either the specifier
or the complement position and so can the predicate. As a result, two basic structures are available, as in (i)
(from Den Dikken 2006:13). The structure under consideration is an instance of (ib).
13
(i)
a.
b.
[RELATOR PHRASE [XP SUBJECT] [R’ RELATOR [YP PREDICATE]]
[RELATOR PHRASE [XP PREDICATE] [R’ RELATOR [YP SUBJECT]]
Den Dikken argues that (36a) is in fact ambiguous between the structure in (ia) and what he calls ‘reverse
predication’. Since it is immaterial to our discussion, I will not further address the issue.
102
far (see Brugè1996, 2002; Campbell 1996; Giusti 1997, 2002; Vangsnes 1999;
Panagiotidis 2000; Rosen 2003; Grohmann and Panagiotidis 2004, 2005; Guardiano 2012;
among others). The status of the demonstrative as DP, coupled with the assumption of the
DP-internal predication analysis of DNCs, leads to the violation of the size restriction
since the demonstrative (i.e., the claimed subject of the nominal predicate) is a DP,
embedded within another DP (i.e., the DNC). If this were the case, any DNC (e.g.,
[[this]SUBJ [prize]PRED]) would contain a DP-subject (i.e., a demonstrative), thus resulting
in a violation of Den Dikken’s size restriction. All DNCs would then be expected to be
ill-formed, if we assume that they are instances of DP-internal predication. Therefore, the
fact that DNCs are grammatical suggest that they are not the cases of DP-internal
predication. 14, 15
This argument based on the English case can be compromised if one develops an
analysis according to which demonstratives are not DPs but something smaller than DP,
or as argued by Jacckendoff (1977) and Bernstein (1997), occupy the D head of the DNC
at least in Standard English. Even if either case is true, we can still fend off such an
The violation of the size restriction can be avoided if one assumes with Leu (2008) that a demonstrative
in Germanic languages (focusing on Swiss German) is a sort of an adjective, more specifically, an extended
adjectival projection rather than a DP. This weakens my argument based on the assumption that the
demonstrative projects a DP. However, it is not clear how far such an analysis can be extended. Rosen
(2003) argues that demonstratives cannot be treated as adjectives on the basis of different behaviors
between demonstratives and adjectives in Michif, the language historically derived from Cree and French.
15
Bernstein observes that a group of words in English (such as the, this, that, these, those, etc) share the thmorpheme and all of them encode third person. This observation leads her to associate the th- morpheme
with third person and the D head. On her account, any type of phrase involving a demonstrative in English
(e.g., this and this book) is headed by the th- morpheme sitting in the head D. I, however, adopt her stance
only partially. That is, I only adopt her idea that the th- morpheme, encoding third person, is associated with
the D head.
Associating person with the D head straightforwardly leads to the hypothesis that all the third person
elements starting with the th- in English project a DP. We can then conclude that the demonstrative, being a
third person element, should be as large as DP. Consequently, we can nullify a potential counter-argument
which relies on the idea that the demonstrative is smaller than NumP and thus DNCs does not violate Den
Dikken’s size restriction.
14
103
argument based on the fact that such an approach cannot be extended to other languages
such as Greek or even other dialects of the same language. Here is why. Above all, in
both Greek and non-Standard English the demonstrative can be collocated with a
reinforcer, and this fact suggests that the demonstrative can form a phrase with a
reinforcer instead of heading the DNC (see for more details sections 2.2.1.2.1 and 4.2.6).
Even if one could somehow show that a demonstrative and a reinforcer do not form a
phrase and the demonstrative is the D head of DNC, the fact that Greek allows a DNC to
be embedded within another DNC (see the discussion of (32b) above) still remains
problematic. The reason is that the DNC embedded within another one is undoubtedly a
DP and thus the construction must be ruled out by Den Dikken’s size restriction; however,
such DNCs are well-formed in the language. Therefore, if we assume that Den Dikken’s
size restriction holds in Greek as well, the fact that DNCs encasing another DNC are
grammatical strongly suggests that DNCs cannot be an instantiation of DP-internal
predication since such DNCs are predicted to be ungrammatical by the size restriction.
Before wrapping up this sub-section concerned with the issue of whether or not the
DNC is an instance of DP-internal predication, a further but brief discussion of the
possibility of application of other theories of “predication” is in order. As discussed
above, I reviewed Roberts’ (2011) predication analysis of DNCs, which basically adopts
Williams’ (1980) idea about predication, and discussed the relevant issue in terms of Den
Dikken’s view of predication. In addition to Williams (1980) and Den Dikken (2006),
there are many other approaches to the issue of what “predication” is and how its
argument position is saturated, including among many others Higgins (1973), Rothstein
104
(1983, 2001), Stowell (1983, 1991), Williams (1983a, 1983b, 1994), Higginbotham
(1987), Safir (1987), Speas (1990), Bowers (1993), Heycock (1994), Stroik (1994),
among many others. Whether “predication” is syntactic or semantic, the problematic
English fact that a DNC cannot embed another DNC remains unexplained. It would take
us too far afield, however, to test in light of all these analyses whether or not DNCs can
be treated as an instance of predication in which the demonstrative is the subject and the
noun is the predicate. I would like to pick one of the above analyses—Stroik (1994)—to
briefly show that assuming a DNC to be an instance of predication does not predict that
the nominal predicate is predicated of the demonstrative.
Stroik (1994) proposes that nominal predicates are well-formed if their lexical and
extended projections (DP, NumP and NP) are properly saturated by an argument which is
in an m-command relationship with the predicate. He shows that NumP and DP can also
function as a predicate and further proposes that the head noun bears the [predication]
feature, which is shared with the higher functional projections such as NumP (via N°-toNum°movement) and/or DP (via syntactic agreement). As a result of saturation and
[predication]-feature-sharing, NumP or DP can also serve as a predicate. For instance, in
(37) (adapted from Stroik 1994:57), the DP my student is demonstrably the predicate of
the small clause; according to Stroik’s analysis, this is possible for two reasons. First, N°to-Num°raising enables the [predication] feature on the noun to be shared with the
NumP. This feature is further shared with the DP via the syntactic agreement between D°
and the NumP. Second, the NumP and the DP with the [predication] feature are
respectively saturated by PRO in [Spec, NumP] and my in [Spec, DP] (PRO is
105
independently motivated (see Stroik 1994:48)). Since all the predicative projections
contained in the DP are properly saturated and the [predication] feature on the noun is
shared, the DP in question is a well-formed predicate. This predicate is then saturated by
her, the subject of the small clause, as in (37a).
(37) a.
b.
I consider [SmallClause [her]k [DP my student]k]
[DP [SPEC my] D°[NumP [SPEC PRO] studentk+Num°[NP tk]]]
DNCs—containing either a simplex demonstrative or another DNC—are, on
Stroik’s account, predicted to be a well-formed predicate, assuming that demonstratives
are in [Spec, DP]. For instance, (29b), repeated in (38), should be able to serve as a
predicate since the [predication] feature on the noun can be shared with the NumP (via
N°-to-Num°raising) and DP (via D°-NumP agreement), and the NP/NumP and DP are
respectively saturated by PRO and this book, in exactly the same way as (37b).
(38) *[DP [SPEC this book] D°[NumP [SPEC PRO] prizek+Num°[NP t k]]]
As we have discussed above (cf. (29)), however, (38) is ungrammatical irrespective of its
argument/predicate status, and this fact would remain unexplained if we considered
DNCs to be predicates within Stroik’s analysis of predications.
It is obvious from the discussion in this sub-section that the demonstrative should
not be treated as the subject of a nominal predicate encased within the same DP.
Specifically, Roberts’ (2011) predication analysis of DNCs cannot be maintained based
106
upon the fact that English DNCs cannot contain another DNC as the subject, which is in
fact predicted to be allowed on Roberts’ account. I have additionally shown that other
theories of predication cannot be imported to account for the properties of DNCs by the
same token.
3.2.1.4 Demonstratives in the Specifier of the Extended Nominal Projection
We have seen so far that the demonstrative cannot be treated as the complement of the
noun, as the highest base-generated element in a DP, or as the subject (or external
argument) of a nominal predicate within DP. Excluding these three views leaves us only
one option—the idea that the demonstrative first merges in the specifier of an extended
nominal projection, higher than NP or nP (i.e., outside the lexical projection) but lower
than D°. This idea is not novel, and has been pursued in much of the literature (Brugè
1996, 2002; Brugèand Giusti 1996; Giusti 1997, 2002; Vangsnes 1999; Rosen 2003;
Grohmann and Panagiotidis 2004, 2005; Guardiano 2012). I take this idea a working
hypothesis throughout this thesis. As mentioned in section 3.2, I assume NumP (Ritter
1991) to be located between D°and NP, and reserve its specifier position, [Spec, NumP]
for adjectives (following Stavrou 1999). That being said, the first merge order of DPinternal elements in Greek is illustrated in (39) (cf. Guardiano’s (2012) universal
underlying DP-internal word order: D > Num > APs > Dem > NP).
107
(39) D > APs > Num > Dem > NP
Following Vangsnes (1999), I label the functional head that introduces
demonstratives (as well as pronouns, as will be discussed in Chapter 4) in its specifier as
dx°, for ‘deixis’. Rosen (2003) calls the functional head that host demonstratives in its
specifier position Dem°. However, the label ‘dx°’ will embody the spirit of the analysis
here better than Rosen’s Dem°label because we are not only dealing with demonstratives
but also with pronouns, and both are deictic elements. In this connection, and to
accommodate the conclusion of Chapter 2 that DNCs and PNCs form a natural class, I
follow Vangsnes’ terminology. Based on the discussion so far, I assume (40) as the base
structure of DNCs:16
(40)
DP
D°
NumP
(AP)
Num’
Num°
dxP
Dem
dx’
dx°
NP
One can posit the existence of nP, between dxP and NP in (40), whose function is to introduce an external
argument or to turn a root into a noun in the sense of Distributed Morphology (Halle and Marantz 1993).
Since its presence in the description of the structure of DNCs (as well as PNCs) is immaterial to our
discussion, I will not include its representation in the structure throughout.
16
108
As we already know, the base structure in (40) does not always correspond to the
surface word order of DNCs. Also, it is well-known that the DP-internal elements in
Greek agree with each other. The next sub-section will be devoted to these issues.
3.2.2
Movement and Agreement within Demonstrative-Noun Constructions
In this section, I will discuss the syntactic derivations which yield DP-internal word
orders in DNCs in Greek. I first discuss how the movement of the demonstrative and the
noun operates in section 3.2.2.1, and how agreement works in section 3.2.2.2.
3.2.2.1 Movement
In this section, I will discuss two types of movement that I employ in order to derive the
linear word orders of DNCs in Greek. In so doing, I follow Panagiotidis’ (2000) analysis.
I first discuss head-movement and then phrasal movement.
Panagiotidis (2000) proposes that the post-nominal demonstrative position is
derived by leaving the demonstrative in-situ while N°raises to Num°(see also Giusti
1997, 2002). Let us discuss the details. As has been mentioned in section 3.2.1.2,
Panagiotidis observes that the presence of an adjective in DNCs has an impact on the
distribution of the demonstrative. Consider (41) and (42) (taken and adapted from
Panagiotidis 2000:728, 730).
109
(41) DNC without an adjective:
a.
I
katiki
afti tis
the inhabitants these the.GEN
‘These inhabitants of the city’
b.
*I
afti katiki
tis
the these inhabitants the.GEN
polis
city.GEN
Greek
polis
city.GEN
(42) DNC with an adjective:
a.
I
nei afti katiki
tis
the new these inhabitants the.GEN
‘These new inhabitants of the city’
b.
*I
nei katiki
afti tis
the new inhabitants these the.GEN
polis
city.GEN
Greek
polis
city.GEN
Compare (41a) and (41b): when there is no adjective present within the DNC, the noun
must precede the demonstrative. As shown in (42), unlike (41), when an adjective appears
within the DNC, it has to be the case instead that the demonstrative precedes the noun.17
As discussed in section 3.2.1.3, Panagiotidis (2000) assumes that the demonstrative
is base-generated in [Spec, NP], and following Stavrou (1999) adjectives are merged in
[Spec, NumP]. (43) represents the base structure of DNCs suggested by Panagiotidis:
Recall that Panagiotidis base-generates the demonstrative in [Spec, NP]; though we reject this idea and
assume [Spec, dxP] as the first merge position of the demonstrative, there is no difference in terms of the
linear order of the DP-internal elements.
17
110
(43)
DP
D°
NumP
AP
Num’
Num°
NP
Dem
N’
N°
XP
He assumes that head-movement is triggered by category feature checking. In the case of
DNCs, N°-raising is triggered to check the [N] category feature. He assumes that Num°
bears a strong [N] feature, which can be checked off in two ways. One is by N°-raising to
Num°yielding [[N°-Num°]-Num°] head adjunction. The other is to fill in [Spec, NumP]
with an adjective. Panagiotidis’ derivations of (41a) and (42a) are illustrated in (44).
(44) Post-nominal demonstrative:
a.
[DP ID°[NumP [Num’ katikiN°+Num°[NP aftiDEM [N’ tN°[DP tis polis]]]]]]
Greek
Pre-nominal demonstrative:
b.
[DP ID°[NumP neiAP [Num’ Num°[NP aftiDEM [N’ katikiN°[DP tis polis]]]]]]
On his approach, it follows that the noun must raise in the absence of an adjective, as in
(44a), while N°-raising is bled by the presence of an adjective, as in (44b). He attributes
the reason for the competition between the merger of an adjective and N°-to-Num°
raising to the assumption that the former pre-empts the latter (cf. Chomsky 1995; Mergeover-Move).
111
Before proceeding to the discussion of phrasal movement, there are two theoretical
issues that deserve a mention. First of all, how N°-to-Num°raising can be blocked by
merging an AP to [Spec, NumP] is not clearly stated. This bleeding relationship is
interesting in that head-movement is bled by external merger of a phrase into the specifier.
I suggest that the idea that there is a competition between the head-movement and the
phrasal merger can be couched in phrasal movement approaches to head-movement such
as Matushansky (2006). Matushansky proposes that head-movement is not a single
operation, but an operation that has two components: syntactic movement and
morphological merger. On the assumption of her approach to head-movement, the target
head of the movement (Y°) first moves to the specifier of the movement-triggering head
(X°) in syntax, as illustrated in (45a). At morphology, the two adjacent heads (X°and Y°)
undergo morphological merger, as shown in (45b).
(45) Syntactic movement:
a.
XP
Morphological merger:
b.
XP
Y°+X°
Y°
X°
YP
YP
ZP
ZP
tY°
tY°
WP
WP
If we assume Matushansky’s approach to head-movement, the reason why N°-to-Num°
raising competes with external merger of AP to [Spec, NumP] is because the landing site
of N°raising is [Spec, NumP]. Either merger of the N°or of an AP in [Spec, NumP]
could check the [N] feature. When both movement and external merger are available
112
options for structure-building at some point in a derivation, the latter is chosen over the
former by the Merge-over-Move preference; hence adjectives are incompatible with N°merger and post-nominal demonstratives.
Second, Panagiotidis’ analysis presented above at first glance seems to be
incompatible with our proposed structure for DP, presented in (40). More specifically, the
postulation of dxP, which hosts the demonstrative in its specifier position, might be
problematic with N°-raising. On Panagiotidis’ approach, N°raises to Num°for the [N]
category feature checking, or stays-in-situ when the category feature can be checked by
something else (an adjective in this case). In this scenario, the postulation of dx°between
N°and Num°may cause a problem. That is, one might think that N°should not be able to
raise when there is a demonstrative contained in the DP. This is because the
demonstrative introduced to [Spec, dxP] by external merger could block the N°-raising to
dx°, in the same way that the external merger of an adjective to [Spec, NumP] does. N°
cannot directly move to Num°skipping the intermediate dx°, due to the Head Movement
Constraint (Travis 1984).
This problem concerning N°-raising only arises if there is one and only one
category feature, [N] in this case, that can be checked. The problem disappears once we
assume that the demonstrative and dx°, in addition to [N], bear a [deictic] category
feature18, which is checked when [Spec, dxP] is filled with a demonstrative (or a pronoun
for PNCs). This allows N°to raise through dx°en route to Num°even in the presence of
a demonstrative since raising N°and merging a demonstrative in [Spec, dxP] are
18
Thanks to Heidi Harley for suggesting this possibility to me.
113
triggered by different category features: the [N] feature which triggers N°-raising and the
[deictic] feature which triggers demonstrative merger. (There is no reason to posit the
[deictic] feature for N°.) If this is the case, demonstrative merger only checks off the
[deictic] category feature and the [N] category feature remains available for checking
with other elements in the derivation (in this case N°). As a consequence, N°-raising to dx°
is allowed, and thus further movement to Num°is made possible without violating the
Head Movement Constraint. However, throughout the thesis, I will present N°-raising
abstracting away from the phrasal movement portion of Matushansky (2006), for
expository reasons. Focusing only on the head-movements in question, the structure is as
illustrated below:
(46) Post-nominal demonstrative:
a.
DP
D°
NumP
N°+dx°+Num° dxP
Dem
tdx°
NP
tN°
114
Pre-nominal demonstrative:
b.
DP
D°
NumP
AP
Num°
dxP
Dem
N°+dx°
NP
tN°
Next, let us turn to the phrasal movement—the movement of the demonstrative. I
follow Campbell (1996) in assuming the TH-Criterion, which is defined as in (47) by
analogy with Rizzi’s (1991) WH-Criterion.
(47) TH-Criterion:
A [+TH] determiner has a [+TH] specifier, and a [+TH] operator specifies a [+TH]
determiner (where [+TH] is defined to be ‘definite’, à la Campbell (1996)).
The TH-Criterion requires the specifier position of the [+TH] head to be filled by a [+TH]
element. In DNCs, this is fulfilled by the dislocation of a [+TH] element—namely, a
demonstrative—to [Spec, DP].
Panagiotidis (2000) adopts Campbell’s TH-Criterion and proposes that this can be
satisfied in either of the following two ways: dislocation of a [+TH] element—namely, a
demonstrative—to [Spec, DP], or merger of a null expletive Operator (Op) in [Spec, DP],
which blocks the movement of demonstrative to [Spec, DP] due to the Merge-over-Move
115
preference. In both cases, chains are created: <Dem…tDem> and <Op…Dem>. The
former strategy is used to derive the word order of the deictic contrastive DNC; the latter
is used to derive the word order of the anaphoric DNC. (48) schematizes the derivation
for both cases:
(48) Pre-article demonstrative:
a.
DP
Dem
Non-pre-article demonstrative:
b.
DP
D’
D°[+TH]
Op
NumP
Num°
D°[+TH]
dxP
Dem
NumP
Num°
dx’
dx°
D’
dxP
Dem
NP
dx’
dx°
N°
NP
N°
I, however, recast Panagiotidis’ analysis in terms of the copy theory of movement
(Chomsky 1993, 1995, 2000; Bobaljik 2002; Nunes 1995, 2004; Vicente 2009), which is
now a standard assumption in syntax. Rosen (2003) also adopts the copy theory of
movement in accounting for the distribution of the demonstrative in the Michif language.
On the assumption of the copy theory of movement, movement is an operation of copy
followed by merge. That is, the target of the movement is copied and the copy is merged
to the landing position. This operation creates a chain of all the copies of a given element:
<X1…X2> (where X stands for a copy).
116
To be precise, I follow Bobaljik (2002) and Vicente (2009) in assuming the
following two principles governing the status of a chain at the syntax-PF and syntax-LF
interfaces. (49) and (50) are as defined in Vicente (2009:81).
(49) Modular Chain Resolution:
The decision concerning which chain link to pronounce or interpret is locally
determined at PF or LF, respectively.
(50) Economy of Pronunciation and Interpretation:
Delete all chain copies at PF up to P-recoverability, and at LF up to Lrecoverability.
Building upon Chomsky’s (1993) idea that either the higher copy or lower copy can be
interpreted at LF, Bobajlik (2002) proposes that the same holds at PF, independently of
the decision made at LF. Hence, there doesn’t have to be a match between LF and PF
with respect to which copy to interpret and pronounce, as stated in (49). Put differently,
while pronunciation and interpretation can target the same copy of a given chain, they can
target different copies as well.
Vicente extends Landau’s (2006) idea regarding how the syntax-PF interface is
regulated, as formulated in (51) (from Landau 2006:57). Given Landau’s (p. 56)
definition of ‘P-recoverability’ in (52), the economy principle in (51) is basically
equivalent to ‘delete all the chain copies of a given element at PF except for the copies
117
that are required to be pronounced by the morpho-phonological well-formedness
considerations’.
(51) Economy of Pronunciation:
Delete all chain copies at PF up to P-recoverability.
(52) P-recoverability:
In a chain <X1…Xi…Xn>, where some Xi is associated with phonetic content, Xi
must be pronounced.
Vicente borrows (51) and (52), and assumes the same process holds at LF. That is, at the
syntax-LF interface, all the chain copies of a given element is deleted except for those
that are needed for interpretation. Vicente’s idea that the same economy principle holds at
LF will be useful in the discussion of the semantic compositionality of DNCs (see section
3.2.2.2).
If we assume that the copy theory of movement can be extended to account for null
operator constructions (see Nunes 1995: Chapter 3 Appendix), the derivation would be
better represented as below:
118
(53) Pre-article demonstrative:
a.
DP
Dem2
Non-pre-article demonstrative:
b.
DP
D’
D°[+TH]
Dem2
NumP
Num°
D°[+TH]
dxP
Dem1
D’
NumP
Num°
dx’
dx°
dxP
Dem1
NP
dx’
dx°
N°
NP
N°
In both cases, a chain of <Dem2…Dem1> is created at the end of syntax. When the
structure is shipped to PF and LF, one of the copies is deleted at both interfaces, and the
decision of which copy is deleted is made locally. That is, PF and LF make an
independent decision with respect to which copy is deleted, allowing thus mismatch
between the pronounced copy at PF and the interpretated copy at LF. Hence, the PF and
LF representation would be as illustrated in (54). PF deletion is indicated by
strikethrough and LF deletion by gray shade.
119
(54) Pre-article demonstrative:
a.
DP
Dem2
Non-pre-article demonstrative:
b.
DP
D’
D°
Dem2
NumP
Num°
D°
dxP
Dem1
D’
NumP
Num°
dx’
dx°
dxP
Dem1
NP
dx’
dx°
N°
NP
N°
This is how the word order is derived in Greek DNCs. In syntax, the demonstrative
always move to [Spec, DP] for the TH-Criterion. However, when the syntactic output is
shipped to PF, only the higher copy is pronounced when the demonstrative occurs in the
pre-article position; only the lower copy is pronounced when it occurs elsewhere.
To summarize, in this sub-section, we have seen how the various linear word orders
within DNCs are derived by a combination of head-movement and phrasal movement.
Now that we have established how DP-internal movement works, let us turn to the
discussion of the agreement within the DP.
3.2.2.2 Agreement
Before inspecting how the DP-internal elements take part in agreement with one another,
I would like to introduce the two theoretical tools that I employ, originally proposed in:
Pesetsky and Torrego (2007) and Hiraiwa (2001). I adopt in particular Pesetsky and
120
Torrego’s (2007) feature-sharing approach to agreement. Their system can be
characterized as follows. There are four possible types of features: uninterpretable and
valued, interpretable and valued, uninterpretable and unvalued, and interpretable and
unvalued. Only unvalued features can probe (irrespective of their (un)interpretability).
When a probe scans its c-command domain and finds a matching goal, Agree (i.e.,
feature-sharing) takes place as defined in (55) (Pesetsky and Torrego 2007:268).
(55) Agree (Feature sharing version):
(i)
An unvalued feature F (a probe) on a head H at syntactic location α (Fα) scans
its c-command domain for another instance of F (a goal) at location β (Fβ)
with which to agree.
(ii)
Replace Fα with Fβ, so that the same feature is present in both locations.
A probe which has agreed with its matching goal can further participate in agreement as a
goal to a higher probe. This property will be important for our purposes later. This
process is as schematized in (56) and (57). The feature on the left c-commands that on the
right; [ ] represents any unvalued feature before Agree; [x] any unvalued feature after
Agree; [y] any valued feature.
(56) … Fα [ ] … Fβ val [ ] … → … Fα val [y] … Fβ val [y] …
121
(57) a.
b.
… Fα [ ] … Fβ [ ] … → … Fα [x] … Fβ [x] …
… Fα [x] … Fβ [x] … Fγ val [y] … →
… Fα val [y] … Fβ val [y] … Fγ val [y] …
(56) illustrates the case in which the goal bears a valued feature. After Agree takes place,
the probe bears the same value as the goal. (57) is the case in which the value remains
undetermined after the first Agree (cf. (57a)); the feature value is eventually determined
by further Agree between any of the unvalued F (Fα [x] or Fβ [x]) and the valued F (Fγ [y]),
as in (57b). As a consequence, the three instances of F end up bearing the same value, [y]
in this case.
I add the concept of MULTIPLE AGREE as proposed in Hiraiwa (2001). Hirawia
(2001) proposes that Agree can take place between one probe and multiple goals. The
definition is given in (58) and the operation is schematized in (59) (Hiraiwa 2001:69-70).
(58) MULTIPLE AGREE/MOVE:
MULTIPLE AGREE (multiple feature checking) with a single probe is a single
simultaneous syntactic operation; AGREE applies to all the matched goals at the
same derivational point derivationally simultaneously. MULTIPLE MOVE
(movement of multiple goals into multiple specifiers of the same probe H) is also a
single simultaneous syntactic operation that applies to all the AGREEd goals.
122
(59) MULTIPLE AGREE:
α
>
β
>
γ
(AGREE (α, β, γ), where α is a probe and both β and γ are matching goals for α.)
If we mix the two concepts of feature-sharing and MULTIPLE AGREE, it seems to be
reasonable to reinterpret the illustrated case of agreement in (57) as MULTIPLE AGREE.
That is, the probe Fα scans its c-command domain and finds its matching goals, Fβ and Fγ,
and the value on Fγ is shared by all the three features, as shown in (60).
(60) … Fα [ ] … Fβ [ ] … Fγ val [ ] … → … Fα val [y] … Fβ val [y] … Fγ val [y] …
Given the above assumptions about Agree, agreement within DNCs proceeds as
illustrated in (61), abstracting away from the movements within DNCs.
(61)
DP
D°
NumP
[iP[ ],iN[ ],iG[ ]]
Num°
dxP
[iN[n]]
Demonstrative
[iP[p],uN[ ],uG[ ]]
dx°
NP
N°
[iG:[g]]
123
The features on D°scan its c-command domain and find multiple matching goals with
(un)valued features, and it enters an agreement relationship with the features on the Num
head, the demonstrative, and the noun simultaneously. Person is valued against the
demonstrative; number is valued against the Num head and the demonstrative; gender is
valued against both the demonstrative and the noun. The person feature on the D head is
valued against whatever the value borne by the demonstrative is, involving Agree
between one probe and one goal. In the other two operations of agreement (number and
gender), MULTIPLE AGREE is in play and thus one single probe (i.e., D°) agrees with
multiple goals (i.e., the demonstrative and the noun). As for number, the value stems
from agreement with the number feature on the Num head, but the value is shared with
the demonstrative, which also agrees with the D head. As for gender, the value originates
from agreement with the gender feature on the noun; again, the value is shared with the
demonstrative, which also agrees with the D head. As a result, all the unvalued phifeatures within the DP are valued. The final phi-feature specification is thus [iP[p], iN[n],
iG[g]] for D°and [iP[p], uN[n], uG[g]] for the demonstrative; they end up bearing the
same values for each phi-feature. (62) shows the schematized version of what has just
been described.
(62) Person – one to one Agree:
a.
D°
Dem(onstrative)
iP[ ]
iP[p]
→
D°
Dem
iP[p]
iP[p]
124
Number – MULTIPLE AGREE:
b.
D°
Num°
Dem
iN[ ]
iN[n]
iN[ ]
→
D°
Num°
Dem
iN[n]
iN[n]
iN[n]
D°
Dem
N°
iG[g]
iG[g]
iG[g]
Gender – MULTIPLE AGREE:
b.
D°
Dem
N°
iG[ ]
iG[ ]
iG[g]
→
I have assumed above that only the person feature is interpretable on the
demonstrative, for which I argue that we can find evidence in the role of the phi-features
on the demonstrative in question. In Chapter 2, we have seen that the presence of the
demonstrative plays an important role in excluding the discourse participants—the
speaker and the hearer—from the set picked out by the DNC, whether the DNC is deictic
contrastive or generic.
(63) (To) John: #Afti/Ekini
i
glossologi einai
eksypni.
these/those
the linguists be.3PL.PRES
smart
‘These/Those linguists are smart.’
(where John is a member of the set denoted by the DNC.)
Greek
If this is so, the exclusion of the speaker and the addressee is set with respect to whatever
the set-theoretic denotation of the DNC is. What is intriguing is the fact that the source of
the exclusion of the speaker and the hearer is the person feature carried by the
demonstrative—that is, third person. The standard view of third person being the lack of
person (due to Benveniste 1966), then, fits well with the exclusion story above. The
125
presence of the demonstrative prevents the speaker and the addressee from being
included in the set picked by the DP. Neither number nor gender seems to have an effect.
If this is so, we can conclude that number and gender features on the demonstrative are
uninterpretable and the morphologically distinct form that reflects number and gender is a
mere result of syntactic agreement inside the DNC.
A question then arises: How can this memebership restriction (i.e., exclusion of the
discourse participants from the denotation of the DNC) be represented in terms of
semantics? I believe that this semantic effect can be straightforwardly captured, if we
assume that the demonstrative is presuppositional.
Since Cooper (1983), it is generally assumed that phi-features on personal
pronouns are interpreted as presupposition (see also Heim and Kratzer 1998). The
semantic denotation of each phi-feature borne by pronouns are given in (64)-(66).
(64) Person (s = speaker; a = addressee):
a.
⟦1st⟧ = λx<e> . x includes s . x
b.
⟦2nd⟧ = λx<e> . x includes a . x
c.
⟦3rd⟧ = λx<e> . x excludes s and a . x
(65) Gender:
a.
⟦masculine⟧ = λx<e> . x is male . x
b.
⟦feminine⟧ = λx<e> . x is female . x
126
(66) Number:
a.
⟦singular⟧ = λx<e> . x is an atom . x
b.
⟦plural⟧ = λx<e> . x is a plurality . x
Each phi-feature is adjoined to the lowest DP (i.e., the pronoun in this case), serving as a
partial identity function of type <e, e>, as shown in (67) (cf. Heim and Kratzer 1998).
(67)
DP<e>
person<e,e> DP<e>
gender<e,e> DP<e>
number<e,e> DP<e>
pronoun
With the above discussion in place, I propose that the demonstrative in DNCs has
the semantic denotation given in (68).
(68) ⟦Demonstrativej⟧g,s,a = λx<e> . ¬(s ≤i g(j)) & ¬(a ≤i g(j)) & g(j) ≤i x & x
The demonstrative adds a presuppositional meaning to the DNC. Assuming the regular
Pro-forms and Traces Rule19, the set denoted by the DNC can include neither the referent
of the demonstrative nor the discourse participants. The semantic representation of DNCs
“If α is a pro-form or trace, i is an index, and g is an assignment whose domain includes i, then ⟦αj⟧g =
g(i)” (Heim and Kratzer 1998:292).
19
127
can be then illustrated as in (69) (the element in gray shade stands for LF deletion).20
Here, I assume that the semantic denotation of the definite article in Greek is an ιoperator, which is taken to be the maximality operator when applied to plural terms (see
the relevant discussion in section 3.3.1).
(69) LF representation:
DP<e>
¬(s ≤i g(j)) & ¬(a ≤i g(j)) & g(j) ≤i ι(λx.linguist(x)) & ι(λx.linguist(x))
Dem<e,e>
D’<e>
ι(λx.linguist(x))
D°<et,e>
λPι(λx.P(x))
NumP<e,t>
λx.linguist(x)
Num°<et,et>
λPλx(P(x))
dxP<e,t>
λx.linguist(x)
Dem
dx’<e,t>
λx.linguist(x)
dx°<et,et>
λPλx(P(x))
NP<e,t>
λx.linguist(x)
Here, I assume that Num° is an identity function which passes up the value of its sister constituent
following Bale (2009) and Scontras (2013). This kind of view of the semantics of plurality has been
originally proposed by Sauerland (2003) and has been taken up by Sauerland et al. (2005) among others. In
Sauerland’s original proposal, the number head (his ϕ) takes a DP complement (for agreement reasons), as
in [ϕP ϕ° [DP D° [NP N°]]]. On his account, plural is the default or unmarked value while singular is the
marked one. In other words, nouns are inherently plural unless specified otherwise. Unlike Sauerland’s
original proposal, I assume with Scontras (2013) that NumP (his #P) is embedded within DP. Scontras
argues that #P must be adjacent to the functional head (his NumP) whose specifier is reserved for numeral
in order to account for the one-ness presupposition imposed by [SG] on Num° (his #°). Adopting Scontras’
proposal does not cause a problem in terms of syntactic agreement, as is clear from our discussion in this
sub-section. I further assume that dx° is semantically vacuous in that its one and only role is to introduce a
demonstrative without making any semantic contribution.
20
128
Recall that we assumed above that LF only requires the highest copy of the demonstrative
to be interpreted. The lowest copy is deleted at LF by the economy condition in (50), and
thus does not concern us with the semantic compositionality of DNCs. By interpreting the
highest copy in the chain created by demonstrative movement, it is ensured that the
membership restriction imposed by the presence of the demonstrative is established with
respect to the denotation of the DNC (which is D’ in (69)). That is, the demonstrative
takes D’ as its complement, and adds the presuppositional meaning to the denotation of
the entire DP. If the tail of the demonstrative chain were to be interpreted, the
membership restriction should be established with respect to the denotation of the
nominal predicate, linguist. As has been discussed in section 2.2.2.3, however, what we
want is to set the membership restriction with respect to the denotation of the DNC.
3.2.3
Interim Summary
To summarize section 3.2, we have developed a syntactic analysis of the DNC structure,
based on the ingredients selectively chosen from previous analyses. We have reached the
conclusion that the demonstrative is base-generated in the specifier position of an
extended nominal projection, which I label ‘dx°’. In order to derive the various word
orders, I employ both N°-raising and demonstrative movement. In the presence of an
adjective within DNCs, N°raises as high as dx°to check the [N] category feature.
Otherwise, N°raises further to Num°. The demonstrative always moves to [Spec, DP] in
syntax to satisfy the TH-Criterion. The in-situ demonstrative effect is achieved by
129
pronouncing the lower copy of the demonstrative chain. The combination of the above
movements yields the word order variations in Greek DNCs, as summarized below:
Demonstrative to [Spec, DP]
Demonstrative in situ
N°-to-dx°
(i) Dem-Article-A-N
(iii) Article-A-Dem-N
N°-to-Num°
(ii) Dem-Article-N
(iv) Article-N-Dem
In Chapter 2, we have discussed the different interpretations DNCs can receive
depending on the position of the demonstrative: (i) and (ii) can receive either deictic
contrastive or generic interpretations while (iii) and (iv) receive an anaphoric
interpretation. In the next section, we will discuss how the different meanings associated
with different word orders obtains.
3.3
Disambiguating the Meanings of Demonstrative-Noun Constructions
Two of the three interpretations of DNCs—deictic contrastive and anaphoric—are easily
distinguished by the surface position of the demonstrative, as has been extensively
discussed in the literature and in Chapter 2. However, a third meaning (i.e., generic) has
been identified in Chapter 2, which has not received any attention in the literature. What
is puzzling concerning the generic interpretation is the fact that the surface form is not
instructive enough to lead to the intended interpretation. In other words, unlike the
distinction between the deictic contrastive and the anaphoric interpretations, there is no
clue as to how one can distinguish the deictic contrastive and the generic interpretations
130
since both have identical surface forms. The natural question to be asked is: How is this
ambiguity resolved? I first argue in section 3.3.1 that the different meanings between the
deictic contrastive DNC and the generic DNC stem from the different semantics of the
definite article in each instance. In order to justify this point, I adopt Giannakidou and
Stavrou’s (1999) view of the semantic duality of definite articles in Greek. I then proceed
to discuss in section 3.3.2 how the anaphoric interpretation of DNCs arises. I propose that
in anaphoric DNCs, the demonstrative moves to [Spec, DP/TopP] adopting Giusti’s
(2005, 2006) Split-DP hypothesis, and the definite article has the same semantics as that
of the deictic contrastive DNC.
3.3.1
Deictic Contrastive and Generic Interpretations
Before proceeding to discussion of the interpretational difference between the deictic
contrastive DNC and the generic DNC, we will discuss Giannakidou and Stavrou’s (1999)
view of the definite article in Greek, which I will adopt in order to disambiguate the
deictic contrastive and generic interpretations. Giannakidou and Stavrou (1999) argues
that the definite articles that occur in generic DPs are in fact contentful rather than
expletive (contra the view that treats definite articles in such contexts as expletives, as
argued by Roussou and Tsimpli (1994) and Longobardi (1994) among others).
Giannakidou and Stavrou examine what they call ‘substantivization’ in Greek,
given in (70) (Giannakidou and Stavrou 1999:296).
131
(70) I
plusii
sinithos
ksexnun apo pu
the rich
usually
forget.3PL from where
‘The rich usually forget where they started from.’
ksekinisan.
started.3PL
Greek
They convincingly show that the substantivization construction, though composed of a
definite article and an adjective without a noun, actually belongs to the category of noun
on the basis of various pieces of evidence.21 Given the apparent face-value of the element
that follows the definite article as an adjective and its nominal behavior, they argue that
substantivization is an instance of category conversion from adjective to noun. Also, they
observe that the semantics of the substantivization construction is generic. With these
conclusions in place, and given that the definite article is obligatory in the
substantivization construction, they argue that the definite article is responsible for the
observed interpretation of the substantivization. That is, the definite article is not at all
expletive, but instead it is an intensional ι-operator.
Outside the generic context, the definite article serves as an extensional ι-operator,
which is taken to be the uniqueness operator when applied to singular terms or to be the
maximality operator when applied to plural terms. Neither function of the ι-operator can
be adequately applied to yield the generic meaning of the DP, however, because generic
nominals are intensional. Since the definite article in the context of substantivization in
Greek turns a property denoted by a noun (derived from an adjective) into a kind,
Giannakidou and Stavrou propose that the definite article is an intensional ι-operator
following Chierchia (1998), which is equivalent to ∩ or a nominalization operator.
21
I will not repeat their argument here; the reader is referred to their section 4.1.
132
Equipped with the dual semantics of the definite article in Greek, we are now in a
position to be able to distinguish the two different interpretations of the DNC with a prearticle demonstrative. When it receives a deictic contrastive interpretation, the definite
article is an extensional ι-operator, while it is ∩ (i.e., an intensional ι-operator) when it
denotes a kind. Since the denotation of the DNC is dependent on the semantics of its head
(i.e., the definite article), then when the definite article functions as ∩, the subject DP
ends up denoting a kind. This in turn enables the subject DP to bear a variable which is
bound by the GN operator, leading to the generic interpretation of the whole sentence.
We now have all the ingredients with which we can distinguish the different
semantics of DNCs containing a pre-article demonstrative. As has been discussed, the
Demonstrative > Article > Adjective > Noun order is ambiguous between two
interpretations: deictic contrastive and generic. However, the structure proposed above in
(48a) does not seem to provide a way to distinguish the two interpretations. Here I
suggest that both the deictic contrastive DNC and the generic DNC have the same
syntactic structure, as shown in (71), but the definite article serves as either an
extensional ι-operator or an intenstional ι-operator.
133
(71) Deictic contrastive and generic DNC:
DP
Dem
D°[+TH]
NumP
Num°
dxP
Dem
dx°
NP
N°
In both deictic contrastive and generic context, D°bears a TH criterial feature and this
feature requires the specifier of D°to be filled. This is fulfilled by the movement of the
demonstrative to [Spec, DP]. Also, N°raises to Num°through dx°. The interpretational
difference arises from the different semantics of D°. Specifically, the deictic contrastive
interpretation emerges with D°functioning as an extensional ι-operator, while the generic
interpretation emerges with D°functioning as an intensional ι-operator.
3.3.2
Deictic Contrastive and Anaphoric Interpretations
In order to account for the anaphoric interpretation of DNCs in Greek, I adopt the SplitDP hypothesis. Let us begin with the introduction of the Split-DP hypothesis. According
to Rizzi (1997), CP is not a homogenous projection but can be decomposed into different
functional heads, each of which encodes different discourse-related information such as
Focus and Topic, as illustrated in (72).
134
(72) ForceP > TopP* > FocP > TopP* > FinP
On independent empirical grounds, it has been proposed that the DP has articulated
structure as well (Aboh 2004; Giusti 2005, 2006; Alexiadou, et al. 2007; among many
others), analogous to the split CP. Given the well-established DP/CP parallelism (Abney
1987; Horrocks and Stavrou 1987; Szabolcsi 1987; to name a few), this is not surprising.
In particular, I assume, à la Giusti (2005), that the DP domain has the structure illustrated
in (73).
(73) DP > TopP* > FocP > TopP* > defP22
On her account, when the left-periphery of the DP is split, the highest DP and the lowest
defP are respectively taken to be the locus for Case and definiteness.
I further follow Giusti (2005) in assuming that adjacent functional heads can be
collapsed into one which can realize multiple features in order to dispense with empty
functional projections. Giusti assumes that the highest DP and the lowest dP can be
realized as one and the same projection in the absence of FocP and TopP, based on
Rizzi’s original idea that ForceP and FinP can be collapsed into one projection. Giusti
further assumes that minimizing the number of projections holds for other cases as well.
For instance, DP and FocP, on the one hand, and TopP and defP, on the other, can be
collapsed into DP/FocP and TopP/defP, respectively.
22
In order to avoid the potential confusion that can be caused by DP/dP, I will use defP for dP instead.
135
Assuming the Split-DP hypothesis and the ambiguous semantics of the definite
article seems to offer a straightforward account for the semantics and syntax of the
anaphoric DNC. Anaphoric DNCs are apparently not generic, hence the definite article
has the semantics of an extensional ι-operator. The purpose of the anaphoric usage of
DNCs is to refer back to something that has been already mentioned in the discourse,
which is reminiscent of topics. That being said, I assume that the demonstrative bears not
only [+TH] but also a [+TOP] feature when the DNC receives an anaphoric interpretation.
Its movement targets the the specifier position of DP/TopP (i.e., the collapsed functional
projection à la Giusti 2005). This movement satisfies the TH-Criterion requirement and at
the same time checks [+TOP] feature. The demonstrative in anaphoric DNCs is assumed
to be co-referential with the discourse or linguistic antecedent, which yields the anaphoric
interpretation. The derivation is illustrated below:
(74) Anaphoric DNC:
DP/TopP
Dem
D°/Top°
NumP
[+TH/+TOP]
Num°
dxP
Dem
dx°
NP
N°
In (74), the head-movement takes place in the same way as we have assumed for the
deictic contrastive and generic DNC. The TH-Criterion is satisfied by merging the
136
demonstrative to the specifier of DP/TopP. As aforementioned, this merger checks the
[+TOP] feature at the same time.23
3.4
Summary
In this chapter, we have reviewed previous analyses of DNCs, and partially adopted some
components of those analyses with modifications. As a result, we have established the
followings for the syntax and the semantics of DNCs:
• The demonstrative first merges in [Spec, dxP] which is an independent functional
phrase posited between NumP and NP.
• The demonstrative bears a valued interpretable person and an unvalued
uninterpretable number and gender, and the person feature contributes to the
meaning of the DNC to the effect that the discourse participants are excluded
from the denotation of the DNC.
• Agreement inside the DNC is operated in terms of feature-sharing and MULTIPLE
AGREE.
• Different word order facts are captured by employing N°-raising (triggered by
category-feature checking) and demonstrative movement (triggered by the THCriterion).
• The three interpretations of DNCs are distinguished as follows:
It is unclear how the syntactic output in (74) is interpreted at LF. I will leave this issue for future
research.
23
137
i. Deictic contrastive DNCs: the D head functions as an extensional ι-operator;
the demonstrative moves to [Spec, DP] and is pronounced in [Spec, DP].
ii. Generic DNCs: the D head functions as an intensional ι-operator (i.e., ∩); the
demonstrative moves to [Spec, DP] and is pronounced in [Spec, DP].
iii. Anaphoric DNCs: the collapsed D/Top head functions as an extensional ιoperator; the demonstrative moves to [Spec, DP/TopP] but is pronounced in
[Spec, dxP]. The higher copy of the demonstrative in [Spec, DP/TopP] is coindexed with the antecedent mentioned in the discourse and provides a
connection to it.
The table below summarizes the relationship between the interpretations of DNCs, the
semantic denotation of definite articles, and the word order.
Spell-out
Demonstrative
in [Spec, DP]
Demonstrative
in [Spec, dxP]
N°-to-dx°
Dem-Articleext.ι-A-N
Dem-Articleint.ι-A-N
N°-to-Num°
Dem-Articleext.ι-N
Dem-Articleint.ι-N
Interpretation
deictic contrastive
generic
Articleext.ι-A-Dem-N
Articleext.ι-N-Dem
anaphoric
In the next chapter, we are going to discuss the syntactic structure of PNCs and
their interpretations. In so doing, I will carry over the analysis of DNCs developed in this
chapter to PNCs, given the conclusion of Chapter 2 that the two constructions form a
natural class.
138
CHAPTER 4
THE SYNTAX OF PRONOUN-NOUN CONSTRUCTIONS
4.1
Introduction
In this chapter, I propose a syntactic account of the structure of PNCs. The analysis will
be primarily based on the analysis of DNCs proposed in Chapter 3. The main idea is that
the pronoun is base-generated in the same position as the demonstrative, [Spec, dxP]. The
pronoun fronts to [Spec, DP] to satisfy the TH-Criterion. The D head of the entire DP
(i.e., PNC) has its unvalued phi-features valued by agreement with the DP-internal
elements: person with the pronoun, number with the Num head, and gender with the N
head. The hypothesis that pronouns, like demonstratives, first merge in the specifier of a
low functional projection might be compromised by the fact that pronouns always surface
in the left-most position in the DP, at least in Greek and English. I marshal evidence for
the low first merge position of the pronoun in the PNC from both theoretical and
empirical perspectives. For empirical motivation, I draw on relevant data from Korean,
which is quite revealing in this regard. I carry over the previously outlined method of
disambiguating DNCs with a pre-article demonstrative to the case of PNCs, whose two
interpretations cannot be distinguished by the surface word order.
This way of analyzing PNCs has important implications for agreement, the locus of
the person feature, and DP/CP parallelism. First, with respect to agreement, I show that
the main predicate agrees with the phi-features of the DP indirectly when the PNC occurs
139
in subject position; DP-internal agreement feeds DP-external agreement. In other words,
all the values of phi-features within the DP are centered on the D head of the entire DP
(PNC) and this DP acts as a goal for the probe T°for predicate-argument agreement.
Second, with respect to the locus of the person feature, the proposed account positions
person on the pronoun, which does not occupy the D head position of the entire DP. This
has a significant implication about where in the DP person is encoded. This approach
suggests that the predominant hypothesis that the D head of the entire DP is the locus for
the person feature cannot be maintained. Lastly, with respect to DP/CP parallelism, I
argue that the movement pattern of the demonstrative and the pronoun within the DP
domain parallels that of wh-phrases within the CP domain, in that anaphoric elements
stay in situ while non-anaphoric elements obligatorily front to the left-periphery of both
DP and CP.
This chapter is organized as follows. In section 4.2, I propose an account for the
base structure of PNC DPs and the syntactic derivation involved, such as agreement and
movement. Furthermore, I motivate the low first merge position of the pronoun in the DP,
providing both theoretical and empirical arguments. In section 4.2.5, I discuss the method
by which I disambiguate the two meanings of PNCs by exploiting the semantic ambiguity
of the definite article in Greek. In section 4.3, I discuss issues related to the DP-internal
distribution of phi-features. A brief discussion of the widely-accepted view of number
and gender features comes first, followed by a detailed discussion of the new view of the
locus for the person feature motivated by this research. In section 4.4, I argue that the fact
that demonstrative/pronoun movement within DP patterns with wh-phrase movement
140
provides further evidence in support of the DP/CP parallelism. In section 4.5, I criticize
two previous analyses of PNCs: the head analysis and the predication analysis. I show
that neither is adequate for Greek PNCs. In section 4.6, I look into the possibility of
extending the proposed account of PNCs based upon Greek data to English PNCs.
4.2
4.2.1
The Syntactic Derivation of Pronoun-Noun Constructions
The Base Structure of Pronoun-Noun Constructions
Given the conclusion that PNCs form a natural class with DNCs in Chapter 2, the
syntactic structure of DNCs established in Chapter 3 is carried over as a hypothesis for
the structure of PNCs. That is to say, the syntactic derivation of PNCs parallels that of
DNCs which contain a pre-article demonstrative. Recall that in Greek the pronoun in
question always surfaces in the left-most position of the DP, patterning together with a
pre-article demonstrative in terms of its distribution, but not with pre- or post-nominal
demonstratives, as has been shown in Chapter 2. Specifically, I propose that the base
structure for PNCs is as illustrated in (1). The base structure for DNCs developed in
Chapter 3 is juxtaposed for comparison in (2).
141
(1)
DP
D°
(2)
NumP
Num°
D°
dxP
Pronoun
DP
NumP
Num°
dx’
dx°
dxP
Demonstrative dx’
NP
dx°
NP
Let us take the structure in (1) as the starting point and discuss the derivation in more
detail.
4.2.2
Movement
With respect to movement inside the PNC, I assume that the same syntactic movement
operations argued for the derivation of the pre-article demonstrative structure discussed
in section 3.2 take place, as illustrated in (3):
(3)
DP
D°[+TH]
NumP
Num°
dxP
Pronoun
dx°
NP
N°
141
142
To begin with, the pronoun first merges in [Spec, dxP] and is moved to [Spec, DP], in
order to satisfy the TH-Criterion, and it is the higher copy of the pronoun that is
pronounced as well as interpreted. N°cyclically raises through dx°en route to Num°in
order to check the [N] category feature. Again, dx°is assumed to bear the [deictic]
category feature, which is checked off when the pronoun is merged in its specifier
position.
4.2.3
Mediated Pronoun-Predicate Agreement
4.2.3.1 Agreement inside the Pronoun-Noun Construction
Taking (1) as given, the derivation proceeds as described below. The features with u are
uninterpretable; those with i are interpretable; an empty pair of brackets [ ] indicates the
feature is unvalued; the value of each phi-feature is indicated by an alphabetical variable
inside a pair of brackets (p for person; n for number; g for gender).
(4)
DP
D°
NumP
[iP[ ],iN[ ],iG[ ]]
Num°
dxP
[iN[n]]
Pronoun
[iP[p],uN[ ],uG[ ]]
dx°
NP
N°
[iG:[g]]
143
With respect to DP-internal agreement, the pattern we have seen within DNCs in Chapter
3 is replicated. Recall that I adopt three theoretical proposals from the literature: (i)
Carstens’ (2000) idea that agreement in both clausal and nominal domains is regulated by
a single system; (ii) Pesetsky and Torrego’s (2007) feature sharing approach to agreement;
and (iii) Hiraiwa’s (2000) MULTIPLE AGREE. The operation then proceeds as follows. The
probe D°scans its c-command domain and finds matching goals with (un)valued features,
and it enters an agreement relationship with the Num head, the pronoun, and the noun.
Person is valued against the pronoun; number is valued against the Num head and the
pronoun; gender is valued against the pronoun and the noun. Person on the D head is
valued against the matching feature on the pronoun, involving Agree between one probe
and one goal. In the latter two instances of agreement (number and gender), MULTIPLE
AGREE is at play and thus one single probe (i.e., D°) agrees with multiple goals (i.e., the
pronoun and the noun). As for number, the value stems from agreement with the Num
head, but the value is shared with the pronoun, which also agrees with the D head. As for
gender, the value originates from agreement with the noun; again, the value is shared
with the pronoun, which also agrees with the D head. As a result, all the unvalued phifeatures within the DP are valued. The final phi-feature specification is thus [uP[p],
uN[n], uG[g]] for D°and [iP[p], uN[n], uG[g]] for the pronoun; they end up bearing the
same phi-features. The process of agreement is as schematized in (5) (the probe for each
Agree instance is in bold).
144
(5)
Person – one to one Agree:
a.
D°
Pronoun
iP[ ]
iP[p]
→
D°
Pronoun
iP[p]
iP[p]
D°
Num°
Pronoun
iN[n]
iN[n]
iN[n]
D°
Pronoun
N°
iG[g]
iG[g]
iG[g]
Number – MULTIPLE AGREE:
b.
D°
Num°
Pronoun
iN[ ]
iN[n]
iN[ ]
→
Gender – MULTIPLE AGREE:
c.
D°
Pronoun
N°
iG[ ]
iG[ ]
iG[g]
→
Another difference comes from the (un)interpretability of the other phi-features of
the pronoun—namely, number and gender. As we have seen in the case of DNCs in
Chapter 3, it has been proposed that it is only person on the demonstrative that is
interpreted in the context of DNCs. Person on the demonstrative, being third person,
makes a semantic contribution to the effect that the discourse participants are excluded
from the set picked out by the DNC. The semantic denotation of the third person feature
and the simplified LF representation of DNCs are respectively repeated in (6) and (7).
(6)
⟦Demonstrativej⟧g,s,a = λx<e> . ¬(s ≤i g(j)) & ¬(a ≤i g(j)) & g(j) ≤i x & x
(where s is the speaker, a the addressee.)
145
(7)
DP<e>
Dem<e,e>
DP<e>
DNC
As has been discussed in Chapter 2, the pronoun in PNCs make a similar semantic
contribution to the PNC interpretation. The only difference between the pronoun in PNCs
and the demonstrative in DNCs is the value of person. Being the first or second person,
the first and second person pronoun in PNCs add to the PNC interpretation the effect that
the discourse participant(s) are properly included in the set picked out by the PNC. Thus
the proposed semantic denotation of the pronoun in PNCs is as given in (8) and the LF
representation of PNCs is given in (9). 1
(8)
a.
⟦wej⟧g,s = λx<e> . s ≤i g(j)) & g(j) ≤i x & x
b.
⟦youj⟧g,a = λx<e> . a ≤i g(j)) & g(j) ≤i x & x
This analysis is similar to Elbourne’s (2005) treatment of the pronoun in question. He proposes that the
pronoun is an <<e,t>,e> type element, with the dentation below:
1
(i)
⟦youj⟧g,a = λf : f ∈ D<e,t> & a ≤i g(j) & f(g(j)) = 1.g(j)
= “this plural you takes an NP with denotation f and gives as the denotation of the whole DP some
contextually salient plural individual j that is conditioned as follows: the addressee a must be part of
j, and j must be f” (Elbourne 2005:43).
As one can see from the semantic type given to the pronoun, he treats the pronoun on par with the definite
article following Postal (1966), according to which the pronoun is the head of the PNC DP.
However, Postal’s syntactic analysis of the pronoun cannot be maintained, as will be discussed in
section 4.5.2. Due to the fact that Postal’s analysis, which Elbourne assumes, is untenable, it follows that
Elbourne’s approach to the semantics of the pronoun under investigation needs to be modified.
146
(9)
LF representation:
DP<e>
s ≤i g(j)) & g(j) ≤i ι(λx.linguist(x)) & ι(λx.linguist(x)), or
a ≤i g(j)) & g(j) ≤i ι(λx.linguist(x)) & ι(λx.linguist(x))
Pronoun<e,e>
D’<e>
ι(λx.linguist(x))
D°<et,e>
λPι(λx.P(x))
Num°<et,et>
λPλx(P(x))
NumP<e,t>
λx.linguist(x)
dxP<e,t>
λx.linguist(x)
Pronoun
dx’<e,t>
λx.linguist(x)
dx°<et,et>
NP<e,t>
λPλx(P(x)) λx.linguist(x)
In this regard, the phi-features of the pronoun other than person—number and gender—
are assumed to be uninterpretable.
4.2.3.2 Agreement outside the Pronoun-Noun Construction
If we are on the right track in postulating the DP-internal agreement operations as
proposed in section 4.2.3.1, a straightforward account follows of how the person feature
on the pronoun is reflected on the inflection of the predicate. It is well-known that in a
language that shows (rich) morphological inflection on the predicate, the inflection is
147
determined in accordance with the pronoun contained in the subject. Consider the Greek
examples in (10), (11) and (12), focusing on subject-predicate agreement.2
(10) [I
glossologi]SUBJECT
the linguists
‘The linguists are smart.’
(11) a.
b.
(12) a.
b.
ine
be.3PL.PRES
exypni.
smart
Greek
[Emis
i
glossologi]SUBJECT
we
the linguists
‘We linguists are smart.’
[Esis
i
glossologi]SUBJECT
you.PL
the linguists
‘We linguists are smart.’
imaste
be.1PL.PRES
exypni.
smart
isaste
be.2PL.PRES
exypni.
smart
*[Emis
i
glossologi]SUBJECT
we
the linguists
‘We linguists are smart.’
*[Esis
i
glossologi]SUBJECT
you.PL
the linguists
‘We linguists are smart.’
ine
be.3PL.PRES
exypni.
smart
ine
be.3PL.PRES
exypni.
smart
Greek
Greek
With a regular DP subject, the verb is inflected for third person, as in (10). On the other
hand, in (11) a PNC subject requires the verb to be inflected for first or second person,
according to the person of the pronoun. Having third person inflection on the verb in the
presence of a PNC subject results in ungrammaticality, as in (12).
This is not surprising at all given the analysis of agreement within PNCs proposed
in section 4.2.3.1. I propose that the agreement between the pronoun and the predicate is
This agreement fact is taken to be the evidence in favor of the head analysis of PNCs. Detailed discussion
of this is postponed until section 4.5.2.
2
148
mediated by the D head of PNCs. That is, the DP-internal agreement ensures all values of
the phi-features scattered around inside the DP are gathered by the D head, which in turn
makes it possible for the predicate to agree with the subject DP. Specifically, the
derivation proceeds in the following way:
(13)
TP
DPPNC
T’
T°[EPP,uP[ ],uN[ ]]
vP
Agree2
DPPNC[iP[p],iN[n],iG[g]]
Pronoun
D°
Agree1
D’
v°
NumP
Pronoun
v’
VP
…
Noun
Let us suppose that DP-internal agreement proceeds as described above. That is, the
unvalued phi-features on the D head of PNCs probe downward and agree with the
matching valued phi-features within the DP. As a result, all the unvalued phi-features are
valued—importantly for us, the person feature is valued via agreement with the pronoun,
and the D head ends up carrying all the values of the phi-features scattered within the DP
(Agree1 in (13)). This PNC subject is introduced in [Spec, vP]. When T°merges with vP,
the T head enters into agreement with the subject DP, in this case a PNC (Agree2 in (13)).
The main point here is that the unvalued person feature on the T head agrees with the
149
matching feature on the pronoun indirectly via D°. In other words, the agreement in terms
of person, number, and gender features is mediated by the head D of the subject DP. 3
4.2.4
Motivating the Low First Merge Position of the Pronoun
The reader may well wonder why we have to assume that the pronoun is merged in a low
position within the DP and raises higher in the structure. This question needs to be
answered adequately for the following reason. In the case of DNCs in Greek, we have
seen strong evidence for assuming the low base-position of the demonstrative given that
the demonstrative can surface in a non-DP-initial position. However, the story is different
when it comes to PNCs: the distribution of the pronoun is strictly limited to the prearticle position.
There are nonetheless (at least) two reasons for claiming that the pronoun merges
low in the structure. I provide these two arguments in the rest of this section. One
argument is based on a purely theory-internal reason (which could be compromised if one
adopted different theoretical tools). The other is motivated by an empirical reason,
considering relevant data drawn from Korean.
On the theoretical side, I argue that the pronoun must be introduced low in the DP
in order for Agree to work properly. More specifically, I assume Pesetsky and Torrego’s
The account for the agreement between the pronoun and T° leads to a straightforward account for the fact
that the pronoun in PNCs in Greek can be omitted, as in (i), as will be fully discussed in Chapter 5.
3
(i)
[(Emis)
i
glossologi]SUBJECT
we
the
linguists
= ‘We linguists won the award.’
≠ ‘The linguists won the award.’
nikisame
to
win.1PL.PST the
epathlo.
award
Greek
150
(2007) feature sharing approach to agreement and Hiraiwa’s (2000) MULTIPLE AGREE as
two theoretical tools regarding how Agree works. In both theories, Agree takes place
between a probe and one or multiple goals if and only if a structural relationship between
the two elements is met. That is, a probe Agrees with a goal when the former ccommands the latter. Under the current assumptions, therefore, we are forced to basegenerate the pronoun somewhere low in the structure, so that the D head can c-command
the pronoun and in turn establish an agreement relationship, despite the fact that pronouns
never surface low (i.e., somewhere between the definite article and the noun in Greek) in
PNCs. This way of justifying the low first merge position of the pronoun in PNCs may be
challenged if one adopts a different stance about how Agree works. That is, there are
some works that argue that a base-generated Spec-Head relationship is a legitimate
syntactic configuration in which Agree can take place (Chung 1998; Koopman 2003,
2006; see also Kratzer 2009). If one adopted this standpoint regarding how Agree works,
one could simply say that the pronoun is base-generated in [Spec, DP] and the D head
agrees with the pronoun in a Spec-Head configuration. We therefore cannot rely solely on
our theoretical framework in order to motivate the low first merge position of the
pronoun.
In order to bolster my stance regarding the low base position of the pronoun and to
fend off a possible counter-proposal that adopts a different theory of Agree, I turn to
empirical evidence from other languages such as Korean. The PNC in Korean in
particular is instructive for our purposes. Korean is one of those languages that allow for
PNCs. As shown in the examples given in (14), the basic word order is Pronoun > Noun.
151
(14) a.
b.
Wuli
enehakcatul
we
linguists
‘We linguists.’
*Enehakcatul
wuli
linguists
we
Korean
When we take into account adjectives, however, Korean PNCs offer revealing examples.
Unlike many other languages including Greek and English, Korean allows two word
orders with PNCs containing an adjective: Pronoun > Adjective > Noun, as in (15a), and
Adjective > Pronoun > Noun, as in (15b).
(15) Pronoun-Adjective-Noun:
a.
Wuli
ttokttokhan
we
smart
‘We smart linguists’
Adjective-Pronoun-Noun:
b.
Ttokttokhan
wuli
smart
we
‘We smart linguists’
enehakcatul
linguists
Korean
enehakcatul
linguists
The two variants are equally syntactically well-formed, and at first glance seem to exhibit
no difference in terms of interpretation. However, there is indeed a subtle difference in
their interpretations that can only be observed in an appropriate context. Consider the
following dialogue:
152
(16) A:
B:
B’:
Etten
what/which
ttokttokhan enehakcatul-i
smart
linguists-NOM
sang-ul
award-ACC
Korean
pass-ass-ni?
receive-PST-INT
‘Which smart linguists won the award?’
Wuli ttokttokhan enehakcatul-i
pass-ass-e.
we smart
linguists-NOM
receive-PST-DECL
‘We smart linguists won (it).’
#Ttokttokhan
wuli enehakcatul-i
pass-ass-e.
smart
we linguists-NOM
receive-PST-DECL
‘We smart linguists won (it).’
As shown above, it is only (16B) that can be an appropriate answer to the question in
(16A). (16B’), though syntactically well-formed, would be semantically at odds with the
provided context. Given that the question is a wh-question that targets the pronoun as the
answer, we can conclude that the pronoun which precedes the adjective receives a focus
interpretation. The reason of the semantic oddity of (16B’) is then due to the fact that the
pronoun is not in a focus position.4
Interestingly, the presence of the pronoun seems to have an impact on whether or not the adjective can be
focused, as in (i).
4
(i)
a.
b.
Etten
enehakcatul-i
sang-ul
pass-ass-ni?
what/which linguist-NOM
award-ACC receive-PST-INT
‘What kind of linguists won the award?’
Ttokttokhan enehakcatul-i
pass-ass-e.
smart
linguists-NOM
receive-PST-DECL
‘The smart linguists won (it).’
Korean
However, the wh-question cannot target the adjective when the DP is an instance of PNCs, as in (ii).
(ii)
*Etten
what/which
wuli
we
enehakcatul-i
linguist-NOM
sang-ul
award-ACC
pass-ass-ni?
receive-PST-INT
Korean
153
Note that the dialogue is not the most natural conversation in Korean. The most
natural answers to the question in (16A) would be those given in (17) which do not repeat
the adjective given in the question.
(17) a.
b.
Wuli-ka
pass-ass-e.
we-NOM receive-PST-DECL
‘We received (it).’
Wuli enehakcatul-i
pass-ass-e.
we linguists-N`OM receive-PST-DECL
‘We linguists received (it).’
Korean
However, despite the reduced naturalness, if the discourse participant is coerced to repeat
the adjective in giving an answer to (16A), it is only (16B) but not (16B’) that can be
understood appropriately. If Pronoun > Adjecitve > Noun order is the one used in the
context where the pronoun is focused, we can draw the conclusion that the other
variation—Adjective > Pronoun > Noun order—is the basic and unmarked word order.
Before closing this section, there is another factor we need to consider. Kang (2005)
argues that Korean adjectives fit well with Cinque’s (2010) story of adjectives, according
to which indirect modifiers, involving reduced relative clauses, are merged in a position
hierarchically superior to direct modifiers. The adjective used above (ttokttokhan ‘smart’)
to test the low position of the pronoun is an indirect modifier. This leads us to wonder if
the pronoun can merge in a higher position than a direct modifier. The answer is negative,
as confirmed by the relative word order between the pronoun and an adjective of direct
modification, cen ‘former’, as shown in (18).
154
(18) Pronoun-Adjective-Noun:
a.
Wuli cen
phyencipcangtul
we former
chief.editors
‘We former chief editors’
Adjective-Pronoun-Noun:
b.
Cen
wuli phyencipcangtul
former
we chief.editors
‘We former chief editors’
Korean
Again, the basic word order is the one in (18b), as confirmed by the oddity of Adjective >
Pronoun > Noun order as an answer to the wh-question that targets the pronoun.
(19) A:
B:
B’:
Etten
what/which
cen phyencipcangtul-i
smart chief.editors-NOM
sang-ul
award-ACC
Korean
pass-ass-ni?
receive-PST-INT
‘Which former chief editors won the award?’
Wuli
cen
phyencipcangtul-i
pass-ass-e.
we
former
chief.editors-NOM
receive-PST-DECL
‘We former chief editors won (it).’
#Cen
wuli phyencipcangtul-i
pass-ass-e.
former
we chief.editors-NOM
receive-PST-DECL
‘We former chief editors won (it).’
The conclusion that is to be drawn from the above discussion of Korean PNCs
containing an adjective is that the pronoun is base-generated in a position lower than the
all adjectives, both indirect and direct in the sense of Cinque (2010), and that the preadjective pronoun is in a derived focus position. I construe the Korean facts to be a piece
of empirical evidence in favor of the hypothesis that the pronoun first merges low in the
DP.
155
4.2.5
Disambiguating the Two Meanings of Pronoun-Noun Constructions
Unlike DNCs, which can receive three distinct interpretations, PNCs receive two
different interpretations: deictic contrastive and generic. Like the deictic contrastive and
generic interpretations of DNCs, surface word order cannot distinguish the two readings
for PNCs either since PNCs require the pronoun to appear in the pre-article position in
Greek. In distinguishing the two readings of PNCs, I carry over my analysis of DNCs.
More specifically, I assume the lexical ambiguity of the definite article following
Giannakidou and Stavrou (1999). According to them, there are two types of definite
articles in Greek: an extensional ι-operator and an intensional ι-operator. The former is
used for referential DPs while the latter is used for generic DPs. The latter is equivalent
to Chierchia’s type-shifting operator ∩ that nominalizes a property into a kind. Consider
(20) with the above assumptions in place:
(20) Deictic contrastive and generic PNC:
DP
D°[+TH]
NumP
Num°
dxP
Pronoun
dx°
NP
N°
156
The head D°agrees with the pronoun in terms of phi-features as described in section
4.2.3.1, and the pronoun fronts to [Spec, DP] to satisfy the TH-Criterion. At PF and LF,
the head of the pronoun chain is respectively pronounced and interpreted.
The two readings arise depending on the denotation of the definite article. That is, a
deictic contrastive reading becomes available if the definite article is an extensional ιoperator while a generic reading becomes available if the definite article is an intensional
ι-operator, namely ∩.
4.2.6
Phrasal Status of Pronouns
We have thus far implicitly assumed that the pronoun under investigation is a phrase on
its own. Recall that the syntactic positions which the pronoun occupies throughout the
derivation are always specifier positions—namely, [Spec, dxP] and [Spec, DP]. In this
section, we are going to remark on the two arguments in favor of the phrasal status of the
pronoun.
First, from a theoretical perspective, the pronoun encased in PNCs should be
treated as a phrase. In Greek, PNCs require a definite article as well as a pronoun. The
general assumption that a definite article is D°leads to the conclusion that what precedes
the definite article should be in the specifier of D°.5
Second, we can find empirical evidence for the view that pronouns are phrasal on
their own. In fact, we have already discussed the evidence in section 2.2.1.2.2. Pronouns
This argument can be compromised if one postulates an additional functional head to host the pronoun
preceding the definite article. This possibility is, however, rejected for the second reason.
5
157
can be collocated with a reinforcer whose role is to modify the deictic property of the
pronouns. Representative examples are given in (21) and (22) for English and Greek.
(21) a.
b.
We here/*there linguists
You here/there linguists
(22) a.
Emis
edho/*eki i
we
here/there the
‘We here/*there linguists.’
Esis
edho/eki i
you
here/there the
‘You here/there linguists.’
b.
glossologi
linguists
Greek
glossologi
linguists
As has been discussed in Chapter 3, Choi (2013) argues that the demonstrative-reinforcer
construction is an instance of a modifiee-modifier relationship. Given the working
hypothesis that DNCs and PNCs form a natural class along with the fact pronounreinforcer constructions behave in the same ways as demonstrative-reinforcer
constructions with respect to the three reinforcerhood diagnostics (section 2.2.1.2.2), I
carry over that treatment of demonstrative-reinforcer constructions to pronoun-reinforcer
constructions. That is, the reinforcer is right-adjoined to the pronominal DP, as illustrated
in (23). RP stands for reinforcer phrase.
158
(23)
DP
DP
D°
RP
NumP
Num
NP
For the purposes of this section, suffice it to say that pronouns are complex with their
phi-features distributed within the DP: person on D°, number on Num°, gender on N°(a
detailed discussion of the phi-feature distribution will be provided in section 4.3). The
possibility of combining a pronoun and a reinforcer also suggests that treating the
pronoun as the head of PNCs cannot be maintained, as will be discussed in section 4.5.2.
To recap section 4.2, the derivation of PNCs involve the same process as that of
DNCs, which is expected from the conclusion of Chapter 2 that the two constructions are
parallel to each other. That is, the pronoun is base-generated in the specifier of a low
functional head dx°and moves to [Spec, DP] for the TH-Criterion. In sections 4.3 and 4.4,
I will discuss two consequences of the proposed account of PNCs.
4.3
The Locus of Person and Two Types of Pronouns
The analysis of the mediated pronoun-predicate agreement has an important implication
for the locus of the person feature within the DP. In much of the literature on pronouns,
these elements are argued to be complex—not atomic (Ritter 1995; Déchaine and
Wiltschko 2002; Panagiotidis 2002; Bernstein 2008; Radford 2009). Despite some
159
differences in the details, the idea commonly shared by these authors is that the phifeatures are distributed within the DP. This idea is in contrast with the claim that all the
phi-features are conflated into a single head, D°(Postal 1966; Abney 1987). (24) is the
complex structure of pronouns proposed by Panagiotidis (2002).
(24)
DP
D°
NumP
[iP[p],uN[ ],uG[ ]]
Num°
NP
[uN[n]]
N°
[uG:[g]]
In (24), each phi-feature is encoded on a different functional/lexical head: person, number,
and gender are borne on the D, Num, and N heads, respectively. This way of designating
a certain feature to a certain functional or lexical head has become the standard view of
the distribution of phi-features within the DP; pronouns, being a DP, also have phifeatures scattered within their own DP. The hypothesis that the Num head encodes
number has been adopted in the literature on the DP since it was first proposed by Ritter
(1991). The gender feature has been argued to be borne on the lexical item (i.e., the N
head) given the arbitrary association between the gender feature and a given noun in
languages with gender systems. For instance, there is no a priori reason that table ‘table’
should be feminine in French.
What then about the person feature, which is the main concern of this sub-section?
The prevailing hypothesis about the locus for person is that the D head encodes person
160
(Ritter 1995; Déchaine and Wiltschko 2002; Panagiotidis 2002; Bernstein 2008;
Longobardi 2008; among others). Let me call this the ‘Person-on-D°hypothesis’. The
main empirical evidence in favor of this hypothesis comes from consideration of the facts
exemplified in (25) (Radford 2009:130).
(25) a.
b.
we linguists = 1st person DP:
We linguists take ourselves/*yourselves/*themselves too seriously, don’t
we/*you/*they?
you linguists = 2nd person DP:
You linguists take yourselves/*ourselves/*themselves too seriously, don’t
you/*we/*they?
The subjects of the sentences in (25) are all PNCs. What is of interest to us is the fact that
the person feature of the PNCs is determined by the person feature of the pronoun. This is
clearly shown by the appropriate form of the reflexive pronouns bound by the PNC
subjects as well as the appropriate form of the pronouns in the tag questions. The
paradigm in (25) supports the idea that the person feature of the PNC is determined by
the pronoun even in languages such as English in which the inflection on the predicate is
not rich enough and thus show no morphological distinction (cf. Greek in (10), (11), and
(12)). This fact, coupled with the assumption that the pronoun is the head D of PNC DPs
(Panagiotidis 2002; see also Déchaine and Wiltschko 2002; Bernstein 2008; Longobardi
2008), leads to the conclusion that the D head is the locus for the person feature. On
Panagiotidis’ account, for instance, a bare pronoun we and a PNC we linguists are as
illustrated in (26) and (27), respectively.
161
(26) we
(27) we linguists
DP
DP
D°
we
NumP
Num°
NP
N°
e
D°
we
NumP
Num°
NP
N°
linguists
Panagiotidis (2002) argues that pronominal DPs always contain the NP portion
(represented as e in (26)), which is responsible for the pronominal interpretation at LF.
He follows Abney (1987) in assuming that pronouns differ from other nominal
expressions (such as common nouns) in terms of whether or not a concept is denoted.
That is, what defines pronouns as pronouns is the lack of such a concept in contrast with
common nouns such as dog, cat, dream, etc. He associates the lack of a concept with the
presence of a covert NP, e.
As a matter of fact, the Person-on-D°hypothesis fares well with the agreement
facts illustrated in (25). As will be extensively discussed below in section 4.5.2, however,
the hypothesis that the pronoun is the head D of PNCs cannot be maintained for
independent reasons. It thus follows that the Person-on-D°hypothesis, which heavily
relies on the untenable assumption that the pronoun heads PNC DPs, is compromised.
Instead, we have developed in section 4.2 an alternative analysis, according to
which the pronoun is not itself the head of a PNC, but instead surfaces in [Spec, DP] as a
result of movement. Under this account, the person feature that determines that of the
entire DP (i.e., PNC) is borne on the the pronoun, rather than the head of PNCs. The head
162
D of PNCs merely obtains the same person value as a result of syntactic agreement with
the pronoun (in the manner discussed in section 4.2.3.1). For this reason, though the head
D of PNCs eventually bears the same person value as the pronoun via agreement, it is the
pronoun that is the true source of person inside the PNC.
With the conclusion just mentioned in place, the question that naturally arises is:
What is the locus for person inside the pronominal DP? I assume with the previous
literature that pronouns have complex internal structure containing D°, Num°, and N°.
Each functional head is associated with a phi-feature: D°encodes person, Num°number,
and N°gender. I further assume with Panagiotidis (2002) that pronominal DPs contain a
covert NP. The internal structure of pronouns in (28) summarizes these assumptions.
(28) Pronoun in the PNC:
DP
D°
NumP
[iP[p],uN[ ],uG[ ]]
Num°
NP
[uN[ ]]
N°
[uG:[ ]]
e
On the face of it, the above structure closely resembles the one in (24). However,
there are two crucial differences. First, the structure in (28) which I propose for the
structure of the pronouns in PNCs always contains a covert N°. Compare the regular
pronominal DP (cf. (26)) and the pronoun in PNCs (cf. (27)): the difference between the
two constructions is whether or not the lexical noun is covert or overt. When the noun is
163
covert, we obtain a regular pronominal DP; when the noun is overt, we obtain a PNC DP.
My account is essentially different from such a view. What follows from the analysis
proposed in section 4.2 is that pronominal DP always contains a covert N°whilst nonpronominal DP always contains an overt N°. That is, whether N°is covert or overt is not
an option. Whenever one sees a pronoun, there is always a covert N°. In the case of PNCs,
a pronominal DP that contains a covert N°is embedded within the larger DP, as is clear
from the proposal in section 4.2.
I instead distinguish pronominal DPs encased in PNCs from regular pronominal
DPs in terms of different phi-feature specifications even though both types have the same
forms. That is, I suggest that we have two different types of pronouns:
(29) Pronoun in the PNC:
DP
D°
(30) Regular pronoun:
DP
NumP
[iP[p],uN[ ],uG[ ]]
Num°
D°
NumP
[iP[p],uN[ ],uG[ ]]
NP
[uN[ ]]
Num°
NP
[iN[n]]
N°
N°
[uG:[ ]]
[iG:[g]]
e
e
The phi-feature specification of pronouns encased in PNCs differs from the regular
pronouns that are used on their own without being a part of a larger DP. In (29), the only
valued phi-feature is person on D°, and the morphological manifestation of all the other
phi-features is simply a reflex of syntactic agreement within the PNC DP that embeds the
pronominal DP in (29), as suggested in section 4.2.3.1. That is, the values of number and
164
gender are later valued by further agreement once it is embedded in the larger DP. Thus,
at the formation of the adnominal pronoun, only person is initially valued, and number
and gender remain unvalued. Instead of valuation, there is a permanent link created
between the probe and the goals (in the sense of Pesetsky and Torrego 2007). The
unvalued features will be valued via DP-internal agreement, as suggested in section
4.2.3.1 (cf. (5)). Hence, the value of number on the Num head and gender on the N head
in (29) will be eventually valued by the number and gender features on the D head in (5),
which in turn will be valued by further valuation of number and gender on the Num and
N heads in the context of the PNC (see (4) and (5)).
By contrast, the regular pronoun contain all the valued interpretable phi-features in
itself, as in (30): not only the D head bears valued person features and unvalued number
and gender features, but also the Num and N heads bear valued number and gender
features, respectively. The D head obtains all the values for each phi-feature via DPinternal agreement.
The approach to the locus of person within the DP proposed here adds another
criterion by which we can demarcate any types of pronominal DPs from non-pronominal
DPs. Specifically, pronominal DPs, whether regular or not, bear valued person features
while non-pronominal DPs bear unvalued person features. When a given non-pronominal
DP is an instance of a PNC, the D head of the PNC obtains its person value via
agreement with the pronoun within, which bears a valued person feature. In contrast,
when a given non-pronominal DP is an instance of a non-PNC, the D head of the DP
obtains third person value by default. This is why non-PNCs are always third person.
165
4.4
Parallelism between DP and CP
Now we have a syntactic analysis of the structure of DNCs and PNCs at hand. We have
seen so far that the distribution of the demonstrative and the pronoun within the DP is
associated with particular interpretations of the DP. Specifically, anaphoric elements stay
in-situ while non-anaphoric elements move to the specifier of DP. This leads to a
difference between DNCs and PNCs: DNCs can be anaphoric and thus the demonstrative
can surface in its base position, while PNCs cannot be anaphoric and thus the pronoun
always surfaces to the left of the definite article. This fact is illustrated by the examples in
(31).
(31) Pronoun-Article-Noun (deictic contrastive or generic):
a.
Emis i
glossologi
we the linguists
‘We linguists’
*Article-Noun-Pronoun:
b.
*I
glossologi emis
the linguists we
(32) Demonstrative-Article-Noun (deictic contrastive or generic):
a.
Afti i
glossologi
these the linguists
‘These linguists’
Article-Noun-Demonstrative (anaphoric):
b.
I
glossologi afti
the linguists these
‘These linguists’
Greek
Greek
166
The above facts raise a question from the perspective of DP/CP parallelism. It has
been argued that the DP can be (to a certain extent) likened to the CP in the literature
(Szabolcsi 1983, 1987, 1994; Abney 1987; Ritter 1991; Giusti 1996, 2005, 2006;
Cardinaletti and Starke 1999; Pesetsky and Torrego 2001; Haegeman and Ürögdi 2010;
among many others). If this is so, we can expect the A’-movement of the demonstrative
and the pronoun to [Spec, DP] to pattern with A’-movement to [Spec, CP]. In this
connection, the movement pattern of the pronoun and the demonstrative to [Spec, DP]
seems to be deviant considering that wh-phrases are known to obligatorily move to [Spec,
CP] in Greek, while we have seen that the demonstrative, at least, may remain in-situ.
These two facts regarding the differing movement patterns inside the DP and CP
can be taken to be an exception to the DP/CP parallelism. However, if we adopt a more
recent view of wh-movement in Greek, the whole picture becomes complete, constituting
another piece of evidence for DP/CP parallelism.
Greek has been traditionally classified as a wh-movement language like English
(Agouraki 1990; Tsimpli 1990, 1995, 1998). Therefore, a sentence with an in-situ whphrase is only expected to be interpreted as an echo question. Compare (33a) and (33b)
with an attention to the distribution of ti ‘what’.
(33) a.
b.
Ti
o
pateras
su
aghorase?
what.ACC the father.NOM yours
bought.2SG
‘What did your father buy?’
O
pateras
su
aghorase
ti?
the father.NOM yours
bought.2SG
what.ACC
= ‘Your father bought what?’
≠ ‘What did your father buy?’
Greek
167
Recently, however, Vlachos (2012; see also his 2008, 2009, 2010), who credits the
original observation to Sinopoulou (2007), argues that Greek is actually a mixed-type
language with regards to wh-movement. More specifically, questions with wh-phrases insitu, which had been considered to only have an echo question interpretation, are actually
ambiguous between an echo question and a true information-seeking question. The
difference in meaning is disambiguated by a prosodic difference: an echo question comes
with a rising intonation while a true information-seeking question comes with a falling
intonation. What is crucial here is the fact that wh-in-situ information-seeking questions
require a pre-established linguistic environment (what Vlachos calls ‘mini-discourse),
unlike wh-movement information-seeking questions. This is clearly characterized in the
short discourse in (34) (from Vlachos 2012:29-30).
(34) a.
b.
c.
Speaker A:
My father, my mother and I went to the store to buy eggs, milk and coffee.
My mother bought the eggs.
Speaker B [with falling intonation]:
Ke o
pateras
su
aghorase
ti?
Greek
and the father.NOM your bought.3SG
what.ACC
‘And what did your father buy?’
Speaker A:
#Aghorase
mila.
bought.3SG
apples.ACC
‘He bought apples.’
As a response to (34b), (34c) is infelicitous, due to the anaphoric property of the in-situ
wh-phrase, ti ‘what’. That is, the possible value of ti ‘what’ in this specific context “must
range over the set of entities already present in the immediate discourse” (Vlachos
168
2012:29). That is, if the answer in (34c) refers to either milk or coffee, the sentence
becomes felicitous in the given context. This fact complies with the fact regarding the
anaphoric demonstrative. As we have seen in Chapter 3, the anaphoric demonstrative is
used to refer back to an entity that has been mentioned in the discourse. (35) illustrates
this point.
(35) Context: A paragraph from a guide book about a Greek town.
a.
I
poli eci [pola istorika
ktiria]
pu
the town has many historical buildings that
b.
c.
Greek
xronologhunte apo ti
vizantini epoci.
date
back to
Byzantine period
‘The town has many historical buildings that date back to Byzantine period.’
[Ta ktiria
afta] episceptonte
kathe xrono ekatondadhes
the buildings these visit.3SG.PRES
every year hundreds
turistes.
tourists
‘These buildings are visited every year by hundreds of tourists.’
???[Afta ta
ktiria]
episceptonte
kathe xrono ekatondadhes
these
the buildings visit.3SG.PRES
every year hundreds
turistes.
tourists
The DNC with an in-situ demonstrative in (35b) is felicitous while the DNC with a
dislocated demonstrative in (35c) is not. And the DNC in (35b) ranges over pola istorika
ktiria ‘many historical buildings’ mentioned in (35a), partially the same as the in-situ whphrase (35b). The only difference between the anaphoric in-situ demonstrative and whphrase is that the former is definite and thus ranges over a unique referent while the latter
is indefinite and thus, as shown in (34), can range over more than one entity.
169
If Vlachos (2012) and I are on the right track, the dislocation to [Spec, DP] and that
to [Spec, CP] pattern together; non-anaphoric elements are dislocated to [Spec, DP/CP],
but anaphoric elements stay in-situ in Greek. This, therefore, supports DP/CP parallelism
with respect to what kind of elements are or are not to be dislocated to the left-periphery
of each domain.
This view makes a prediction about the pattern of wh-movement and
demonstrative/pronoun movement in other languages. That is, it is expected that in true
wh-movement languages, all the demonstratives should appear in the DP-initial position
as a result of movement, since the base position is never associated with an anaphoric
interpretation. This prediction is borne out in English, as shown in (36).
(36) a.
b.
c.
These/We smart linguists
*Smart these/we linguists
*Smart linguists these/we
Note that the presence of an adjective is an indicator for (non-)movement of
demonstratives, assuming that adjectives are base-generated in the specifier position of an
extended nominal projection (cf. Cinque 1994, 2010), and that the base-position of
demonstratives as well as that of pronouns is lower than the functional projections that
host adjectives (see Chapter 3 and section 4.2). That is, if a demonstrative follows an
adjective or a noun, it suggests that the demonstrative has not undergone movement.
Given this, the word orders in (36b) and (36c) indicate no movement of the
demonstrative or pronoun and thus the examples are ungrammatical. On the contrary,
170
(36a), with the DP-initial demonstrative/pronoun, is grammatical. This pattern is
expected given that English is consistently wh-movement language.
I also find that the movement to [Spec, DP/CP] patterns together in a wh-in-situ
language like Korean. As we have seen in section 4.2.4, Korean allows for both Pronoun
> Adjective > Noun and Adjecitve > Pronoun > Noun word orders. The fact is that DNCs
in Korean allow for both orders as well.
(37) a.
b.
Ku/Wuli cen
phyencipcangtul
these/we former
chief.editors
‘These/We former chief editors’
Cen
ku/wuli
phyencipcangtul
former
these/we chief.editors
‘These/We former chief editors’
Korean
But I have shown that (37b) represents the basic word order while (37a) represents a
derived word order in which the pronoun is focused. This is exemplified in the paradigm
in (38): it is only Demonstrative/Pronoun > Adjective > Noun order that can be a
felicitous answer to the wh-question in (38A) that targets the demonstrative/pronoun.
(38) A:
B:
Etten
what/which
cen
former
phyencipcangtul-i
chief.editors-NOM
pass-ass-ni?
receive-PST-INT
‘Which former chief editors won the award?’
Ku/Wuli cen
phyencipcangtul-i
these/we former
chief.editors-NOM
‘These/We smart chief editors won (it).’
sang-ul
award-ACC
pass-ass-e.
receive-PST-DECL
Korean
171
B’:
#Cen
ku/wuli
phyencipcangtul-i
former
these/we chief.editors-NOM
‘These/We former chief editors won (it).’
pass-ass-e.
receive-PST-DECL
Given that Korean is a wh-scrambling language, I assume that (37a) is derived via
demonstrative/pronoun-scrambling. This way of viewing the word order variations within
DNCs and PNCs in Korean seems to lend further support to DP/CP parallelism.6
6
Treating the movement of the demonstrative to [Spec, DP] on par with wh-movement is not new.
Panagiotidis (2000) draws a parallel between the movement of the demonstrative and wh-movement. As
has been established in Chapter 3, the demonstrative may stay in situ or move to [Spec, DP] in Greek.
Panagiotidis liken this fact to the optional wh-movement in French. It is known that wh-phrases may or
may not move to [Spec, CP], as shown in (i) (from Mathieu 1999:441).
(i)
a.
b.
Tu
vois qui
ce
soir?
you see
who this evening
Qui vous voyez tqui
ce
soir?
who you see
this evening
‘Who are you seeing tonight?’
French
Panagiotidis’ idea that there is a parallel between the DP and the CP in terms of movement to their leftperipheries seems to be on the right track. However, the picture appears to be incomplete in the sense that
what he is comparing is wh-movement in French and demonstrative movement in Greek. It would make
more sense that DP is compared with CP within a given language. Also, incorporating PNCs, which are
argued to form a natural class with DNCs, into the discussion makes the picture more complete.
The claim for DP/CP parallelism presented in this sub-section seems to predict that French, which
allows optional wh-movement, should optionally allow for the demonstrative movement as well. At first
glance, this prediction appears to be completely ruled out because the demonstrative in French, irrespective
of the meaning, surfaces only in the left-most position—namely, its landing position if we assume the same
syntactic analysis for DNCs as developed in Chapter 3. The distribution of the demonstrative in French,
abstracting away from the collocation with a reinforcer, is fixed rather than flexible, as in (ii)0.
(ii)
a.
b.
Ce
livre
this/that
book
‘This/That book’
*Livre
ce
book
this/that
French
My claim for DP/CP parallelism considers not only the surface distributions of demonstratives and whphrases in a given language but also the interpretations associated with each element. Recall that in our
discussion of DP/CP parallelism based on the data from Greek, English and Korean, we were concerned
with the distribution of anaphoric demonstratives and wh-phrases whose potential referent(s) were
provided by the previously established discourse. With respect to French, it seems that specific
interpretation rather than anaphoric interpretation should be employed as a criterion to test the DP/CP
parallelism because the distribution of wh-phrases in French is associated with specificity. Mathieu (2004)
172
The summary of the distribution of the demonstrative and the pronoun within
DNCs and PNCs is given in the table below:
Greek
DP
CP
English
DP
CP
Korean
DP
CP
4.5
Demonstrative
Pronoun
wh-phrase
Demonstrative
Pronoun
wh-phrase
Demonstrative
Pronoun
wh-phrase
[Spec, DP/CP]
✓
✓
✓
✓
✓
✓
✓
✓
✓
in-situ
✓
✓
✓
✓
✓
Previous Analyses of Pronoun-Noun Constructions
This section addresses problems with two existing approaches to PNCs, namely, the
predication analysis and the head analysis. The former has been proposed for Greek
(Panagiotidis and Marinis 2011) and Japanese (Noguchi 1997; Furuya 2009), and the
latter for English (Postal 1969; Radford 2009; among others) and Iberian Spanish
(Longobardi 2008). I will review each approach and show that neither is tenable by
focusing on Greek.
argues that dislocated wh-phrases are associated with a specific reading whilst wh-phrases in situ have a
non-specific reading. Given that demonstratives are specific (as well as definite), the fact that they occur
only in the dislocated (i.e., the left-most) position in French comports well with the fact that specific whphrases surface in the dislocated position.
173
4.5.1
Predication Analysis
It has been proposed that the PNC in Greek is actually a case of determiner spreading and
thus of DP-internal predication (Panagiotidis and Marinis 2011). Determiner spreading is
a phenomenon in which multiple occurrences of a definite article within a single DP are
allowed when there is an adjective, as in (39) (cf. the single article in the DP in (40)).
(39) Determiner spreading:
Ta meghala ta
spitia
the big.ones the houses
‘the big house’
(40) Monadic DP:
Ta meghala
the big
‘the big house’
spitia
houses
Greek
Greek
Determiner spreading can receive either a predicative reading or a restrictive reading. (39)
can be paraphrased as ‘the big ones that have the property of being the houses’ on the
predicative reading, and as ‘the big ones of the houses’ on the restrictive reading. In
either case, the denotation of determiner spreading involves an interaction between the
two sets denoted by each sub-DP. Panagiotidis and Marinis (2011) proposes the structure
in (41), in which there is a subject-predicate relationship. That is, the first sub-DP ta
meghala ‘the big.ones’ is the subject and the second sub-DP ta spitia ‘the houses’ is the
174
predicate. As a result the highest DP in (42) denotes the intersection between the subject
and the predicate.
(41)
DP
DP
D’
ta meghala
‘the big.ones’
D°
ta
‘the’
NP
spitia
‘houses’
Panagiotidis and Marinis argue that close appositions, exemplified in (42a), are
instances of DP-internal predication, given that (42a) denotes an intersection of the two
sub-DPs, oi aetoi ‘the eagles’ and ta pulia ‘the birds’ (see also Lekakou and Szendrői
(2012)7). That is, (42a) means some things that are the eagles and the birds. Hence, (42b),
which is basically identical to (41), is adopted for the structure of (42a).
(42) a.
Oi aetoi
ta
pulia
the eagles
the birds
‘The eagles that are the birds’
Greek
Lekakou and Szendrői (2012) treats determiner spreading as a case of close apposition. On either view,
then, determiner spreading is equivalent to close appositions.
7
175
b.
DP
DP
D’
oi aetoi
‘the eagles’
D°
ta
‘the’
NP
pulia
‘birds’
Without a detailed discussion, Panagiotidis and Marinis claim that PNCs are
semantically identical to determiner spreading. They therefore argue that the PNC is also
a case of DP-internal predication, in which the pronoun is the subject and the following
noun a predicate, as illustrated in (43).
(43)
DP
D’
DP
emis
‘we’
D°
i
‘the’
NP
glossologi
‘linguists’
The hypothesis that the structure of PNCs involve DP-internal predication and
thus form a natural class with determiner spreading and close appositions immediately
brings some problems to the fore. That is, PNCs behave differently at least in three
aspects from determiner spreading and close appositions, the clear instances of DPinternal predication. First, determiner spreading allows for free word order between the
176
two sub-DPs, as in (44) (as was discussed in section 3.2.1.3). In contrast, PNCs require a
strict word order, i.e., Pronoun > Noun, as in (45).
(44) Determiner spreading:
a.
[Ta meghala]
the big.ones
b.
[Ta spitia]
the houses
[ta
the
[ta
the
(45) PNC:
a.
[Emis]
[i
we
the
b.
*[I glossologi]
the linguists
glossologi]
linguists
[emis]
we
spitia]
houses
meghala]
big.ones
Greek
Greek
Second, in a sentence, the predicate can agree with either sub-DP in determiner
spreading cases, as in (46). In contrast, when a PNC occurs in a sentence, the predicate
can only agree with the pronoun, as in (47).
(46) Close apposition:
Oi aetoi
ta
the eagles.M the
pulia
birds.N
ine
be.3PL.PRES
Greek
megaloprepa/megaloprepo.
majestic.M/majestic.N
‘The eagles that are the birds are majestic.’
(47) PNC:
Emis
i
glossologi piname/*pinane.
we
the linguists be.1PL.PRES.hungry/be.3PL.PRES.hungry
‘We linguists are hungry.’
Greek
177
These first two problems are also noted in Lekakou and Szendrői (2012).
Third, noun phrases containing determiner spreading do not allow for an
exception, which is a property of generic PNCs (see section 2.2.2.2). Thus, a continuation
that stipulates an exception, as in (48). (48) is semantically ill-formed since it is
contradictory. In contrast, PNCs can be generically interpreted and thus do allow for an
exception, as in (49).
(48) #Ta meghala
the big.ones
ta
the
spitia
houses
ine
be.3PL.PRES
akriva,
expensive
Greek
alla ohi afto to
spiti.
but not this the house
‘The big ones of the houses are expensive, but not this house.’
(where afto to spiti ‘this house’ refers to one of ta meghala ta spitia ‘the big ones of
the houses.’)
(49) Emis i
glossologi imaste
we the linguists be.1PL.PRES
‘We linguists are smart, but not me.’
exypni,
smart
alla ohi
but not
ego.
I
Greek
For the above reasons, therefore, I conclude that the predication analysis cannot be
maintained for Greek PNCs. Before inspecting the validity of the head analysis, I would
like to add a comment. If the predication analysis of PNCs is correct, we would expect
DNCs also to be an instance of predication, given the conclusion of Chapter 2 that the
two constructions form a natural class. However, this prediction is incompatible with the
conclusion of section 3.2.1.3 that DNCs also cannot be a predication in which the
demonstrative is the subject and the noun is the predicate. Hence, pursuing the
178
predication analysis of PNCs would render all the similarities between DNCs and PNCs
discussed in Chapter 2 merely coincidental.
4.5.2
Head Analysis
In the literature on English (Postal 1966, 1969; Pesetsky 1978; Abney 1987; Déchaine &
Wiltschko 2002; Panagiotidis 2002; Bernstein 2008; Radford 2009), it has been proposed
that the pronouns in PNCs are the head D of the construction. The motivation for the
analyses cited above is the fact that in English, the pronouns are in complementary
distribution with definite articles and demonstratives, both of which are treated as
members of D°in English (Jackendoff 1977). As seen in (50), only the variants in (50a)
are grammatical while all the other possible combinations of the three elements are not
acceptable.
(50) a.
b.
c.
d.
e.
The/These/We linguists
*The these/we linguists
*We the/these linguists
*These the/we linguists
*The these we linguists
f.
g.
h.
i.
j.
*The we these linguists
*These the we linguists
*These we the linguists
*We the these linguists
*We these the linguists
Assuming the DP hypothesis in the sense of Abney (1987) along with Ritter’s (1991)
NumP, the structure of (50a) would be (51).
179
(51)
DP
D°
NumP
we
the Num°
NP
these
linguists
The head analysis, however, cannot be extended to Greek. First, the head analysis
predicts that definite articles should be in complementary distribution with pronouns in
PNCs. As has been emphasized in Chapter 2, in Greek, a definite article is instead
obligatorily required in PNCs. The examples are repeated in (52).
(52) a.
b.
Emis/Esis i
glossologi
we/you
the linguists
‘We/You linguists’
*Emis/Esis
glossologi
we/you
linguists
Greek
Second, the head analysis of PNCs cannot capture the facts concerning pronounreinforcer constructions. As has been discussed in section 4.2.6, pronouns are taken to be
phrasal given the fact that they can form a modification relationship with a reinforcer, as
shown in (53).
(53) a.
b.
Emis
edho i
glossologi
we
here the linguists
‘We here/*there linguists.’
Esis
edho/eki i
glossologi
you
here/there the linguists
‘You here/there linguists.’
Greek
180
Let us put aside for now the issue of the obligatory occurrence of a definite article in
Greek PNCs, by following Stavrou (1995) in assuming that there is a functional head (Y°)
associated with the definite article, as in (54). Assuming the head analysis of PNCs, it is
not clear how the structure can provide a syntactic position for a reinforcer.
(54)
DP
D°
emis
‘we’
YP
i glossologi
‘linguists’
One could possibly assume an additional functional head (X°) whose role is to host a
reinforcer between DP and NumP, as shown in (55).
(55)
DP
D°
XP
emis
‘we’ X°
edho
‘here’
YP
glossologi
‘linguists’
Although this structure can capture the word order facts, that does not capture the
modifier-modifiee relationship between the reinforcer and the pronoun.
Third, if the head analysis is correct and thus pronouns are to be treated in the same
way as definite articles, it is then expected that pronouns should be able to spread within
181
DP in languages that allow determiner spreading. As exemplified in (56), however,
multiple occurrences of a pronoun within a single DP yield ungrammaticality.
(56) a.
b.
c.
Emis
i
exypni
glossologi
we
the smart
linguists
‘We smart linguists’
*Emis
i
emis exypni
glossologi
we
the we smart
linguists
*Emis
i
emis
exypni
emis glossologi
we
the we
smart
we linguists
Greek
Unsurprisingly but as expected given the analysis proposed here, the DNC in Greek
also requires a definite article, as has been discussed in Chapter 2 and reported in ample
literature. Also, demonstratives can be collocated with a reinforcer as well and this fact
straightforwardly repudiates the head analysis of DNCs by the same reasoning.
Moreover, demonstrative spreading is not attested in Greek, either. Like the PNC case in
(56), demonstratives cannot occur more than once within the same DP, as shown in (57).
(57) a.
b.
c.
Afti
i
exypni
glossologi
these
the smart
linguists
‘These smart linguists’
*Afti
i
afti exypni
glossologi
these
the these smart
linguists
*Afti
i
afti exypni
afti glossologi
these
the these smart
these linguists
Greek
These two facts further support the main hypothesis of this thesis that PNCs and DNCs
form a natural class and thus the pronoun and the demonstrative in such constructions
182
must be categorically distinct from the definite article.
To sum up, neither the predication analysis nor the head analysis is adequate for
PNCs in Greek. In what follows, I will probe the possibility of applying the proposed
analysis to PNCs in Standard English, Japanese, and Korean.
4.6
Revisiting English
So far, we have been focusing on PNCs in Greek. It has been shown that Greek crucially
differs from English in that it requires the occurrence of a definite article immediately
following the pronoun in PNCs. This fact was important evidence for rejecting the head
analysis (section 4.5.2). However, the head analysis does not cause a problem as far as
we are concerned with English PNCs. If the analysis presented in section 4.2 is correct—
at least for Greek—we now have two analyses for PNCs: one for Greek and the other for
English. Then, the task left for us is to determine whether we can have one single analysis
for PNCs that works for both Greek and English. We have already seen in section 4.5.2
that the head hypothesis for English is incompatible with Greek. In what follows, I will
provide examples from English that also speak against the head analysis and will
examine the possibility of extending the current analysis, which was constructed upon
Greek PNCs, to English.
The head analysis of PNCs seems to be compatible with English PNCs since it can
capture the fact that the pronoun encased in PNCs cannot co-occur with definite articles
and demonstratives, both of which are members of D°in English (Jackendoff 1977). This
183
fact renders useless for our purposes in this sub-section the argument based on
complementary distribution between definite articles, demonstratives, and pronouns in
English. However, the fact concerning the pronoun-reinforcer collocation can suggest a
different avenue to pursue for PNCs in English. We have assumed with Choi (2013) that
the relationship between pronoun and reinforcer is an instance of modification
relationship and thus the collocation, forming a constituent, stands as a phrase and
occupies [Spec, DP] but not D°. If so, the idea that PNCs are headed by the pronoun
cannot be maintained, at least in non-Standard English. The proposed analysis in which
pronouns are argued to occupy [Spec, DP] within PNCs is free from this problem.
In Greek and non-Standard English, which allow for the pronoun-reinforcer
collocation, the phrasal status of pronouns is rather easily identifiable. A remaining issue
is how to deal with Standard English, in which pronouns cannot be collocated with a
reinforcer. I will argue that even in Standard English, pronouns must occupy [Spec, DP]
of PNCs, based on the interpretive similarity between demonstratives and such, and
demonstratives in the context of DNCs and pronouns in the context of DNCs. Alexiadou
et al., (2007) argues that demonstratives in English precedes the head D and thus are in
the left periphery of the DP, based on the examples in (58).
(58) a.
b.
I did not expect this reaction.
I did not expect such a reaction.
(Alexiadou et al., 2007:108)
This in (58a) and such in (58b) are similar in that both “point to an element known from
the discourse context” (Alexiadou et al., 2007:108). What should be noted here is the fact
184
that such in (58b) precedes the indefinite article, or a. On the assumption that indefinite
articles are the head D, it straightforwardly follows that such is in [Spec, DP]. Given that
this and such are interpreted in a similar way, they conclude that this also occupies the
same position as such, namely, [Spec, DP].8 Following the line of reasoning above, I
conclude that pronouns also occupy [Spec, DP] of PNCs, given the conclusion that DNCs
and PNCs form a natural class (Chapter 2). This conclusion can be argued to hold for
both Standard and non-Standard English, since the fact regarding the similar
interpretation between this and such holds for both Standard and non-Standard Englishes.
4.7
Summary
In this chapter, I have developed an analysis for the base structure of PNCs and its
syntactic derivation. I have carried over the analysis of DNCs presented in Chapter 3, on
the basis of the parallels between DNCs and PNCs, the conclusion of Chapter 2. On the
current analysis, I proposed that pronouns merge in the specifier of a low functional
phrase, i.e., [Spec, dxP], and undergo movement and surface in [Spec, DP]. This shift in
the view of the structure of PNCs leads to a new view of how Agree works. All the phifeatures on the subject PNC controls the inflection on the agreeing predicate via
The question remains why demonstratives can co-occur with a definite article in Greek while they cannot
in English. This difference is probably due to the Doubly-filled DP filter effect. Alexiadou et al., (2007)
provides an account for why DP-initial demonstratives can co-occur with a definite article in Greek but not
in Spanish by parameterizing the filter. In Greek, the value for the parameter is set to be negative, and thus
filling [Spec, DP] and the head D at the same time is allowed. On the other hand, in Spanish, the value of
the parameter is set to be positive, and thus [Spec, DP] and the head D cannot be overtly realized at the
same time. If this is the case, the value for the filter in English is presumably set to be positive, like
Spanish, and thus demonstratives cannot co-occur with a definite article in English.
8
185
mediating D°. Moreover, positioning the valued pronoun feature of PNCs in [Spec, DP]
has a consequence for the distribution of person within the DP. Lastly, differing
movement patterns in the DP and CP domains, which seem to weaken the DP/CP
parallelism, in fact support it once we adopt the idea that Greek allows both dislocated
and in-situ wh-phrases for true information-seeking questions. I have then shown that the
two previous analyses—the head analysis and predication analysis—are not adequate for
Greek PNCs, and that the proposed account can be applied to English PNCs.
186
CHAPTER 5
PRO-DROP IN PRONOUN-NOUN CONSTRUCTIONS
5.1
Introduction
This chapter addresses the optionality of the pronoun in PNCs. As noted earlier in
Chapters 1 and 4, the pronouns of PNCs can be omitted, the discussion of which has been
delayed till this chapter. The main claim of this chapter is that such optional pronouns,
being an instance of pro-drop strictly regulated by agreement, are licensed by T°, with the
aid of D°which I call a ‘mediating licenser’, under some conditions. I will first show that
such optionality patterns with the well-known pro-drop phenomenon (section 5.2). By
investigating the syntactic environments in which the pronoun in question can remain
silent in Greek and PNC data drawn from other languages, I will define the conditions
which regulate such optional pronouns (section 5.3). I will then show that a syntactic
account straightforwardly follows if we combine the present syntactic analysis of PNCs
presented in Chapter 4 with an existing pro-based theory of null subject phenomena
(section 5.4). I will also justify the choice of pro-drop theory made above by showing that
a competing hypothesis about regular pro-drop cannot be extended to account for PNC
pro-drop. I discuss why the presence of a reinforcer blocks PNC pro-drop, and why
DNCs and PNCs behave differently with respect to the optionality of
demonstrative/pronoun. I finally summarize the findings of this chapter (section 5.5).
187
5.2
The Optionality of the Pronoun in Pronoun-Noun Constructions as Pro-Drop
The fact that the pronoun in PNCs is optional in Greek, and that when it does not surface
its content can be recovered from the information encoded by the verbal inflection, as
briefly mentioned in Chapter 4, resembles what we see in the well-known pro-drop
phenomenon1. As it turns out, Greek is, not coincidentally, one of the classical pro-drop
languages. Compare the Greek examples illustrated in (1) and (2).
(1)
a.
b.
c.
(2)
a.
b.
c.
[Emis]SUBJECT
imaste
we
be.1PL.PRES
‘We are smart.’
[Esis]SUBJECT
isaste
you.PL
be.2PL.PRES
‘You are smart.’
[Afti]SUBJECT
einai
they
be.3PL.PRES
‘They are smart.’
Imaste
exypni.
be.1PL.PRES
smart
‘We are smart.’
Isaste
exypni.
be.2PL.PRES
smart
‘You are smart.’
Einai
exypni.
be.3PL.PRES
smart
‘They are smart.’
exypni.
smart
Greek
exypni.
smart
exypni.
smart
Greek
This kind of phenomenon is also known as the ‘null subject phenomenon’. However, I will adopt the more
neutral term ‘pro-drop’ throughout this paper. The reason is that the former is limited to the empty category
licensing of a pronoun in the subject position of a clause, which the latter does not imply. As shown in
Chapter 4, the pronoun in PNCs is not the subject of a nominal predicate, nor is it by itself the subject of
the clause.
1
188
The sentences exemplified in (2) illustrate pro-drop in the clausal domain; the examples
in (1) are the sentences corresponding to (2) with an overt pronominal subject. In spite of
the lack of an overt subject in (2), the interpretation remains the same due to the rich
verbal inflection, from which discourse participants can draw information about the
missing subject pronoun. Under an appropriate context—when the subject receives an
emphatic interpretation—it is preferred to overtly express the pronoun with stress, along
with the verbal inflection, as shown in (1).2 Hence, the sentences in (1) are better used
when the pronominal subject receives stress, whereas the sentences in (2) are used in
other contexts.
Under current theories of agreement, it is generally assumed that verbs (or,
predicates) are inflected as a result of agreement with a DP (the subject, in most cases).
Taking this as given, one of the mainstream syntactic analyses of the pro-drop
phenomenon posits the existence of an empty DP category, pro, which controls subject-
2
A sample of comments on the the option of overtly expressing the pronoun subject from the literature:
“In many languages the distinction between the three persons is found not only in pronouns, but in
verbs as well, thus in Latin (amo, amos, amat), Italian, Hebrew, Finnish, etc. In such languages
many sentences have no explicit indication of the subject, and ego amo, tu amas is at first said only
when it is necessary or desirable to lay special stress on the idea, “I, thou.”” Jespersen (1924:213)
“Principle (5) [Avoid Pronoun] might be regarded as a subcase of a conversational principle of
deletion-up-to-recoverability, but there is some reason to believe that it functions as a principle of
grammar.” Chomsky (1981:65)
“Roughly speaking, the use of pronounced material is legitimate only when necessary to convey the
intended meaning, within the constraints of UG and of the particular grammar. This imples that,
given the existence of a zero pronominal option, in languages like Italian the overt form will be
limited to the cases in which it is necessary, i.e. when the pronominal subject, being focal or
contrastive must bear stress (evidently, the zero element cannot bear stress).” Rizzi (1989:73-74)
189
verb agreement (see Holmberg 2005, among many others).3 This view of the pro-drop
phenomenon is illustrated in (3). The subject position of each sentence is filled with pro.
(3)
a.
b.
c.
pro
Imaste
we
be.1PL.PRES
‘We are smart.’
pro
Isaste
you.PL
be.2PL.PRES
‘You are smart.’
pro
Einai
they
be.3PL.PRES
‘They are smart.’
exypni.
smart
Greek
exypni.
smart
exypni.
smart
What we have seen so far can be easily replicated to account for the omitted
pronoun within PNCs.4 Just as in the sentential domain, the PNC pronoun can be omitted
without affecting the interpretation, as already shown above. The verbal inflection
provides information about the missing element. Compare (4) and (5).
(4)
a.
b.
[Emis
i
glossologi]SUBJECT
we
the linguists
‘We linguists are smart.’
[Esis
i
glossologi]SUBJECT
you.PL
the linguists
‘You linguists are smart.’
imaste
be.1PL.PRES
exypni.
smart
isaste
be.2PL.PRES
exypni.
smart
Greek
See Alexiadou and Anagnostopolou (1998) and its variants for an alternative account of pro-drop
phenomena. I adopt pro-based theories without further discussion for now and in section 5.4.1. But, this
choice will be justified in section 5.4.3.
4
Since PNCs cannot be formed out of a third person plural pronoun and a noun, I will only include PNC
examples containing the first person and second person plural pronoun. However, the main thesis presented
in Chapter 3 is that PNCs must be treated on a par with DNCs, which in turn suggests the possibility of
dropping the demonstrative. I will, however, defer a discussion of the pro-drop issue related to DNCs to
section 5.4.4.2.
3
190
(5)
a.
b.
[I
glossologi]SUBJECT
the linguists
‘We linguists are smart.’
[I
glossologi]SUBJECT
the linguists
‘You linguists are smart.’
imaste
be.1PL.PRES
exypni.
smart
isaste
be.2PL.PRES
exypni.
smart
Greek
Second, when the pronoun receives an emphatic interpretation with accompanying
stress, the overt spell-out of the pronoun within PNCs is preferred, as has been shown in
(4). Again, omitting the pronoun portion of a PNC is allowed when the pronoun does not
receive an emphatic interpretation and stress.
Given the similarities between the unexpressed pronoun in both domains, I
conclude that the optional pronoun in PNCs constitutes an instance of pro-drop occurring
within the subject nominal. I have implicitly assumed so far the pro-based approach to
the pro-drop phenomenon (cf. fn. 3). We are then led to posit an empty category pro, the
source of verbal inflection, within the subject PNC, as schematized in (6) (where the
subjects are bracketed). Given that subject pro occupies the same position as its
corresponding overt pronoun, we assume that the position of pro within PNCs is exactly
the same position as that of the overt pronoun in PNCs. That is, pro appears within the
subject DP in the specifier position preceding the definite article.
(6)
a.
b.
[pro
I
glossologi]SUBJECT
we
the linguists
‘We linguists are smart.’
[pro
I
glossologi]SUBJECT
you.PL
the linguists
‘You linguists are smart.’
imaste
be.1PL.PRES
exypni.
smart
isaste
be.2PL.PRES
exypni.
smart
Greek
191
Given the parallel between the optional pronoun in PNCs and traditional pro-drop, I
conclude that the former is an instance of pro-drop. I will thus call this ‘PNC pro-drop’.
In the case of PNC pro-drop, just like the null subject phenomenon, we can observe
an asymmetrical tendency depending on the interpretation. We have seen that the PNC
can receive either deictic contrastive or generic interpretation. Among the two, the
generic reading is far more compatible with PNC pro-drop. In other words, the pronoun
of a deictic contrastive PNC has a great tendency to be overtly expressed. The same holds
true for pronoun-reinforcer constructions, which only permit a deictic contrastive
interpretation. Compare (7) and (8).
(7)
a.
b.
(8)
a.
b.
[Emis
we
edho i
here the
glossologi]SUBJECT
linguists
exypni.
smart
‘We here linguists are smart.’
[Esis
eki i
glossologi]SUBJECT
you.PL
there the linguists
‘You there linguists are smart.’
*[Edho
here
*[Eki
there
i
the
i
the
glossologi]SUBJECT
linguists
glossologi]SUBJECT
linguists
imaste
be.1PL.PRES
isaste
be.2PL.PRES
imaste
be.1PL.PRES
isaste
be.2PL.PRES
Greek
exypni.
smart
exypni.
smart
exypni.
smart
Greek
Once a reinforcer is added, as in (7), the deictic contrastive reading is reinforced. In such
cases, omitting the pronoun yields ungrammaticality, as in (8). This issue will be taken up
in more detail in section 5.4.4.1.
192
What is naturally expected from the above conclusion that PNC pro-drop is related
to traditional pro-drop is that non-pro-drop languages such as English will disallow PNC
pro-drop. Consider the example under (9).
(9)
Linguists are smart.
≠ ‘We/You linguists are smart.’
The sentence in (9) can only receive an existential or generic reading. However, when the
intended meaning for the subject is a PNC interpretation (on any reading), the sentence is
ungrammatical; the lack of rich verbal inflection in English prevents recovery of the
content of the missing information conveyed by the pronoun.
Even though we have reached the conclusion that the optional pronoun in PNCs is
an instance of pro-drop, there is a nontrivial difference between PNC pro-drop and
classical pro-drop, with respect to what can be dropped. In both cases, what is omitted is
limited to the pronoun; however, the two phenomena crucially differ if we consider the
grammatical function of the target pronoun. Classical pro-drop targets the entire subject,
which corresponds to a pronoun. (This is the reason why it is also known as the nullsubject phenomenon.) By contrast, PNC pro-drop targets only a portion of the subject –
namely, the pronoun portion of the PNCs, which is not obviously the subject of the clause.
Given that PNC pro-drop differs from the classical pro-drop in this respect despite
the similarities shared by the two phenomena, PNC pro-drop needs to be further
examined. This task will be carried out in what follows.
193
5.3
Defining Conditions on PNC Pro-Drop
PNC pro-drop is not freely allowed. Instead, there are certain conditions to which the prodrop in question is subject. This is not surprising at all, given that regular pro-drop—the
null subject phenomenon—is only possible under certain circumstances as well. For
instance, in languages such as Italian, Greek, and Spanish, a null subject is allowed due to
the rich inflection on the verb,5 whereas languages such as English, which lack rich
verbal inflection, typically disallow such pro-drop. To make a long story short, two
conditions must be met for PNC pro-drop in a given language. Informally speaking, first,
the language must be a Greek-type pro-drop language, on the one hand, and, second, a
definite article must co-occur with the pronoun in PNCs. In what follows, I will explicate
these two conditions on the basis of the facts concerning the (un)availability of PNC prodrop from a cross-linguistic perspective. Not only English and Greek but also Chinese,
German, Italian, Korean, Russian, Spanish, and the like will be considered.
5.3.1
The Dependency of PNC Pro-Drop on the Clausal Domain Pro-Drop
We have established in section 5.2 that the unexpressed pronoun in PNCs involves prodrop within the subject nominal. Even though PNC pro-drop takes place within the
nominal, we cannot determine the conditions on dropping the pronoun in PNCs if we
This is actually known to be an inadequate characterization of pro-drop languages. Some languages show
rich verbal inflection just like the above-mentioned languages, but do not allow pro-drop. Others allow prodrop in the complete absence of inflection. This issue will be addressed in section 5.3.2.2.
5
194
limit our attention to the PNC itself. In other words, PNC pro-drop is closely related to
the pro-drop in the clausal domain.
In the first place, the crucial evidence that shows that PNC pro-drop is dependent
on the pro-drop in the clausal domain comes from the consideration of the impossibility
of PNC pro-drop when there is no predicate expressed. The conclusion is that a predicate
must be present in order for PNC pro-drop to be allowed. That is, the pro-drop under
discussion is not allowed in isolation. This point is attested by the examples in (10) and
(11) below.
(10) Emis/Esis
i
we/you
the
‘We/You linguists’
glossologi
linguists
(11) I
glossologi
the linguists
‘The linguists’
≠ ‘We/You linguists’
Greek
Greek
The examples in (10) illustrate PNCs in isolation, not coupled with a predicate. If we
omit the pronouns from the examples in (10), we cannot maintain the PNC interpretation,
as in (11). The example in (11) is only grammatical when the intended meaning is ‘the
linguists’; it means neither ‘we linguists’ nor ‘you linguists’.
The mere presence of a predicate does not suffice to license PNC pro-drop,
however. PNC pro-drop is observed only when the PNC is in an agreement relationship
with the predicate. The reason for this must be that recovering the content of the missing
195
element, the pronoun, is not possible when agreement is absent. If so, it is expected that
in Greek—in which the predicate only agrees with the subject, but not with the object—
the pronoun embedded in PNC is not omissible if the PNC is in the object position. The
examples in (12) and (13) clearly illustrate this point.
(12) a.
b.
I
the
Anthi
Anthi
epenese
praise.3SG.PST
glossologus]OBJECT.
linguists
‘Anthi praised us linguists.’
I
Anthi
epenese
the Anthi
praise.3SG.PST
‘Anthi praised us linguists.’
(13) I
Anthi
epenese
the Anthi
praise.3SG.PST
= ‘Anthi praised the linguists.’
≠ ‘Anthi praised us linguists.’
≠ ‘Anthi praised you linguists.’
[emas
us
tus
the
[esas
you.ACC
tus
the
Greek
glossologus]OBJECT.
linguists
[tus glossologus]OBJECT.
the linguists
Greek
As mentioned above, the predicate does not agree with an object in Greek. Thus, the
absence of the pronouns in (12) results in the sentence in (13), whose object cannot be
interpreted as pro-dropped PNC. They are grammatical only when the intended meaning
is ‘Anthi praised the linguists’.
The profile of PNC pro-drop described so far clearly demarcates PNC pro-drop
from the superficially similar case of optional possessive pronouns, which is worth
196
remarking on. There are some languages that allow for omission of possessive pronouns.
One such language is Turkish. Consider the examples in (14).
(14) a.
b.
c.
(Ben-im) ev-im
1SG-1.GEN house-1
‘my house’
(Biz-im) ev-im-iz
1PL-1.GEN house-1-PL
‘our house’
(Sen-in)
ev-in
2SG-GEN
house-2
‘your(SG) house’
d.
e.
f.
(Siz-in)
ev-in-iz
Turkish
2PL-GEN
house-2-PL
‘your(PL) house’
(On-un)
ev-i
3SG-GEN
house-3
‘His/Her house’
(On-lar-ın)
ev-ler-i
3SG-PL-GEN
house-PL-3
‘Their house’
As indicated by the parentheses surrounding the possessive pronouns in (14), they are all
optional. This type of pro-drop in Turkish is, however, different from PNC pro-drop, in
that the former is not dependent on the presence of the predicate. That is, the possessivepronoun-drop is licensed DP-internally, regardless of the DP-external environment. The
reason why such pro-drop is possible in Turkish is presumably due to the rich agreement
morphology on possesses. Note that the inflectional suffixes are in bold. This being said,
we will leave this phenomenon aside.
In summary, the preliminary condition on PNC pro-drop, established by the Greek
facts and the contrastive behavior between Greek and English, is summarized under (15).
(15) PNC pro-drop generalization (to be revised):
PNC pro-drop is allowed when the predicate agrees with the PNC in a pro-drop
language.
197
5.3.2
Not All Pro-Drop Languages Allow PNC Pro-Drop
The above condition seems to suffice at this point. As it will turn out, however, it needs
to be refined due to the loose definition of what we have been calling ‘pro-drop
languages’, which need to be sub-divided into four distinct sub-groups. Before we
proceed to discuss how we want to update the condition, a brief introduction to the
different types of pro-drop languages is needed. A reexamination of the condition in (15)
will then follow.
5.3.2.1 Typology of Pro-Drop Languages
The null subject phenomenon has been a primary interest of linguists, and research over
many years has discovered that the so-called ‘null subject’ or ‘pro-drop’ languages do not
form a homogeneous group, since not all pro-drop languages behave in the same ways.
Instead, it has been suggested that in order to achieve better descriptive adequacy, we
need to break down the pro-drop languages into more elaborated subclasses. I distinguish
four different types of pro-drop languages following Roberts and Holmberg (2010):
consistent pro-drop languages, expletive pro-drop languages, partial pro-drop languages,
and radical pro-drop languages. Let us characterize each type one by one.6
6
Roberts and Holmberg (2010:6-12), to be precise, uses the following terms for each type of pro-drop
languages: ‘consistent null-subject languages’, ‘expletive/semi null subjects’, ‘discourse/radical pro-drop’,
and ‘partial null-subject languages’.
198
To begin with, consistent pro-drop languages are represented by Italian, Greek,
Spanish, etc. Consistent pro-drop languages are characterized by rich agreement
morphology on the verb and unexpressed subject pronouns in all persons in all tenses.
The representative examples have been provided above in (3) for Greek, reproduced
below:
(16) a.
b.
c.
(Emis)
imaste
we
be.1PL.PRES
‘We are smart.’
(Esis)
isaste
you.PL
be.2PL.PRES
‘You are smart.’
(Afti)
einai
they
be.3PL.PRES
‘They are smart.’
exypni.
smart
Greek
exypni.
smart
exypni.
smart
The second type is partial null subject languages, represented by Finnish, Hebrew,
Russian, etc. These languages allow first and second person pronouns to be dropped in
any finite context, while third person pronouns can only be dropped under restricted
conditions.7 The Modern Hebrew (Hebrew) examples in (17) illustrate this point (from
Gutman (2004:464)). 8
7
Finnish is a partial pro-drop language, as illustrated in (i) (from Holmberg (2005:539)).
(i)
a.
b.
(Minä)
puhun
I
speak.1SG
‘I speak English.’
(Sinä)
puhut
you.SG
speak.2SG
‘You speak English.’
englantia.
English
englantia.
English
Finnish
199
(17) a.
b.
c.
pro
Nixshalti
ba-mivxan
failed.1SG
in-the.test
‘I failed the history test.’
pro Nixshalta
ba-mivxan
failed.2SG.M
in-the.test
‘You failed the history test.’
*pro Nixshal/Nixshela ba-mivxan
failed.3SG.M/F in-the.test
‘He/She failed the history test.’
be-historia.
in-history
Hebrew
be-historia.
in-history
be-historia.
in-history
First and second person pronouns can be omitted, as shown in (17a) and (17b), while
third person pronouns cannot, as shown in (17c).
The third type is expletive pro-drop languages, represented by German, some
varieties of Dutch, and some creoles such as Cape Verdean, Haitian, etc. In this type, an
expletive pronoun can be suppressed, while referential pronouns must be overtly
c.
d.
e.
f.
*(Hän)
puhut
he/she
speak.3SG
‘He/She speaks English.’
(Me)
puhumme
we
speak.1PL
‘We speak English.’
(Te)
puhutte
you.PL
speak.2PL
‘You speak English.’
*(He)
puhuvat
they
speak.3PL
‘They speak English.’
englantia.
English
englantia.
English
englantia.
English
englantia.
English
The ban on dropping the third person pronouns ((ic) and (if) disappears, when the pronoun in question is
bound by a higher argument, as in (ii) (from Holmberg 2005:539).
(ii)
Pekkai
väittää
[että häni/j/∅i/*j
Pekka
claims
that he
‘Pekka claims that he speaks English well.’
puhuu
speaks
englantia
English
hyvin]. Finnish
well
The overt pronoun hän ‘he’ can be either bound by the higher argument Pekka or not. By contrast, the
pronoun can be unexpressed only when it refers to the higher argument Pekka.
8
Hebrew lacks verbal inflection for phi-features in the present tense, and thus no pronouns can be omitted
in the tense.
200
expressed. Consider the German examples in (18) (from Roberts and Holmberg (2010:8);
originally from Cardinaletti (1990:5-6)).9
(18) a.
b.
Gestern
wurde
(*es) getanzt.
yesterday was
it
danced.
‘Yesterday there was dancing.’
Gestern
war
*(es) geschlossen.
yesterday was
it
closed.
‘Yesterday it was closed.’
German
In (18a), the expletive pronoun es ‘it’ must be suppressed. By contrast, the referential
pronoun es ‘it’ in (18b) must be expressed. Likewise, first and second person pronouns
cannot be omitted in German, as shown in (19).
(19) a.
*(Wir)
sind
we
be.1PL
‘We are smart.’
intelligent.
smart
German
In (19a), the omission of the expletive pronoun es ‘it’ is obligatory, rather than optional. However, this
obligatory suppression of the expletive is specific to the construction in (19a), i.e., to impersonal passive
constructions in German. In this construction, an expletive is only allowed to occur in Fore Field, typically
considered to correspond to the CP domain. As shown in (i), the occurrence of an expletive is allowed even
in an impersonal passive sentence, if the position of the expletive belongs to Fore Field.
9
(i)
Es
wurde
getanzt.
it
was
danced.
‘There was dancing.’
German
The fact that the use of an expletive in impersonal passive constructions depends on the position of the
expletive does not hold for other constructions. For instance, in weather sentences, an expletive is required
regardless of its position:
(ii)
Getern
regnete
yesterday
rained
‘Yesterday, it rained.’
*(es).
it
German
201
b.
*(Ihr)
seid
you.PL
be.2PL
‘You are smart.’
intelligent.
smart
The fourth type is radical pro-drop languages. Languages of this type include
Chinese, Japanese, Korean, Thai, etc, which are characterized by a lack of agreement
morphology on the verb. Radical pro-drop languages allow for any argument—not
limited to just subjects—to be dropped. Consider (20) (from Huang (1984:533)).
(20) Speaker A: Zhangsan kanjian
Lisi
Zhangsan see
Lisi
‘Did Zhangsan see Lisi?’
Speaker B: a.
Ta kanjian
ta
he see
he
‘He saw him.’
b.
e
Kanjian
ta
he see
he
‘He saw him.’
c.
Ta kanjian
e
he see
he
‘He saw him.’
d.
e
Kanjian
e
he see
he
‘He saw him.’
le
LE
ma?
Chinese
Q
le.
LE
le.
LE
le.
LE
le.
LE
As shown in (20), either the subject pronoun, as in (20b), or the object pronoun, as in
(20c), can be omitted. Or, both the subject and the object can be dropped simultaneously,
as in (20d). The typology of pro-drop languages is as summarized in the table below:
202
Types
Consistent
pro-drop
Languages
Italian, Greek, Spanish,
Turkish10
Properties
Partial
pro-drop
Finnish, Hebrew,
Russian
Expletive
pro-drop
German, some varieties
of Dutch and Afrikaans
Definite subject pronouns can be
dropped; the verbal inflection is rich.
Only first and second person pronouns
can be dropped; dropping third person
pronouns is highly restricted.
Only expletives can be dropped;
personal pronouns cannot be dropped.
Radical
pro-drop
Chinese, Japanese,
Korean, Thai
Any argument can be dropped,
regardless of its grammatical function.
5.3.2.2 Redefining the First Condition
Since we have seen above the (un)availability of PNC pro-drop in English (a non-prodrop language) and Greek (a consistent pro-drop language), let us now examine the
pattern of PNC pro-drop in the other three types of pro-drop languages one by one, the
result of which will help us to revise the condition in (15). Before proceeding to
investigate each type of language, I would like to acknowledge that a wider study would
be required to confirm the conclusion of this sub-section. For practical reasons, I only
look into 11 languages.
Roberts and Holmberg classify Turkish as a consistent pro-drop language (see also Kornfilt (2003)). As
they recognized, however, Öztürk (2001, 2008) argues that Turkish is a radical pro-drop language. Since it
is immaterial to our discussion, I simply take the former stance.
10
203
5.3.2.2.1 PNC Pro-Drop in Partial Pro-Drop Languages
Let us begin with scrutiny of the relevant data from Hebrew, which is, as shown above, a
partial pro-drop language. As Hebrew allows for first and second person pronouns to be
unexpressed, we would expect, given (15), to see PNC pro-drop in the language for at
least these pronouns. This is, however, not borne out, as illustrated by the contrasting
grammaticality of (21) and (22).11
(21) a.
b.
Anachnu ha-talmidim
we
the-pupils
nitstarekh
have.to.FUT.1PL.M
la’avod
work
kasheh
hashanah.
hard
this.year
‘We pupils have to work hard this year.’
Atem
ha-talmidim
titstarkhu
you.PL
the-pupils
have.to.FUT.2PL.M
la’avod
work
Hebrew
kasheh
hard
hashanah.
this.year
‘You pupils have to work hard this year.’
11
Interestingly, the PNC may or may not contain a definite article in Hebrew. Hence, (i) is grammatical.
(i)
Anachnu
talmidim
nitstarekh
we
pupils
have.to.FUT.1PL.M
‘We pupils have to work hard this year.’
la’avod
work
kasheh
hard
hashanah.
this.year
Its absence/presence correlates with the two different interpretations of the PNC. With a definite article, the
PNC receives a generic interpretation; without a definite article, the PNC receives a deictic contrastive
interpretation. It is not clear where this compositional difference with regard to the PNC in this language
stems from. I leave this issue for future investigation.
204
(22) a.
b.
*Ha-talmidim
the-pupils
nitstarekh
have.to.FUT.1PL.M
la’avod
work
kasheh
hard
hashanah
this.year
*Ha-talmidim
the-pupils
titstarkhu
have.to.FUT.2PL.M
la’avod
work
kasheh
hard
Hebrew
hashanah.
this.year
Russian, another partial pro-drop language, also forbids PNC pro-drop, leading to the
conclusion that PNC pro-drop is not available in partial pro-drop languages.
(23) a.
b.
My
we
lingvisty
linguists
izucha-em
study-1PL.PRES
structur-u
structure-ACC
Russian
predlozhenij
nauchnym sposob-om
sentences
scientific way-INST
‘We linguists study the structure of languages in a scientific way.’
Vy lingvisty izucha-ete
structur-u
predlozhenij
you linguists study-2PL.PRES structure-ACC
sentences
nauchnym sposob-om
scientific way-INST
‘You linguists study the structure of languages in a scientific way.’
(24) a.
b.
*Lingvisty izucha-em
linguists study-1PL.PRES
structur-u
structure-ACC
predlozhenij
sentences
Russian
nauchnym sposob-om
scientific way-INST
Intended: ‘We linguists study the structure of languages in a scientific way.’
*Lingvisty izucha-em
structur-u
predlozhenij
linguists study-1PL.PRES structure-ACC
sentences
nauchnym sposob-om
scientific way-INST
Intended: ‘You linguists study the structure of languages in a scientific way.’
205
To summarize, the Hebrew and Russian facts discussed in this section suggest that partial
pro-drop languages disallow PNC pro-drop.
5.3.2.2.2 PNC Pro-Drop in Expletive Pro-Drop Languages
What about PNC pro-drop in expletive null subject languages? Let us take the German
language for example. Compare (25) and (26).
(25) a.
b.
(26) a.
b.
Wir
Linguisten sind
we
linguists be.1PL
‘We linguists are smart.’
Ihr
Linguisten bist
you.PL
linguists be.2PL
‘You linguists are smart.’
intelligent.
smart
German
intelligent.
smart
*Linguisten
sind
intelligent.
linguists
be.1PL
smart
Intended: ‘We linguists are smart.’
*Linguisten
bist
intelligent.
linguists
be.2PL
smart
Intended: ‘You linguists are smart.’
German
As shown in (26), neither sentence can receive a PNC reading. Given that first and
second person pronoun subjects cannot be suppressed in expletive languages, despite the
rich verbal agreement morphology, the fact that PNC pro-drop is disallowed in German is
not unexpected.
206
5.3.2.2.3 PNC Pro-Drop in Radical Pro-Drop Languages
In radical pro-drop languages, arguments—regardless of their grammatical function (i.e.,
subject, object, etc)—can be radically dropped, as has been illustrated above in (23) in
section 5.3.2.1. This characteristic of languages of this type might leads us to expect that
PNC pro-drop would be quite freely allowed in such languages, under similar discourse
conditions to regular pro-drop. The result is, however, quite the opposite: PNC pro-drop
is disallowed in radical pro-drop languages. Let us compare the Chinese examples in (27)
and (28).
(27) a.
b.
Women
yuyanxuejia
ying-le
we
linguists
win-PER
‘We linguists won the award.’
Nimen
yuyanxuejia
ying-le
you
linguists
win-PER
‘You linguists won the award.’
(28) *Yuyanxuejia
ying-le
diyiming.
linguists
win-PER first.prize
Intended: ‘We/You linguists won the award.’
diyiming.
first.prize
Chinese
diyiming.
first.prize
Chinese
As shown in (27) and (28), the PNC reading is unavailable without the overtly expressed
pronoun. The sentence in (28) is only ambiguous between the definite reading (‘The
linguists won the award’) and the existential reading (‘Some linguists won the award’).
Exactly the same pattern can be reproduced in other radical pro-drop languages such as
Japanese, as in (29)-(30); adapted from Furuya (2009:43)), and Korean, as in (31)-(32).
207
(29) a.
b.
Watasitati zyosei-wa
mina
we
woman-TOP
all
‘We women are all full-timers.’
Anatatati zyosei-wa
mina
you
woman-TOP
all
‘You women are all full-timers.’
seisyasin
full.timer
desu.
seisyasin
full.timer
desu.
COP
COP
(30) *Zyosei-wa
mina seisyasin desu.
woman-TOP
all full.timer COP
Intended: ‘We/You women are all full-timers.’
(31) a.
b.
Wuri
enehakcatul-un
we
linguists-TOP
‘We linguists are smart.’
Nehuy
enehakcatul-un
you
linguists-TOP
‘You linguists are smart.’
Japanese
Japanese
ttokttokha-ta.
smart-DECL
Korean
ttokttokha-ta.
smart-DECL
(32) *Enehakcatul-un ttokttokha-ta.
linguists-TOP
smart-DECL
Intended: ‘We/You linguists are smart.’
Korean
We have seen that PNC pro-drop is not allowed in every type of pro-drop
languages. Instead, only one type of pro-drop language allows PNC pro-drop, namely
consistent pro-drop languages. This fact must be taken into account in revising the
condition in (15) so that non-consistent pro-drop languages as well as non-pro-drop
languages like English can be properly ruled out. Our tentative conclusion is summarized
in (33).
208
(33) PNC pro-drop generalization (to be revised):
PNC pro-drop is allowed when the predicate agrees with the PNC in consistent prodrop languages.
5.3.3
The Dependency of PNC Pro-Drop on the Definite Article
Given the above discussion, it is very tempting to conclude that PNC pro-drop should be
allowed in any consistent pro-drop language. It is, however, not the case that all
consistent pro-drop languages allow for PNC pro-drop. This suggests the existence of
another factor which regulates PNC pro-drop. Unlike the previously identified conditions
in (33), this factor is not concerned with the status of pro-drop in the clausal domain, but
rather with the structure of the PNC itself.
The additional restriction can be identified by comparing PNCs in Italian with
those in Greek. It is well known that Italian, another consistent pro-drop language, allows
unexpressed subject pronoun of sentences, as shown in (34) (from Haegeman 1991).
(34) a.
b.
c.
(Io)
parlo.
I
speak.1SG
‘I speak.’
(Tu)
parli.
you.SG
speak.2SG
‘You speak.’
(Lei)
parla.
she
speak.3SG
‘She speaks.’
d.
e.
f.
(Noi)
parliamo.
we
speak.1PL
‘We speak.’
(Voi)
parlate.
you.PL
speak.2PL
‘You speak.’
(Loro)
Parlano.
they
speak.3PL
‘They speak.
Italian
209
However, a pronoun embedded within a subject PNC must be overtly spelled-out, even if
there is a predicate with rich agreement morphology coindexing the pronoun, as
exemplified in (35) and (36).
(35) a.
b.
(36) a.
b.
Noi linguisti
siamo
we linguists be.1PL.PRES
‘We linguists are smart.’
Voi linguisti
siete
you linguists be.2PL.PRES
‘You linguists are smart.’
intelligenti.
smart
Italian
intelligenti.
smart
*Linguisti siamo
intelligenti.
linguists be.1PL.PRES
smart
Intended: ‘We linguists are smart.’
*Linguisti siete
intelligenti.
linguists be.2PL.PRES
smart
Intended: ‘You linguists are smart.’
Italian
As shown in (36), the sentences under (35) become unacceptable if the pronouns are not
overtly expressed.12
Where then does this different behavior of PNC pro-drop between Greek and
Italian come from? I seek an answer in the different surface configurations of the PNC in
the two languages. Italian is crucially different from Greek in that Italian does not tolerate
the overt presence of a definite article within PNCs. This point is illustrated in the
grammaticality contrast between (35) and (37).
Note that the distribution of bare plural noun phrases in Italian is highly restricted (Longobardi 1994; see
also Alexopoulou and Folli 2010).
12
210
(37) a.
b.
c.
d.
*Noi
we
*I
the
*Voi
you
*I
the
i
the
noi
we
i
the
voi
you
linguisti
linguists
linguisti
linguists
linguisti
linguists
linguisti
linguists
siamo
be.1PL.PRES
siamo
be.1PL.PRES
siete
be.2PL.PRES
siete
be.2PL.PRES
intelligenti.
smart
intelligenti.
smart
intelligenti.
smart
intelligenti.
smart
Italian
Combining a pronoun with a bare plural noun (which is presumably embedded by a
covert D°, under the analysis developed in Chapter 4), exemplified in (35), is the only
option for forming a PNC in Italian. The addition of a definite article either before or
after the pronoun gives rise to ungrammatical sentences, as shown in (37). This structural
difference between Greek and Italian PNCs leads to the conclusion that an overt definite
article in PNCs is necessary to allow PNC pro-drop.
Spanish and Turkish (consistent pro-drop languages) pattern with Greek and Italian,
respectively, in terms of the surface make-up of their PNC. That is, Spanish PNCs must
contain a definite article, as illustrated in (38). Without the article, the sentences become
ungrammatical, as in (39).
(38) a.
b.
Nosotros los lingüistas
we
the linguists
‘We linguists are smart.’
Vosotros los lingüistas
you
the linguists
‘You linguists are smart.’
somos
be.1PL
listos.
smart
sois
be.2PL
listos.
smart
Spanish
211
(39) a.
b.
*Nosotros lingüistas somos
listos.
we
linguists be.1PL
smart
Intended: ‘We linguists are smart.’
*Vosotros lingüistas sois
listos.
you
linguists be.2PL
smart
Intended: ‘You linguists are smart.’
Spanish
As expected, Spanish allows PNC pro-drop, as in (40).
(40) a.
b.
Los lingüistas somos
the linguists be.1PL
‘We linguists are smart.’
≠ ‘The linguists are smart.’
Los lingüistas sois
the linguists be.2PL
‘You linguists are smart.’
≠ ‘The linguists are smart.’
listos.
smart
Spanish
listos.
smart
In the case of Turkish, an article-less language, PNCs can never contain a definite article;
it is impossible to construct an ungrammatical version of (41) due to the lack of definite
article in Turkish.
(41) a.
b.
Biz
dilbilimciler
we
linguists
‘We linguists are smart.’
Siz
dilbilimciler
you
linguists
‘You linguists are smart.’
akıllıyızdır.
smart.1PL
Turkish
akıllısınızdir.
smart.2PL
And, if we omit the pronoun, a PNC interpretation becomes unavailable, despite the rich
agreement on the predicate, as in (42).
212
(42) a.
b.
*Dilbilimciler
akıllıyızdır.
linguists
smart.COP.1PL.EPIS
Intended: ‘We linguists are smart.’
*Dilbilimciler
akıllısınızdir.
linguists
smart.COP.2PL.EPIS
Intended: ‘You linguists are smart.’
Turkish
The above Italian and Spanish facts strongly suggest that we need a separate rule
that distinguishes two types of consistent pro-drop languages: ones that allow PNC prodrop and the others that disallow PNC pro-drop. The variation regarding the structure of
PNCs observed in Italian, Turkish, Greek and Spanish straightforwardly lets us amend
our condition so as to exclude Italian and Turkish, the consistent pro-drop languages
whose PNCs do not take a definite article. The updated generalization is stated in (43)
below.
(43) PNC pro-drop generalization (final version):
PNC pro-drop is allowed only when
a.
the predicate agrees with the PNC in consistent pro-drop languages, AND
b.
the PNC contains a definite article.
As indicated by AND, the two constraints must be satisfied in a conjunctive manner: PNC
pro-drop is not allowed unless both of the conditions are satisfied. Greek and Spanish are
the cases in which both (43a) and (43b) are satisfied, whereas Italian and Turkish satisfy
only (43a) but not (43b). It is further supported by Hebrew that both conditions must be
satisfied. Among the languages we have discussed so far, Hebrew is a proper testing
213
ground. As we saw in (21) and (22), reproduced in (44) and (45), even though Hebrew
PNCs can contain a definite article, as in (44), PNC pro-drop is disallowed, as in (45).
However, Hebrew is only a partial pro-drop language, not a consistent pro-drop language,
so it does not satisfy condition (43a), though it does satisfy (43a).
(44) a.
b.
(45) a.
b.
Anachnu ha-talmidim
we
the-pupils
nitstarekh
have.to.FUT.1PL.M
la’avod
work
kasheh
hashanah.
hard
this.year
‘We pupils have to work hard this year.’
Atem
ha-talmidim
titstarkhu
you.PL
the-pupils
have.to.FUT.2PL.M
hashanah.
this.year
‘You pupils have to work hard this year.’
la’avod
work
*Ha-talmidim
the-pupils
nitstarekh
have.to.FUT.1PL.M
la’avod
work
kasheh
hard
hashanah
this.year
*Ha-talmidim
the-pupils
titstarkhu
have.to.FUT.2PL.M
la’avod
work
kasheh
hard
Hebrew
kasheh
hard
Hebrew
hashanah.
this.year
Other languages in my sample do not qualify as a testing ground for this
‘conjunctive satisfaction’ condition. Our other partial pro-drop language (Russian), as
well as the radical pro-drop languages discussed above (Chinese, Japanese, Korean, and
Turkish), are article-less languages, and thus forming PNCs with a definite article is in
principle impossible. In the case of German, an expletive pro-drop language, PNCs do
not feature a definite article, as shown in (46).
214
(46) a.
b.
c.
d.
*Wir
we
*Die
the
*Ihr
you.PL
*Die
the
die
the
wir
we
die
the
Ihr
you.PL
Linguisten
linguists
Linguisten
linguists
Linguisten
linguists
Linguisten
linguists
sind
be.1PL
sind
be.1PL
bist
be.2PL
bist
be.2PL
intelligent.
smart
intelligent.
smart
intelligent.
smart
intelligent.
smart
German
Thus, it appears that the additional condition on the structure of PNCs does not make any
conflicting predictions for PNC pro-drop at least in the languages discussed so far.
5.3.4
Interim Summary
In this section, I have shown that the optional pronoun in PNCs is an instance of pro-drop.
Also, I have identified two conditions, which must be satisfied conjunctively in order for
PNC pro-drop to be allowed. The conditions are as follows: (i) the PNC must occupy the
position which the predicate agrees with in a consistent pro-drop language; (ii) the PNC
must contain a definite article in the language. Among those data discussed above, it is
only Greek and Spanish that satisfy all the two conditions, and as a result they are the
only two languages that allow PNC pro-drop. The table below summarizes the typology
of PNC pro-drop:
215
Languages
Greek
Spanish
Italian
Turkish
Hebrew
Russian
German
Chinese
Japanese
Korean
English
Pro-Drop Type
Consistent
Consistent
Consistent
Consistent
Partial
Partial
Expletive
Radical
Radical
Radical
Non-pro-drop
Definite Article in PNC
PNC pro-drop
✓
✓
✓
✓
✓ or
The above table seems to be far from complete. There are some logically possible
languages which have not been exemplified in our sample:
i. Expletive pro-drop languages whose PNC contains a definite article
ii. Radical pro-drop languages whose PNC contains a definite article
iii. Non-pro-drop languages whose PNC contains a definite article
Even though (ii) is logically possible, it does not seem that such languages exist given
that one salient property shared by radical pro-drop languages—at least those discussed
above—is that they lack articles.13
This is a tentative conclusion, and thus it might be hasty to conclude that the type (ii) languages do not
exist. For instance, Turkish can be arguably classified as a radical pro-drop language (cf., fn. 10), even
though Turkish does not share the hallmark property of radical pro-drop languages, namely the lack of rich
verbal agreement. This border-line case suggests that there can be an exception to the typical property of
radical pro-drop languages. That is, there can be a radical pro-drop language which has definite articles its
lexical inventory and employs it in forming PNCs. Even if so, however, the generalization in (43) still
13
216
What is clear at this point is that the PNC Pro-Drop Generalization in (46) makes a
very strong prediction, which is that PNC pro-drop is not an available option in any of the
three additional types of logically possible languages that have not been examined here,
since none of them satisfies both of the two sub-conditions. In the next section, I will
attempt a syntactic account of PNC pro-drop assuming that the PNC pro-drop
generalization is correct as stated.
5.4
Analysis: Mediated Pro-Drop
In this section, I propose a syntactic account for PNC pro-drop. Specifically, I show that
the syntax of PNCs proposed in Chapter 4 in combination with a pro-based theory of null
subjects can provide a straightforward account for the phenomenon in question. In
section 5.4.1, I lay out the theoretical tools for my analysis: a pro-based null subject
theory and the syntactic structure for PNCs established in Chapter 4. In section 5.4.2, I
show how the combination of the two tools can account for PNC pro-drop. In section
5.4.3, I argue why a pro-based theory should be adopted over what I call the ‘V°-raising
hypothesis’ of null subject phenomena. In section 5.4.4, I discuss cases in which PNC
pro-drop is not allowed even when the PNC pro-drop conditions are conjunctively
satisfied.
predicts that the type (ii) languages would not permit PNC pro-drop since they are not consistent pro-drop
languages.
217
5.4.1
Two Cornerstones for an Analysis of PNC Pro-Drop
5.4.1.1 Pro-Drop
There have been numerous attempts to account for pro-drop in terms of syntactic
relations (Perlmutter 1971; Chomsky 1981; Rizzi 1982, 1986, among many others;
Holmberg 2005; cf. Holmberg 2010; Roberts 2010).
One of the most influential theories of pro-drop was proposed in Rizzi (1986).
Rizzi’s theory consists of two essential components: a licensing condition and an
identification condition, as in (47) (as stated in Holmberg (2005:536)).
(47) Pro-drop parameter:
a.
Licensing
Pro is Case-marked by Xy°, where y is parameterized.
b.
Identification
Pro inherits the phi-feature values of Xy°(if it has phi-features; if not, pro
gets a default interpretation, typically arb).
Rizzi’s pro-drop parameter in (47) has been proposed in order to account for not only null
subjects but also arbitrary null objects. The choice of the pro-licensing Xy°is a property
specific to a given language. If we limit our attention to null subjects, which is our
primary concern here, T°(or INFL°) is a pro-licensing Xy°in Italian, allowing null
218
subject pronouns. By contrast, T°in English is not such a head and thus disallows null
subject pronouns. Under Rizzi’s theory, pro is unspecified for phi-features, and the prolicensing T°identifies the content of pro, which is possible due to the rich verbal
inflection, which manifests the phi-features on T°. Consider the trees in (48).
(48) a.
TP
pro
b.
T’
TP
pro
T’
[φ]
T°
…
[φ]
T°
…
[φ]
(48a) represents the status of pro with no specified phi-features; (48b) represents the
situation in which the content of pro has been identified by T°, in accordance with
whatever the values of the phi-features on T°are.
Holmberg (2005) points out that Rizzi’s (1986) pro-drop parameter does not fit in
with the current theory of agreement, in which there is a distinction between
uninterpretable/unvalued and interpretable/valued features (in the sense of Chomsky
1995, 2000, 2001a,b). Uninterpretable phi-features must be removed, since they are
uninterpretable and thus cannot be read off at LF. T°bears uninterpretable phi-features
whereas DP bears interpretable phi-features. If T°’s uninterpretable phi-features, after
being valued for phonological manifestation, must be eliminated before LF, there is no
way that T°can identify the content of pro. For this reason, Holmberg reformulates the
status of pro. It is argued that pro bears interpretable/valued phi-features, as do ordinary
pronouns, the only difference being that the former is simply phonologically null. T°,
219
bearing unvalued phi-features, enters an agreement relation with pro and has its unvalued
phi-features valued. Pro moves to [Spec, TP] in order to satisfy the EPP on T°when EPP
is present. He further argues that what distinguishes consistent pro-drop languages from
the other languages is the presence of a D-feature on T°, which enables the definite
interpretation of the null subject. He assumes pro to be a type of weak pronoun (in the
sense of Cardinaletti and Starke 1999) (φP), which thus lacks definiteness. The D-feature
on T°renders such a pro definite via Agree. The whole process is illustrated in (49).
(49)
TP
pro
T’
T°
vP
pro
v’
v°
...
T°scans its c-command domain and values its unvalued phi-features by the virtue of
agreement with pro, which bears the matching valued phi-features. At the same time,
pro’s unvalued D-feature is valued against T°’s valued D-feature. Pro moves up to [Spec,
TP] for EPP reasons. Holmberg compares this version of pro-drop theory with an
alternative approach in which the way in which the EPP on T°is satisfied is
parameterized (Alexiadou and Anagnostopolou 1998; inter alia), and argues for the probased approach for Finnish. I simply adopt the pro-based account in a broad sense over
the alternative approach, but I will return to justify this choice below.
220
5.4.1.2
Mediated Agreement
In Chapter 4, we developed a syntactic account of the structure of the PNC and its
derivation. The main idea is that the pronoun is base-generated in the specifier position of
an extended nominal projection (i.e., [Spec, dxP]) and moves to [Spec, DP] in order to
satisfy the TH-Criterion. It has been proposed that D°enters the derivation with unvalued
phi-features, and values them by entering into an Agree relation with one or multiple
matching goals—in this case, the pronoun, Num°and N°—which bear the matching phifeatures. It has been argued that DP-internal agreement feeds DP-external agreement
between T°and the pronoun. The derivation is illustrated in (50).
(50)
TP
DPPNC
T’
T°[EPP,φ]
vP
Agree2
DPPNC[φ]
Pronoun
D’
v’
v°
D°[+TH,φ] NumP
Agree1
Pronoun Noun
VP
…
221
5.4.2
Mediated Pro-Drop
The mediated agreement that involves both DP-internal and DP-external agreement,
coupled with the reformulated pro-drop analysis, leads to a straightforward account for
PNC pro-drop. In a nutshell, the process of mediated pro-drop is the mirror image of that
of mediated agreement. I propose that PNC pro-drop is licensed in a Spec-Head
configuration by the mediating head D, as the pronoun-verb agreement is mediated by D°,
as is illustrated in (51).
(51)
TP
DP
pro
License2
T’
License1
D’
D°
Agree1
T°
dxP
pro
Agree2
dx’
dx°
vP
DP
v’
v°
VP
NP
In (51), T°licenses pro in its specifier position. Since unlike the conventional pro-drop,
the nominal in [Spec, TP] is not a pronoun, but a lexical DP, the whole subject cannot be
omitted. Instead, the head of the lexical subject DP further scans its specifier position and
it eventually licenses pro.
222
Mediated pro-drop is made possible due to the TH-Criterion and mediated
agreement, as proposed in Chapter 4. The TH-Criterion ensures a Spec-Head
configuration established between the mediating licenser D°and its licensee pro.
Mediated agreement creates permanent links between probe and goal (i.e., between T°
and DP, and between D°and pro). The primary reason for assuming a null pronoun
subject pro is to satisfy the EPP on T°, which forces a Spec-Head configuration between
them. This is obviously the case for the DP-internal syntax in (51), in which the pronoun
occupies [Spec, DP] to satisfy the TH-Criterion on D°. Based on this, I propose that the
permanent link created by feature-sharing agreement between T°, D°and pro(noun) not
only enables T°to indirectly agree with the phi-features embedded in DP but also assures
extended pro-licensing via the mediating D°. Put differently, as the feature-sharing
approach to agreement indirectly connects two elements (T°and pro(noun)) via the
intervening head (D°) which serve both as a probe for pro(noun) and a goal for T°, the
pro-drop can be licensed indirectly guided by the permanent link: The intervening head D
licenses the pro in [Spec, DP]. In the case of DPs of non-PNC-type, no mediated prolicensing is necessary: the fact that the pro(noun) does not exist inside the non-PNC-type
DPs implies the absence of DP-internal agreement as well as Spec-Head configuration
between D°and any pro(noun). That is, there is no permanent agreement link created
between D°and a pronominal element inside a non-PNC-type DP. For the same reason,
mediated pro-licensing is not required DP-internally when the subject is itself a pro(noun),
since pro-drop in this case is licensed directly by T°.
223
To recap, what is of importance is the second condition of PNC Pro-Drop
Generalization in (46). That is, PNC pro-drop cannot be licensed without an overt
definite article. Crucially, T°, which eventually licenses PNC pro-drop, is not in a SpecHead configuration with its licensee, but D°is. The permanent link created between T°
and D°enables T°to be in an indirect Spec-Head configuration with its licensee through
the mediating licensor D°.
5.4.3
PNC Pro-Drop and Theories of Pro-Drop
The current analysis has an important theoretical implication for theories of pro-drop.
There are mainly two approaches to pro-drop. On the one hand, it has been argued that a
phonetically null syntactic element—pro—has to be posited so that the EPP can be
satisfied in null subject constructions (Holmberg 2005; see also Chomsky 1981; Huang
1984; Rizzi 1982, 1986; among many others). The current analysis of PNC pro-drop is
based on this approach. On the other hand, it has been argued that the EPP can be
satisfied by V°-to-T° raising, and thus we can eliminate the need for pro (Alexiadou and
Anagnostopoulou 1998; see also Jelinek 1984; Barbosa 1995; Manzini and Savoia 2002;
Platzack 2003, 2004; among many others). I will call the former the ‘pro-hypothesis’ and
the latter the ‘V°-raising hypothesis’. In what follows, it will be shown that only the prohypothesis, not the V-raising hypothesis, can be extended to account for PNC pro-drop.
According to the V°-raising hypothesis, the EPP is satisfied, for instance in Greek
and Spanish, by head movement—that is, V°-raising to T°, and thus there is no need to
224
postulate that T° projects its specifier, which is in turn needs to be filled by pro (or an
overt subject). Thus, the V°-raising hypothesis dispenses with pro. I have argued above
that PNC pro-drop is mediated by D°. If the V°-raising hypothesis were to be extended to
account for PNC pro-drop, the natural hypothesis would be that the counterpart to V°raising in the IP domain would be N°-raising within DP. This hypothetical N°-raising
would satisfy the TH-Criterion on D° (which requires [Spec, DP] to be filled by a
syntactic element) in the same way that V°-raising is taken to satisfy the EPP feature on
T° in the V°-raising hypothesis.
However, it appears that extending the V°-raising hypothesis to account for PNC
pro-drop in this way is impossible. First of all, if N°-raising is to satisfy the TH-Criterion
on D°, N° would have to raise as high as D°. N°-to-D° raising (through intermediate
functional heads) is not attested in Greek and Spanish, however. Hence, the TH-Criterion
on D° could not be satisfied, and the null pronoun within PNC could not be licensed.
Second, even if we assume that N°-raising to the functional head of some extended
nominal projection, lower than D°, can mediate PNC pro-drop, we immediately
encounter a critical problem. It has been argued that N°-to-Num° raising is necessary in
order to derive the post-nominal position of demonstratives, but such head movement is
blocked in the presence of an adjective in [Spec, NumP] (cf. Chapter 4 and Panagiotidis
2000). This means that when there is an adjective within PNCs, there is no N°-to-Num°
raising. Let us then suppose for the moment that N°-to-Num° raising can somehow
mediate PNC pro-drop. The prediction is then that PNC pro-drop should not be allowed
in the presence of an adjective. The reason is as follows: N°-to-Num° raising is
225
responsible for licensing PNC pro-drop, but the presence of an adjective indicates no N°to-Num° raising. For this reason, when there is an adjective present within PNCs, PNC
pro-drop should not be licensed. This prediction is not borne out, however. PNC pro-drop
can still be licensed even when an adjective occurs pre-nominally within PNCs (i.e., in
the absence of N°-to-Num° raising) in Greek, as shown in (54).
(52) a.
b.
(Emis)
we
i
the
nei
young
glossologi imaste
linguists be.1PL.PRES
exypni.
smart
= ‘We young linguists are smart.’
≠ ‘The young linguists are smart.’
(Esis)
i
nei
linguists
you
the young
linguists
= ‘You young linguists are smart.’
≠ ‘The young linguists are smart.’
isaste
be.2PL.PRES
Greek
exypni.
smart
In (52), the PNCs contain an adjective, which indicates the absence of N°-to-Num°
raising. 14 However, PNC pro-drop is allowed, contrary to the prediction made by
Note that Noun > Adjective word order is only allowed in determiner spreading constructions, as in (ia)
and (ib).
14
(i)
a.
b.
c.
d.
Emis
we
Esis
you
*Emis
we
*Emis
we
i
the
i
the
i
the
i
the
glossologi
linguists
linguists
linguists
glossologi
linguists
linguists
linguists
i
the
i
the
nei
young
nei
young
nei
imaste
exypni.
young be.1PL.PRES smart
nei
isaste
exypni.
young be.2PL.PRES smart
imaste
exypni.
be.1PL.PRES smart
isaste
exypni.
be.2PL.PRES smart
Greek
The idea that N°-raising derives the Noun > Adjective word order in (ia) and (ib) can be maintained if we
assume Giusti’s (2002) account of determiner spreading: the adjective is articulated as a result of
definiteness agreement and is base-generated in the specifier of a functional phrase, and its post-nominal
226
extending the V°-raising hypothesis to the nominal domain. As a matter of fact, the
absence/presence of an adjective within PNCs has nothing to do with PNC pro-drop.
A similar line of argument can be implemented for another PNC pro-drop language,
namely Spanish. Let us assume for our purposes that the typical post-nominal adjective
position in this language is derived by N°-raising (Cinque 1994).15 On this assumption,
the presence of a post-nominal adjective suggests that N°has raised to a higher functional
head. However, it turns out that the relative word order between a noun and an adjective
does not have any effect to the availability of PNC pro-drop; PNC pro-drop is licensed
with the (relatively restricted) set of pre-nominal adjectives in Spanish as well as with the
post-nominal ones.
(53) a.
b.
(Nosotros) los
we
the
lingüistas jóvenes
linguists young
listos.
smart
= ‘We young linguists are smart.’
≠ ‘The young linguists are smart.’
(Vosotros) los lingüistas jóvenes
you
the linguists young
= ‘You young linguists are smart.’
≠ ‘The young linguists are smart.’
somos
be.1PL.PRES
sois
be.1PL.PRES
Spanish
listos.
smart
position is derived by raising N° to a higher functional phrase.
15
N°-raising as a syntactic operation in Romance and Germanic languages is questioned in the literature
(Larmarche 1991; Bouchard 1998, 2002; Shlonsky 2004; Cinque 2005, 2010; among others). If N°-raising
does not exist in these languages, there would be no way to satisfy the TH-Criterion on the D° of PNCs in
any event. If we follow this literature, then, we are also led to choose the pro-based account over the V°raising hypothesis as our tool to account for PNC pro-drop.
227
(54) a.
b.
(Nosotros) los jóvenes
lingüistas
we
the young
linguists
= ‘We young linguists are smart.’
≠ ‘The young linguists are smart.’
(Vosotros) los jóvenes
lingüistas
you
the young
linguists
= ‘You young linguists are smart.’
≠ ‘The young linguists are smart.’
somos
be.1PL.PRES
listos.
smart
sois
be.1PL.PRES
listos.
smart
Spanish
As illustrated in (53) and (54), PNC pro-drop is allowed regardless of the DP-internal
word order. That is, whether the adjective precedes or follows the noun, PNC pro-drop is
licensed. This again suggests that the V°-raising hypothesis cannot be extended to
account for PNC pro-drop.
In this section, I have shown that the V°-raising hypothesis for pro-drop in the
clausal domain cannot be extended to account for pro-drop in the nominal domain. Given
this, I conclude that the traditional pro-based account is more adequate for PNC pro-drop
since it can be extended to account for its main properties.
5.4.4
When PNC Pro-Drop Is Blocked
We have established two conditions to which PNC pro-drop is subject, and provided a
syntactic account of PNC pro-drop. However, there are two cases in which PNC pro-drop
is banned even in Greek—our representative language that satisfies both the conditions.
The two cases are as follows: (i) when the pronoun is modified by a reinforcer; and (ii)
when the DP under consideration is not a PNC but a DNC. This section will address these
issues. The former case is straightforwardly explained if we assume the proposed analysis
228
of PNC pro-drop. For the latter case, I will attribute the behavior of the two cases to
semantics/pragmatics interface issues.
5.4.4.1 Reinforcers Block PNC Pro-Drop
As we have seen in Chapter 3, pronouns in PNCs can be modified by a reinforcer, and in
this case, such constructions receive a contrastive interpretation with a focal stress. In
these cases, the pronoun cannot be omitted, as shown above in (7) and (8) (reproduced in
(55) and (56)).
(55) a.
b.
(56) a.
b.
[Emis
we
edho i
here the
glossologi]SUBJECT
linguists
exypni.
smart
‘We here linguists are smart.’
[Esis
eki i
glossologi]SUBJECT
you.PL
there the linguists
‘You there linguists are smart.’
imaste
be.1PL.PRES
isaste
be.2PL.PRES
*[Edho
here
i
the
glossologi]SUBJECT
linguists
imaste
be.1PL.PRES
exypni.
smart
*[Eki
there
i
the
glossologi]SUBJECT
linguists
isaste
be.2PL.PRES
Greek
exypni.
smart
Greek
exypni.
smart
229
The cases in (55) and (56) can be taken to be the same as the cases in which emphatic
subject pronouns in consistent pro-drop languages are spelled-out overtly rather than
unexpressed, as discussed above in section 5.2. Each example is reproduced in (57) (for
the clausal domain) and (58) (for the nominal domain).
(57) a.
b.
c.
(58) a.
b.
[EMIS]SUBJECT imaste
we
be.1PL.PRES
‘WE are smart.’
[ESIS]SUBJECT
isaste
you.PL
be.2PL.PRES
‘YOU are smart.’
[AFTI]SUBJECT einai
they
be.3PL.PRES
‘THEY are smart.’
[EMIS
we
i
the
exypni.
smart
Greek
exypni.
smart
exypni.
smart
glossologi]SUBJECT
linguists
exypni.
smart
‘WE linguists are smart.’
[ESIS
i
glossologi]SUBJECT
you.PL
the linguists
‘YOU linguists are smart.’
imaste
be.1PL.PRES
isaste
be.2PL.PRES
Greek
exypni.
smart
The reason why that the pronouns in (57) and (58) are overtly expressed is presumably
that the “emphatic” information associated with the pronoun cannot be delivered by other
means. In the case of ordinary pro-drop, the information carried by an unstressed pronoun
can be recovered by the rich verbal agreement morphology. However, the inflectional
morphology does not reflect the discourse-sensitive information delivered by emphasis,
230
and no other constituent retains this information if the pronoun is omitted. That is to say,
if we delete the emphatic pronoun, only information related to the phi-features of the
pronoun can be recovered, but not the emphatic information.
The above explanation which appeals to the recoverability of the emphatic
information is not extendable to the cases in (55) and (56), however. Recall that the
primary function of reinforcers is to add an emphatic feel to the interpretation of a
pronoun. This means that the presence of a reinforcer can allow one to retrieve the
emphatic content. If so, it is predicted that PNC pro-drop should be allowed in (55) and
(56), contrary to fact, since the phi-feature information and the emphatic content can be
recovered from the verbal inflection and the reinforcer, respectively.
I argue instead that the structural configuration required for PNC pro-drop to be
licensed is not satisfied in (55). Specifically, the pronoun-reinforcer construction is
phrasal as an instance of modifiee-modifier relationship (see Chapter 4), and the string of
a pronoun and a reinforcer occupies as a constituent [Spec, DP] of the PNC, as
schematized below (abstracting away from the internal structure of the pronounreinforcer construction).
(59)
PNC
DP
pronoun- D°
reinforcer
D’
NumP
231
If this is the structure of the PNC containing a pronoun-reinforcer construction, then the
current analysis of PNC pro-drop predicts the pronoun cannot be dropped when modified
by a reinforcer. The reason is that what is in a Spec-Head configuration with the
mediating licenser D°is the pronoun-reinforcer construction rather than the pronoun. The
proposed mediated PNC pro-drop is strictly regulated by the Spec-Head configuration
established between the licensee and the medicating licenser. Hence, the target of PNC
pro-drop is predicted to the entire DP in the specifier of the PNC DP, rather than the
pronoun itself in (55). Therefore, omitting the pronoun in the context of pronounreinforcer constructions cannot be licensed, as shown in (56). Note that omitting the
pronoun-reinforcer construction embedded in the PNC is disallowed due to the above
mentioned recoverability issue.
5.4.4.2 Demonstratives Are Never Dropped
Let us turn to the fact that demonstratives are never omitted in the context of DNCs. The
main idea of this thesis is that PNCs must be treated on par with DNCs (as extensively
discussed in Chapter 2). Given this hypothesis, the analysis of DNCs according to which
the demonstrative is base-generated in the same position as the pronoun of PNCs (i.e.,
[Spec, dxP]) (cf. Chapter 3), and the discussion of PNC pro-drop in this chapter, predict
that dropping should be allowed. However, this prediction is incorrect, as shown in the
examples in (60) and (61).
232
(60) a.
b.
Afti
i
glossologi
these
the linguists
‘These linguists are smart.’
Ekini
i
glossologi
those
the linguists
‘Those linguists are smart.’
ine
be.3PL.PRES
exypni.
smart
ine
be.3PL.PRES
exypni.
smart
(61) I
glossologi ine
exypni.
the linguists be.3PL.PRES
smart
‘The linguists are smart.’ or ‘Linguists are smart.’
≠ ‘These/Those linguists are smart.’
Greek
Greek
If we take out the demonstratives from the sentences in (60), those sentences can never
retain the meaning the demonstratives contributed, whether the interpretation deictic
contrastive or generic. Instead, the subject DP in (61) can only be interpreted to be either
as generic or definite, as indicated by the English translation. The situation here sharply
contrasts with the situation of PNC pro-drop (the examples are repeated in (62) for
convenience).
(62) a.
b.
[I
glossologi]SUBJECT
the linguists
‘We linguists are smart.’
≠ ‘The linguists are smart.’
[I
glossologi]SUBJECT
the linguists
‘You linguists are smart.’
≠ ‘The linguists are smart.’
imaste
be.1PL.PRES
exypni.
smart
isaste
be.2PL.PRES
exypni.
smart
Greek
The subjects of the sentences in (62) unambiguously receive a PNC interpretation with or
without a pronoun.
233
The question to be asked at this point is: whence stems this difference between
PNCs and DNCs with respect to the (un)availability of omitting the pronoun and
demonstrative? The answer to this question lies in considering whether or not the omitted
content is recoverable. As has been shown in section 5.3, PNC pro-drop is only allowed
in consistent pro-drop languages whose rich verbal inflection permits recovery of the
information which is missing due to the absence of the pronoun. That is, the content of
the missing pronoun in examples like (62) can be easily recovered by virtue of the
information encoded on the verb. This recoverability of the missing information therefore
makes it possible to omit the pronoun in the context of PNCs.
By contrast, if the demonstrative of DNCs is dropped, the resulting sentence is
identical to a sentence whose subject is a regular definite DP, as can be seen above by
comparing (60) and (61). There is no clue from the form of the verb as to what is missing
in (61). In (62), the fact that the sentence is grammatical in spite of the person feature
mismatch between the subject and the verb reveals that the agreement controller is null.
In (61), there is no such mismatch since both the missing element (i.e., demonstrative)
and the string-identical non-DNC subject bears the same person feature. For this reason,
dropping demonstratives involves removing the only source of the missing information
on which the discourse participants can count—that is, the demonstrative itself. As we
have seen in Chapter 2, the presence of the demonstrative has the effect of excluding the
discourse participant(s) from the set denoted by the DNC. This semantic contribution is
not recoverable once the demonstrative is omitted due to the grammatical expressions
given in (61). For this reason, if the speaker utters the sentence in (61) with the intention
234
to express the meaning denoted by the sentences in (60), the addressee is not able to infer
the missing information. This is the reason why DNC pro-drop is impossible even in
languages where DNCs satisfy the both conditions of PNC pro-drop.
5.5
Summary
In this chapter, I addressed the issue of the optionality of the pronoun contained in PNCs.
I have identified two licensing conditions for PNC pro-drop by investigating the syntactic
environments in Greek and other languages such as English, Korean, Hebrew, Spanish,
etc. I have shown that PNC pro-drop can receive an adequate account if we adopt the
syntax of PNCs proposed in Chapter 4 and the pro-based approach to null subject
phenomenon. I have also shown that the V°-raising approach to null subject phenomenon,
which is the major competitor of the pro-based approach, cannot be extended to account
for PNC pro-drop. Finally, I have addressed cases in which PNC pro-drop is disallowed,
involving focused pronouns combined with a reinforcer, and addressed the question of
why demonstratives are never dropped. I have argued that pro-drop fails due to syntactic
reasons for the former case, and pragmatic reasons concerning recoverability of deleted
information for the latter case.
235
CHAPTER 6
CONCLUDING REMARKS
6.1
Summary
The primary goals of this dissertation have been a) to prove that the PNC must be treated
on par with the DNC, b) to argue on that basis for a particular analysis of PNC structure,
and c) to explore the empirical and theoretical consequences of the proposed analysis.
In so doing, this present study addresses the contrasting facts concerning the PNC
in English and Greek. On the face of it, the pronoun in the PNC appears to pattern with
the definite article in English while it is clearly distinct from the definite article in Greek.
As an attempt at providing a unified analysis of the PNC in these languages, this study
argues in favor of an analysis which treats the PNC as involving the same structure as the
DNC. As I have demonstrated in Chapter 2, the PNC and the DNC have syntactic and
semantic similarities in common. First, a demonstrative cannot be added to the the PNC;
nor can a pronoun be added to the DNC. This suggests that the two elements are in
complementary distribution. In Greek, in both cases, a definite article is required for the
construction to be well-formed. Second, both demonstratives and pronouns can be used in
combination with a reinforcer, and show parallel requirements in that combination. Third,
both constructions can receive either a deictic contrastive interpretation or a generic
interpretation. Lastly, the presence of a demonstrative or a pronoun imposes the same
type of membership restriction effect on the denotation of the DP: the only
236
interpretational difference between the demonstrative and the pronoun is in the value of
the person feature: third person for demonstratives and first or second person for
pronouns.
A formal analysis of the above facts began with a discussion of the ample literature
on the DNC. In the course of building up an analysis of the DNC, I have singled out
untenable sub-claims from the previous literature. The resulting proposal for the DNC on
the basis of the selected assumptions is that the demonstrative is base-generated in a low
position within the DP and its surface position is determined by movement of the
demonstrative and the noun head, largely following the analyses suggested by Giusti
(1997, 2002), Panagiotidis (2000) and Rosen (2003). This analysis is then carried over to
the PNC: the pronoun is introduced into the derivation of the PNC and the word order is
derived by the same set of movements. In spite of the fact that the pronoun always occurs
in the left-most position in Greek and English, I show that the pronoun can appear low in
the structure in some languages such as Korean, corroborating the proposed low firstmerge position for the pronoun.
Subsequently, a detailed discussion of the consequences that derive from the
proposed analysis of the PNC and DNC follows. First, the proposal leads to an inevitable
change in our view regarding where in a given DP the valued person feature comes from.
The main claim is that the valued person-feature of a given DP originates from the
pronoun/demonstrative embedded by the DP. This is in contrast to the predominant
hypothesis that the D head of a given DP provides the value for person (Panagiotidis
2002; Longobardi 2008; among others). Although the D head of the DP does end up
237
bearing a person feature in the proposed analysis, it does so via entering into an Agree
relation with the low pronoun or demonstrative, not by virtue of bringing that feather
from the numeration.
Second, the proposed analysis of the PNC/DNC necessitates a revised view of
agreement between a PNC/DNC subject and an inflected verb. I have proposed a
mediated agreement operation, in which DP-internal agreement feeds DP-external
agreement. Mediated agreement involves two agreement processes: The D head collects
all the value of the phi-features via DP-internal agreement, and subsequently the T head
agrees with these features on the D head. The proposed mediated agreement is
implemented by assuming Pesetsky and Torrego’s (2007) feature sharing approach to
agreement, which allows a single syntactic element to function both as a probe and as a
goal for distinct agreement operations. This fares well with mediated agreement in which
the D head serves as a probe for DP-internal agreement, but as a goal for subsequent
agreement probed by the T head.
Third, the proposed parallel treatment of the PNC and the DNC corroborates the
DP/CP parallelism hypothesis proposed by an increasing number of authors (e.g.,
Szabolcsi 1983, 1987, 1994). As noted in Chapter 2, the fact that in Greek the pronoun
always surfaces in its dislocated position while the demonstrative can surface either in a
dislocated position or in its base position stands out as an exception to the parallel
treatment of the PNC and the DNC. This apparent exception is explained if the difference
is reconsidered in connection with the movement pattern of wh-phrases in the language.
Recent research reveals that Greek allows two types of information seeking questions: wh
238
ex-situ and wh in-situ. The potential answer to wh in-situ questions must target a
linguistic antecedent in the previously established discourse (Vlachos 2012). In this sense,
in-situ wh-phrases are anaphoric. With this conclusion in place, the fact that on its
anaphoric use, the demonstrative can appear in its base-position is not a surprise. Instead,
this fact can be directly connected to the fact that in-situ wh-phrases are anaphorically
interpreted. Further, because the pronoun in the PNC lacks an anaphoric interpretation, it
cannot appear in the base-position. This is the source of the difference in positional
possibilities between the PNC and the DNC.
Fourth, the membership restriction effect that arises due to the presence of a
pronoun/demonstrative is captured by treating it as involving presuppositional semantics.
That is, the pronoun or demonstrative adds a presupposition to the meaning of the PNC or
DNC. First and second person pronouns specify that the speaker or the addressee is a
member of the set picked out by the PNC, while demonstratives, being third person,
specify that neither the speaker nor the addressee is a member of the set picked out by the
DNC.
Lastly, the peculiar facts concerning the optionality of the pronoun embedded in the
PNC is discussed. I have argued that the fact that the pronoun in question can be dropped
only in some languages (Greek and Spanish) but not in others (Chinese, English, German,
Italian, Japanese, Korean, Hebrew, Russian, and Turkish) can be accounted for within the
proposed analysis of the PNC in conjunction with extant theories of pro-drop which
assume pro. Investigating the conditions in which the pronoun in question can be
suppressed in the languages just mentioned, I draw the generalization that PNC pro-drop
239
is licensed only in consistent pro-drop languages whose PNC obligatorily requires a
definite article. I further argue that the proposed analysis of the PNC in combination with
a pro-based theory of null subject phenomenon (e.g., Rizzi 1986) provides a
straightforward account for the facts. Specifically, I propose that this kind of pro-drop is
mediated by the D head, by virtue of its role in mediating agreement. I discuss this
particular pro-drop phenomenon as a testing ground for the rival theories of pro-drop
(e.g., Alexiadou and Anagnostopoulou 1998), according to which pro can be licensed by
head-movement in some languages. I conclude that the latter type of pro-drop theories
cannot be adequately extended to account for pro-drop within the PNC.
6.2
Questions for Future Research
While the present study provides answers for some questions related to PNCs, there are
some remaining questions which have not been touched on.
First, recall from Chapter 2 that we have identified two differences between PNCs
and DNCs. I have shown that one of them—the fact that the demonstrative can appear
either high or low in the structure while the pronoun can only appear high in Greek—is in
fact not a difference at all (section 4.4). The second difference between the two
constructions has to do with the availability of the generic interpretation. Its availability
in DNCs is determined by an interaction between the humanness and number features.
That is, DNCs whose referent is [SG] and [+HUMAN] cannot receive a generic
240
interpretation, while those whose referent is [SG] can typically be generically interpreted
as long as the other feature is [-HUMAN]. This pattern remains to be accounted for.
Second, why is it that only some pronouns can participate in the formation of PNCs
in the languages that permit them, while others cannot? The restrictions imposed on
which pronouns can be a part of PNCs in the languages investigated in this dissertation
varies. For example, third person pronouns in Greek and Turkish, singular personal
pronouns and the third person plural pronoun in English, German, Hebrew, Italian,
Korean, Russian and Spanish, and all the singular pronouns in Chinese and Japanese are
banned from forming PNCs.
Some attempts have been made to account for this kind of restriction in the
literature. A pragmatic account of the singular/plural asymmetry in English has been
proposed by Pesetsky (1978); the reason that singular pronouns cannot participate in the
formation of PNCs is attributed to “the obligatory unique reference of singular pronouns
as opposed to the unspecified reference of plural pronouns” (Pesetsky 1978:352). This
idea is adopted by Noguchi (1997:776) as an account for the Japanese case. in contrast,
Bernstein (2008) proposes a morpho-syntactic account for the English case. She proposes
that PNCs can be formed with pronouns that satisfy two conditions: they “must display
overt person marking … and … must agree in number with the noun” (Bernstein
2008:225). However, none of these accounts is extendable to account for the restrictions
imposed on PNCs in other languages (e.g., Greek), as also noted by Lyons (1999). At the
moment, a universal account for the pattern does not seem to be available.
241
Finally, it is interesting to note that demonstratives as well as DNCs are universal
across languages, while PNCs are not available in all languages. For instance, unlike the
languages inspected in this thesis, Modern Irish (Irish) does not accept PNCs. What is the
source of this unexpected difference, given the similarities we have emphasized between
DNCs and PNCs? One might speculate that the absence of PNC in a given language ties
in with the absence of strong pronouns in that language (in the sense of Cardinaletti and
Starke 1999). However, Irish does have strong pronouns; nevertheless, PNCs are still illformed in this language.
These remaining questions require a great deal of extensive research both crossand intra-linguistically; I will not pursue answers to these questions in this venue. Instead,
I will leave these issues for future research.
242
REFERENCES
Abels, Klaus, and Ad Neeleman. 2009. Universal 20 without the LCA. In: Merging
features: Computation, interpretation, and acquisition. ed. J. M. Brucart, A.
Gavarro, and J. Sola, 60–79. New York: Oxford University Press.
Abels, Klaus, and Ad Neeleman. 2012. Linear Asymmetries and the LCA. Syntax 15:25–
74.
Abney, Stephen. 1987. The English noun phrase in its sentential aspect. Doctoral
dissertation. MIT, Cambridge, MA.
Aboh, Enoch O. 2004. Topic and focus within D. Linguistics in the Netherlands 21:1–12.
Agouraki, Y. 1990. On the projection of maximal categories: the case of CP and FP in
Modern Greek. UCL Working Papers in Linguistics 2:183–200.
Alexiadou, Artemis, and Elena Anagnostopoulou. 1998. Parametrizing AGR: Word order,
V-movement and EPP-checking. Natural Language and Linguistic Theory
16:491–539.
Alexiadou, Artemis, Liliane Haegeman, and Melita Stavrou. 2007. Noun phrase in the
generative perspective. Berlin: Walter De Gruyter.
Alexopoulou, Theodora, and Raffaela Folli. 2010. Indefinite topics and the syntax of
nominals in Italian and Greek. In Proceedings of The Twenty-Eighth West Coast
Conference on Formal Linguistics, M. B. Washburn, S. Ouwayda, C. Ouyang, B.
Yin, C. Ipek, L. Marston, and A. Walker. Department of Linguistics, Southern
California University, Los Angeles, CA.
Bale, Alan. 2009. Yet more evidence for the emptiness of plurality. In Proceedings of the
Thirty-Eighth Annual Meeting of the NELS, Vol. 1, ed. A. Schardl, M. Walkow
and M. Abdurrahman, 75–88. Amherst: GLSA.
Barbosa, Pilar. 1995. Null subjects. Doctoral dissertation. MIT, Cambridge, MA
Benveniste, Émile. 1971. The nature of pronouns. Problems in general linguistics, 217–
222. FL: University of Miami Press.
Bernstein, Judy. 1997. Demonstratives and reinforcers in Romance and Germanic
languages. Lingua 102:87–113.
Bernstein, Judy. 2001. The DP Hypothesis: Identifying Clausal Properties in the Nominal
Domain. In The handbook of contemporary syntactic theory. ed. M. Baltin and C.
243
Collins, 536–561. Oxford: Blackwell.
Bernstein, J. 2008. English th- Forms. In Essays on nominal determination: From
morphology to discourse management, ed. H. Müller and A. Klinge, 213–232.
Amsterdam: John Benjamins.
Biberauer, Theresa, Anders Holmberg, and Ian Roberts. 2007. Disharmonic word-order
systems and the Final-over-Final-Constraint (FOFC). In Proceedings of XXXIII
Incontro di Grammatica Generativa, 86–105.
Biberauer, Theresa, Anders Holmberg, and I. Roberts. 2008. Structure and linearization in
disharmonic word orders. In Proceedings of Twenty-Sixth Western Coast
Conference on Formal Linguistics, 96–104. Somerville, MA: Cascadilla
Proceedings Project.
Biberauer, Theresa, Anders Holmberg, and Ian Roberts. 2010. A syntactic universal and
its consequences. Ms., Universities of Cambridge and Newcastle.
Biberauer, Theresa, Michelle Sheehan, and G. Newton. 2010. Impossible changes and
impossible borrowings: The Final-over-Final Constraint. In Continuity and
Change in Grammar, ed. A. Breitbarth, C. Lucas, S. Watts, and D. Willis, 35–60.
Amsterdam: John Benjamins.
Bobaljik, Jonathan. 2002. A-chains at the PF interface: Copies and ‘covert’ movement.
Natural Language and Linguistic Theory 20:197–267.
Bouchard, Denis. 1998. The distribution and interpretation of adjectives in French: A
consequence of bare phrase structure. Probus 10:139–183.
Bouchard, Denis. 2002. Adjectives, number and interfaces: Why languages vary.
Linguistic Variation 61. Amsterdam/Boston: North Holland Linguistic Series.
Bowdle, Brian, and Gregory Ward. 1995. Generic Demonstratives. In Proceedings of the
Twenty-First Annual Meeting of the Berkeley Linguistics Society, 32–43.
Linguistics Department, University of California, Berkeley, CA.
Bowers, John. 1993. The Syntax of Predication. Linguistic Inquiry 24:591–656.
Brugè, Laura. 1996. Demonstrative movement in Spanish: a comparative approach.
University of Venice Working Papers in Linguistics 6:1–53.
Brugè, Laura. 2002. The position of demonstratives in the extended nominal projection in
functional structure. In DP and IP: The cartography of syntactic structures, Vol.
1, ed. G. Cinque, 15–53. Oxford/New York: Oxford University Press.
Brugè, Laura, and Giuliana Giusti. 1996. On demonstratives. Talk presented at the 19th
244
GLOW Colloquium 1996, Athens, GLOW Newsletter 36, pp. 24–25.
Campbell, R., 1996. Specificity operators in SpecDP. Studia Linguistica 50: 161–188.
Campbell, R., 1998. A null pronominal in the noun phrase. Linguistic Inquiry 29: 153–
160.
Campos, Héctor, and Melita Stavrou. 2004. Polydefinite constructions in Modern Greek
and Aromanian. In Balkan syntax and semantics, ed. O. M. Tomić. 136–173.
Amsterdam: Benjamins.
Cardinaletti, Anna. 1990. Impersonal constructions and sentential arguments in German.
Padua: Unipress.
Carstens, Vicki. 2000. Concord in Minimalist Theory. Linguistic Inquiry 31:319–355.
Cardinaletti, Anna, and Michal Starke. 1999. The typology of structural deficiency: A
case study of the three classes of pronouns. In Clitics in the languages of Europe,
ed. H. van Riemsdijk, 145–235. Berlin: Mouton de Gruyter.
Carnie, Andrew. 2007. Syntax: A generative introduction (2nd edition). Oxford:
Blackwell Publishing.
Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language
Semantics 6:339–405.
Choi, Jaehoon. 2013. Demonstrative-reinforcer constructions. The Linguistic Association
of Korea Journal 21:1–27.
Chomsky, Noam. 1981. Lectures on government and binding: The pisa lectures.
Dordrecht: Foris Publications.
Chomsky, Noam. 1993. A minimalist program for linguistic theory. In The view from
Building 20: Essays in linguistics in honor of Sylvain Bromberger, ed. K. Hale
and S. J. Keyser, 1–52. Cambridge, MA: MIT Press. [Also published in
Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press.]
Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press.
Chomsky, Noam. 2000. Minimalist inquiries: the framework. In Step by step: Essays on
minimalist syntax in honor of Howard Lasnik, H. Lasnik, R. Martin, D. Michaels
and J. Uriagereka, 89–156. Cambridge, MA: MIT Press.
Chomsky, Noam. 2001a. Derivation by phase. In Ken Hale: A life in language, ed. M.
Kenstowicz, 1–52. Cambridge, MA: MIT Press.
245
Chomsky, Noam. 2001b. Beyond explanatory adequacy. In MIT Occasional Papers in
Linguistics, Vol. 20, 1–28. Cambridge, MA: MIT Department of Linguistics.
Chung, Sandra. 1998. The design of agreement: Evidence from Chamorro. Chicago:
University of Chicago Press.
Cinque, Guglielmo. 1994. On the evidence for partial N-movement in the Romance DP.
In Paths towards universal grammar, ed. G. Cinque, J. Koster, Y. Pollock, L.
Rizzi, and R. Zanuttini, 85–110. Washington, DC: Georgetown University Press.
Cinque, Guglielmo. 2005. Deriving Greenberg’s universal 20 and its exceptions.
Linguistic Inquiry 36:315–332.
Cinque, Guglielmo. 2010. The syntax of adjectives. A comparative study. Cambridge,
MA: MIT Press.
Cooper, Robin. 1983. Quantification and syntactic theory. Dordrecht: Reidel.
Déchaine, Rose-Marie, and Martina Wiltschko. 2002. Decomposing pronouns. Linguistic
Inquiry 33:409–442.
Delorme, Evelyne, and Ray Dougherty. 1972. An appositive NP construction.
Foundations of Language 8:2–29.
Diessel, Holger. 1999. Demonstratives. Form, function, and grammaticalization.
Amsterdam: John Benjamins.
Den Dikken, Marcel. 1998. Predicate Inversion in DP. In Possessors, predicates and
movement in the determiner phrase, ed. A. Alexiadou and C. Wilder, 177–214.
Amsterdam: John Benjamins.
Den Dikken, Marcel. 2006. Relators and linkers: the syntax of predication, predicate
inversion, and copulas. Cambridge, MA: MIT Press.
Dryer, Matthew. 2009. The branching direction theory revisited. In Universals of
language today, ed. S. Scalise, E. Magni and A. Bisetto, 185–207. Berlin:
Springer.
Elbourne, Paul. 2005. Situations and individuals. Cambridge, MA: MIT Press.
Furuya, Kaori. 2009. The DP hypothesis through the lens of Japanese nominal
collocation constructions. Doctoral dissertation. CUNY, New York, New York.
Giannakidou, A. and Stavrou, M. 1999. Nominalization and ellipsis in the MG DP.
Linguistic Review 16:295–331.
246
Giusti, Giuliana. 1996. Is there a FocusP and a TopicP in the Noun Phrase structure?
University of Venice Working Papers in Linguistics 6:105–128.
Giusti, Giuliana. 1997. The categorial status of determiners. In The new comparative
syntax. ed. L. Haegeman, 95–123. London and New York: Longman.
Giusti, Giuliana. 2002. The functional structure of noun phrases: A bare phrase structure
approach. In DP and IP: The cartography of syntactic structures, Vol. 1, ed. G.
Cinque, 54–90. Oxford/New York: Oxford University Press.
Giusti, Giuliana. 2005. At the left periphery of the Rumanian noun phrase. In
Proceedings of the Tense and Aspect Conference, 23–49.
Giusti, Giuliana. 2006. Parallels in clausal and nominal periphery. In Phases of
interpretation, ed. M. Frascarelli, 163–186. Berlin: Mouton de Gruyter.
Gorrie, Colin F. 2014. Ideal-Typology. Doctoral dissertation. The University of Arizona.
Greenberg, Joseph. 1963. Some universals of grammar with particular reference to the
order of meaningful elements. In Universals of language, ed. by J. Greenberg,
73–113. Cambridge, MA: MIT Press.
Guardiano, Cristina. 2012. Demonstratives, word order and the DP between syntax and
semantics: Crosslinguistic remarks. Studies in Greek Linguistics 32:100–115.
Gutman, Eynat. 2004. Third person null subjects in Hebrew, Finnish and Rumanian: an
accessibility theoretic account. Journal of Linguistics 40: 463–90.
Grohmann, K. Kleanthes, and Panagiotidis, Phoevos. 2004. Demonstrative doubling in
Modern Greek. In University of Maryland Working Papers in Linguistics 13, ed.
P. Chandra, T. Fujii, U. Soltan, M. Yoshida, 109–131. University of Maryland,
College Park, MD.
Grohmann, K. Kleanthes, and Panagiotidis, Phoevos. 2005. An anti-locality approach to
MG demonstratives. In Contributions to the 30th Incontro di Grammatica
Generativa. ed. L. Brugè, G. Giusti, N. Munaro, W. Schweikert, G. Turano, 243–
263. Cafoscarina, Venice.
Haegeman, Liliane. 1991. Introduction to government and binding theory. Oxford: Basil
Blackwell.
Haegeman, Liliane. and Barbara Ürögdi. 2010. Referential CPs and DPs: An operator
movement account. Theoretical Linguistics 36: 111–152.
Halle, Moris, and Alec Marantz. 1993. Distributed Morphology and the pieces of
inflection. In The view from Building 20: Essays in linguistics in honor of Sylvain
247
Bromberger, ed. K. Hale and S. J. Keyser, 111–176. Cambridge, MA: MIT Press
Hawkins, John. 1983. Word order universals. New York: Academic Press.
Hawkins, John. 1994. A performance theory of order and constituency. Cambridge:
Cambridge University Press.
Harley, Heidi. and Ritter, Elizabeth. 2002. Person and number in pronouns: a featuregeometric analysis. Language 78: 482–526.
Heim, Irene, and Angelika Kratzer. 1998. Semantics in generative grammar. Oxford:
Blackwell.
Heycock, Caroline. 1994. The internal structure of small clauses. Proceedings of NELS
25:223–238. Amherst: GLSA.
Higginbotham, James. 1987. Indefiniteness and predication. In The representation of
(in)definiteness, ed. E. Reuland, and A. Meulen, 43–70. Cambridge MA: MIT
Press.
Higgins, Francis Roger. 1973. The pseudo-cleft construction in English. Doctoral
dissertation, MIT, Cambridge, MA.
Hiraiwa, Ken. 2001. Multiple Agree and the defective intervention constraint in Japanese.
MIT Working Papers in Linguistics 40:67–80
Holmberg, Anders. 2000. Deriving OV Order in Finnish. In The derivation of VO and
OV, ed. P. Svenonius, 123–152. New York: Oxford University Press.
Holmberg, Anders. 2005. Is there a little pro? Evidence from Finnish. Linguistic Inquiry
36: 533–64.
Holmberg, Anders. 2010. Null subject parameters. In Parametric variation: Null subjects
in minimalist theory, ed. T. Biberauer, A, Holmberg, I. Roberts, and M. Sheehan,
88–125. Cambridge: Cambridge University Press.
Horrocks, Geoffrey, and Melita Stavrou. 1987. Bounding theory and Greek syntax:
Evidence from wh-movement in NP. Journal of Linguistics 23:79–108.
Huang, James. 1984. On the distribution and reference of empty pronouns. Linguistic
Inquiry 15:531–574.
Jackendoff, Ray. 1977. X-bar syntax. Cambridge, MA: MIT Press.
Jespersen, Otto. 1909-49. A modern English grammar on historical principles, I-VII.
Londong/Copenhagen: Allen and Unwin.
248
Jelinek, Eloise. 1984. Empty categories, case, and configurationality. Natural Language
and Linguistic Theory 2:39–76.
Jelinek, Eloise. 2006. The pronominal argument parameter. In Arguments and Agreement,
ed. P. Ackema, P. Brandt, M. Schoorlemmer, and F. Weerman, 261–288. Oxford:
Oxford University Press.
Julien, Marit. 2007. On the relation between morphology and syntax. In The Oxford
Handbook of Linguistic Interfaces, ed. G. Ramchand and C. Reiss, 209–238.
Oxford: Oxford University Press.
Kang, Soon-Haeng. 2005. On the adjectives in Korean. University of Venice Working
Papers in Linguistics 15:153–169.
Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press.
Kayne, Richard. 2000a. A note on prepositions, complementizers, and word order
universals. In Parameters and Universals, 314–326. New York: Oxford
University Press.
Kayne, Richard. 2000b. Prepositional complementizers as attractors. In Parameters and
Universals, 282–328. New York: Oxford University Press.
Kayne, Richard. 2002. On some prepositions that look DP-internal: English of and
French de. Catalan Journal of Linguistics 1:71–115.
Kayne, Richard. 2004. Prepositions as probes. In Structures and beyond. The cartography
of syntactic structures, Vol. 3, ed. A. Belletti, 192–212. New York: Oxford
University Press.
Koopman, Hilda. 2003. The locality of agreement and the structure of the DP in Maasai.
In The role of agreement in natural language (Proceedings of TLS 5), ed. W.
Griffin, 207–227. Department of Linguistics, University of Texas, Austin, TX.
Available at http://uts.cc.utexas.edu/~tls/2001tls/Koopman.pdf.
Koopman, Hilda. 2006. Agreement configurations: In defense of ‘‘spec head’’. In
Agreement systems, ed. C. Boeckx, 159–199. Amsterdam: John Benjamins.
Kornfilt, Jaklin. 2003. Turkish. In International encyclopedia of linguistics, Vol. IV, 2nd
edition. ed. W. Frawley. Oxford: Oxford University Press.
Kratzer, Angelika. 2009. Making a pronoun: Fake indexicals as windows into the
properties of pronouns. Linguistic Inquiry 40: 187–237.
Larmarche, Jacques. 1991. Problems for N-movement to Num-P. Probus 3: 215–236.
249
Landau, Ian. 2006. Chain resolution in Hebrew V(P)-fronting. Syntax 9: 32–66.
Lekakou, M. and Szendrői, K. 2012. Polydefinites in Greek: Ellipsis, close apposition
and expletive determiners. Journal of Linguistics 48: 107–149.
Leu, Thomas. 2008. The internal syntax of determiners. Doctoral dissertation. New York
University, New York, NY.
Longobardi, Giuseppe. 1994. Reference and proper names: a theory of N-movement in
syntax and logical form. Linguistic Inquiry 25:609–665.
Longobardi, Giuseppe. 2008. Reference to individuals, person, and the variety of
mapping parameters. In Essays on nominal determination, ed. A. Klinge and H.
Müller, 189–211. Amsterdam: John Benjamins.
Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press.
Lyons, John. 1977. Semantics. Cambridge: Cambridge University Press.
Manolessou, Io. 2001. The evolution of the demonstrative system in MG. Journal of
Modern Greek Linguistics 2: 119–148.
Manzini, Rita, and Leonardo Savoia. 2002. Parameters of subject inflection in Italian
dialects. In Subjects, expletives and the EPP, ed. P. Svenonius, 157–200. Oxford:
Oxford University Press.
Mathieu, Eric. 1999. French WH in-situ and its restrictions. Ms., UCL.
Mathieu. Eric. 2004. The mapping of form and interpretation: the case of optional WHmovement in French. Lingua 114:1090–1132.
Matushansky, Ora. 2006. Head movement in linguistic theory. Linguistic Inquiry 37: 69–
109.
Medeiros, David. 2012. Economy of Command. Doctoral dissertation. The University of
Arizona, Tucson, AZ.
Murasugi, Keiko. 1991. Noun phrases in Japanese and English: A study in syntax,
learnability and acquisition. Doctoral dissertation. The University of
Connecticut, Storrs, CT.
Noguchi, Tohru. 1997. Two types of pronouns and visible binding. Language 43: 770–
797.
Nunes, Jairo. 1995. The copy theory of movement and linearization of chains in the
minimalist program. Doctoral dissertation. University of Maryland, College
250
Park, MD.
Nunes, Jairo. 2000. Linearization of chains and sideward movement. Cambridge, MA:
MIT Press.
Öztürk, Balkız. 2001. Turkish as a non-pro-drop language. In The verb in Turkish, ed. E.
Taylan, 239–260. Amsterdam: Benjamins.
Öztürk, Balkız. 2008. Non-configurationality: free word order and argument drop in
Turkish. In The limits of syntactic variation, ed. T. Biberauer, 411–440.
Amsterdam: Benjamins.
Panagiotidis, Phoevos. 2000. Demonstrative determiners and operators: The case of MG.
Lingua 110:717–742.
Panagiotidis, Phoevos. 2002. Pronouns, clitics and empty nouns. Amsterdam: John
Benjamins.
Panagiotidis, Phoevos. and Marinis. 2011. Determiner spreading as DP-predication.
Studia Linguistica 65: 268–298.
Perlmutter, David. 1971. Deep and surface constraints in syntax. New York: Holt,
Rinehart and Winston.
Pesetsky, David. 1978. Category switching and so-called pronoun. Proceedings of CLS
14: 350–360.
Pesetsky, David, and Esther Torrego. 2001. T-to-C movement: Causes and consequences.
In Ken Hale: A life in language., ed. M. Kenstowicz, 355–426. Cambridge, MA:
MIT Press.
Pesetsky, David, and Esther Torrego. 2007. The syntax of valuation and the
interpretability of features. In Phrasal and clausal architecture: Syntactic
derivation and interpretation, ed. S. Karimi, V. Samiian, and W. K. Wilkins,
262–294. Amsterdam: John Benjamins.
Platzack, Christer. 2003. Agreement and null subjects. In Tromsø University Working
Papers on Language and Linguistics Vol. 31, No. 2 (Proceedings of SCL 19).
Platzack, Christer. 2004. Agreement and the person phrase hypothesis. Working Papers in
Scandinavian Syntax 73: 83–112.
Postal, Paul. 1966. On so-called ‘pronouns’ in English. In Georgetown university
monograph series on languages and linguistics, Vol 19, ed. F. Dinneen, 177–206.
Washington DC: Georgetown University Press.
251
Postal, Paul. 1969. On so-called ‘pronouns’ in English. In Modern studies in English, ed.
D. Reible, and S. Schane, 201–224. Englewood Cliffs, N.J: Prentic Hall.
Radford, Andrew. 2009. Analysing English sentences: A minimalist approach.
Cambridge: Cambridge University Press.
Ritter, Elizabeth. 1991. Two functional categories in noun phrases: Evidence from
Modern Hebrew. In Syntax and Semantics 25: Perspectives on phrase structure,
ed. S. Rothstein, 37–62. New York: Academic Press.
Ritter, Elizabeth. 1995. On the syntactic category of pronouns and agreement. Natural
Language and Linguistic Theory 13: 405–443.
Rizzi, Luigi. 1982. Issues in Italian syntax. Dordrecht: Foris Publications.
Rizzi, Luigi. 1986. Null objects in Italian and the theory of pro. Linguistic Inquiry 17:
501–557.
Rizzi, Luigi. 1989. The new comparative syntax: principles and parameters of universal
grammar. Cahiers Ferdinand de Saussure 43: 65–78.
Rizzi, Luigi. 1991. Residual verb second and the wh-criterion. Ms., University of
Geneva, Geneva.
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed.
L. Haegeman, 281–337. Dordrecht: Kluwer.
Roberts, Ian. 2010. A deletion analysis of null subjects. In Parametric variation: Null
subjects in minimalist theory, ed. T. Biberauer, A, Holmberg, I. Roberts, and M.
Sheehan, 58–87. Cambridge: Cambridge University Press.
Roberts, Ian, and Anders Holmberg. 2010. Introduction: parameters in minimalist theory.
In Parametric variation: Null subjects in minimalist theory, ed. T. Biberauer, A,
Holmberg, I. Roberts, and M. Sheehan, 1–57. Cambridge: Cambridge University
Press.
Roberts, Ian. 2011. FOFC in DP: Universal 20 and the nature of demontratives. Ms.,
Cambridge University. Available at: http://ling.auf.net/lingbuzz/001502
Rosen, Nicole. 2003. Demonstrative position in Michif. The Canadian Journal of
Linguistics 48:39–69.
Rothstein, Susan. 1983. The syntactic forms of predication. Doctoral dissertation. MIT,
Cambridge, MA.
Rothstein, Susan. 2001. Predicates and their subjects (Studies in Linguistics and
252
Philosophy 74). Dordrecht/Boston: Kluwer.
Roussou, Anna, and Ianthi Maria Tsimpli. 1994. On the interaction of case and
definiteness in Modern Greek. Amsterdam Studies in the Theory and History of
Linguistic Science Series 4:69–76.
Safir, Ken. 1987. What explains the definiteness effect. In The representation of
(in)definiteness, ed. E. Reuland, and A. Meulen, 71–97. Cambridge MA: MIT
Press.
Sauerland, Uli. 2003. A new semantics for number. In Proceedings of the Thirteenth
Semantics and Linguistic Theory, ed. R. B. Young and Y. Zhou, 258–275. Ithaca,
NY: CLC Publications.
Sauerland, Uli, Jan Anderssen, and Kazuko Yatsushiro. 2005. The plural is semantically
unmarked. In Linguistic evidence: Empirical, Theoretical and Computational
Perspectives, ed. S. Kesper and M. Reis, 413–434. Berlin: Mouton de Gruyter.
Scontras, Gregory. 2013. A unified semantics for number marking, numerals, and
nominal structure. The Proceedings of Sinn und Bedeutung 17:545–562.
Senft, Gunter. 1986. Kilivila: The language of the Trobriand Islanders. Berlin: Mouton
de Gruyter.
Shlonsky, Ur. 2004. The form of the Semitic noun phrase. Lingua 114: 1465–1526.
Speas, Margaret. 1990. Phrase structure in natural language. Dordrecht: Kluwer
Academic Publishers.
Stavrou-Sifaki, Melita. 1995. Epexegesis vs. apposition in Modern Greek. In Scientific
bulletin of the school of philology Vol. 5, 217–250. Thessaloniki: Aristotle
University.
Stavrou, Melita. 1999. The position and serialization of APs in the DP: Evidence from
Greek. In Studies in Greek Syntax, ed. A. Alexiaoud, G. Horrocks, and M.
Stavrou, 201–225. Amsterdam: Benjamins.
Steddy, Sam, and Vieri Samek-Lodovici. 2011. On the ungrammaticality of remnant
movement in the derivation of Greenberg’s Universal 20. Linguistic Inquiry 42:
445–469.
Stirling, Lesley, and Rodney Huddleston. 2002. Deixis and anaphora. In The Cambridge
grammar of the English language, ed. R. Huddleston and G. Pullum, 1449–1564.
Cambridge: Cambridge University Press.
Stowell, Tim. 1983. Subjects across categories. The Linguistic Review 2:285–312.
253
Stowell, Tim. 1991. Small clause restructuring. In ed. Principles and parameters in
comparative grammar, ed. R. Friedin, 182–218. Cambridge, MA: MIT Press.
Stroik, Tim. 1994. Saturation, predication and the DP hypothesis. Linguistic Analysis
24:39–61.
Szabolcsi, Anna. 1983. The possessor that ran away from home. Linguistic Review 3: 89–
102.
Szabolcsi, Anna. 1987. Functional categories in the noun phrase. In Approaches to
Hungarian, Vol 2: Theories and analyses, ed. I. Kenesei, 167–190. Szeged: Jate.
Szabolcsi, Anna. 1994. The noun phrase. In The Syntactic Structure of Hungarian, Vol.
27: Syntax and semantics, ed. F. Kiefer, and K. Kiss, 179–275. San Diego:
Academic Press.
Tsimpli, Ianthi Maria. 1990. The clause structure and word order of Modern Greek. UCL
Working Papers in Linguistics 2:226–255.
Tsimpli, Ianthi Maria. 1995. Focusing in Modern Greek. In Discourse Configurational
Languages, ed. K. Kiss, 176–206. Oxford: Oxford University Press.
Tsimpli, Ianthi Maria. 1998. Individual and functional readings for focus, wh- and
negative operators: Evidence from Greek. In Theme in Greek Linguistics Vol. 2,
ed. G. Horrocks, B. Joseph, and I. Philippaki-Warburton, 197–227. Amsterdam:
John Benjamins.
Travis, Lisa. 1984. Parameters and effects of word order variation. Doctoral dissertation.
MIT, Cambridge, MA.
Tryon, Darrell T. 1967. Nengone grammar. Canberra: Australian National University.
Tryon, Darrell T. 1971a. Dehu grammar. Canberra: Australian National University.
Tryon, Darrell T. 1971b. Iai grammar. Canberra: Australian National University.
Vangsnes, Øystein. 1999. The identification of functional categories. Doctoral
dissertation. University of Bergen.
Vicente, Luis. 2009. A note on the copy vs. multidominance theories of movement.
Catalan Jounal of Linguistics 8:75–97.
Vlachos, Christos. 2008. Wh-in-situ inquiries in a wh-movement Language: the case of
Greek. Ms., University of Patras. Available at: http://ling.auf.net/lingBuzz/.
Vlachos, Christos. 2009. Wh-in-situ questions in Greek. In Proceedings of the First
254
Patras International Conference for Graduate Students in Linguistics, 180–196.
University of Patras.
Vlachos, Christos. 2010. Wh-in-situ: the case of Greek. In Movement and clitics, V.
Torrens, L. Escobar, A. Gavarró, and J. Gutiérrez, 84–111. Newcastle:
Cambridge Scholars Publishing.
Vlachos, Christos. 2012. Wh-constructions and the division of labour between syntax and
the interfaces. Doctoral dissertation. University of Patras.
Williams, Edwin. 1980 .Predication. Linguistic Inquiry 11:203–238.
Williams, Edwin. 1983a. Against small clauses. Linguistic Inquiry 14:287–308.
Williams, Edwin. 1983b. Semantic vs. syntactic categories. Linguistics and Philosophy
6:423–446.
Williams, Edwin. 1994. Thematic structure in syntax. Cambridge, MA: MIT Press.
Zamparelli, Roberto. 2002. Definite and bare kind-denoting noun phrases. In The
Proceedings of Going Romance 2000, Oxford University Press.
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
advertisement
© 2021