INTERROGATIVE FEATURES

INTERROGATIVE FEATURES
INTERROGATIVE FEATURES
by
Jason Robert Ginsburg
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
2009
2
THE UNIVERSITY OF ARIZONA
GRADUATE COLLEGE
As members of the Final Examination Committee, we certify that we have read the
dissertation prepared by Jason Robert Ginsburg
entitled Interrogative Features
and recommend that it be accepted as fulfilling the dissertation requirement for the
Degree of Doctor of Philosophy.
Date: 15 December 2008
Simin Karimi
Date: 15 December 2008
Andrew Barss
Date: 15 December 2008
Andrew Carnie
Date: 15 December 2008
Heidi Harley
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: 15 December 2008
Dissertation Director: Simin Karimi
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of 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 accurate acknowledgment of 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: Jason Robert Ginsburg
4
ACKNOWLEDGEMENTS
I would like to thank the members of my dissertation committee Simin Karimi,
Andrew Barss, Andrew Carnie, and Heidi Harley for their excellent guidance. I
would especially like to thank my advisor Simin Karimi for all of her help. She
read numerous drafts of my work and provided excellent comments. Our numerous
meetings led to many improvements in this work. I thank Heidi Harley for her
excellent suggestions and clear explanations. Thanks go to Andrew Carnie for his
detailed comments and suggestions. I also thank my language consultants: Dalila
Ayoun, Rusty Barrett, Junko Ginsburg, Simin Karimi, Sunjing Ji, and Hyun Kyoung
Jung. I am indebted to all of my professors at the University of Arizona. Special
thanks go to Sandiway Fong for his instruction in my minor of Computational
Linguistics and Mike Hammond for his guidance, both in the Spam lab and as
department head. Much thanks go to my classmates at the University of Arizona. I
thank my parents for their support and encouragement. Lastly, I am deeply indebted
to my wife Junko for her patience and support.
5
DEDICATION
To Junko and Rachel.
6
TABLE OF CONTENTS
ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
CHAPTER 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
CHAPTER 2 Theoretical background . . . . . . . . . . .
2.1 Introduction . . . . . . . . . . . . . . . . . . . . .
2.2 Operations of the faculty of language . . . . . . .
2.2.1 Lexical items . . . . . . . . . . . . . . . .
2.2.2 Numeration . . . . . . . . . . . . . . . . .
2.2.3 Select . . . . . . . . . . . . . . . . . . . .
2.2.4 Merge, Move, and Agree . . . . . . . . . .
2.3 Phrase structure . . . . . . . . . . . . . . . . . .
2.4 The left periphery . . . . . . . . . . . . . . . . .
2.5 Basic clause structure . . . . . . . . . . . . . . . .
2.6 Island effects . . . . . . . . . . . . . . . . . . . .
2.6.1 Government and Binding Theory accounts
2.6.2 Minimalist Program accounts . . . . . . .
2.6.3 The absence of island-effects . . . . . . . .
2.7 Intervention effects . . . . . . . . . . . . . . . . .
2.8 Head movement . . . . . . . . . . . . . . . . . .
2.9 Conclusion . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
20
20
21
22
22
23
23
25
30
38
43
44
47
64
66
68
70
CHAPTER 3 Qu-features and Qu-morphemes . . .
3.1 Introduction . . . . . . . . . . . . . . . . . .
3.2 Evidence for a Qu-feature . . . . . . . . . .
3.3 Multiple Qu-morphemes in a single language
3.4 Conclusion . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
71
71
71
75
82
CHAPTER 4 Yes/no constructions . . . . . . .
4.1 Introduction . . . . . . . . . . . . . . . .
4.2 Types of Qu-morphemes . . . . . . . . .
4.2.1 Overt Qu-morphemes . . . . . . .
4.2.2 Covert Qu-morphemes . . . . . .
4.2.3 An overt or covert Qu-morpheme
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
83
83
83
85
92
96
.
.
.
.
.
.
.
.
.
.
.
.
7
TABLE OF CONTENTS – Continued
.
.
.
.
.
.
.
.
.
.
.
.
98
99
99
101
103
108
123
127
131
132
134
134
CHAPTER 5 Qu-morphemes in wh-constructions . . . . . . . . . . . . . . .
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Clause-peripheral Qu-morphemes . . . . . . . . . . . . . . . . . . . .
5.2.1 Mandarin . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.2 Japanese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.1 The standard American English facts . . . . . . . . . . . . . .
5.3.2 Accounts of the subject/non-subject wh-construction assymmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.3 Analysis of English facts . . . . . . . . . . . . . . . . . . . . .
5.3.4 Ozark English . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4 Qu-morphemes in various positions . . . . . . . . . . . . . . . . . . .
5.5 Qu-morphemes and island effects . . . . . . . . . . . . . . . . . . . .
5.6 Remaining problems regarding TP-internal Qu-morphemes in Sinhala
5.6.1 Embedded wh-constructions . . . . . . . . . . . . . . . . . . .
5.6.2 Matrix wh-constructions . . . . . . . . . . . . . . . . . . . . .
5.6.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.7 Comparison with previous analyses . . . . . . . . . . . . . . . . . . .
5.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
136
136
138
138
139
141
141
141
CHAPTER 6 Wh-features in single wh-constructions
6.1 Introduction . . . . . . . . . . . . . . . . . . .
6.2 Agree . . . . . . . . . . . . . . . . . . . . . .
6.3 Wh-feature movement . . . . . . . . . . . . .
191
191
197
204
4.3
4.4
4.5
4.2.4 Summary . . . . . .
Qu-morphemes in Typ . . .
4.3.1 Modern Japanese . .
4.3.2 English . . . . . . . .
TP-internal Qu-morphemes
4.4.1 Sinhala . . . . . . . .
4.4.2 Okinawan . . . . . .
4.4.3 Premodern Japanese
4.4.4 Tupı́ . . . . . . . . .
4.4.5 Ewen . . . . . . . . .
4.4.6 Summary . . . . . .
Conclusion . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
145
150
154
158
159
170
177
177
181
183
183
189
8
TABLE OF CONTENTS – Continued
6.4
6.5
6.6
6.7
6.8
Wh-phrasal movement . . . . . . . . . . . . . . . . . . . . . . . . .
Partial wh-movement . . . . . . . . . . . . . . . . . . . . . . . . . .
6.5.1 German . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.5.2 Albanian . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.5.3 Remaining issues concerning partial wh-movement . . . . .
Wh-phrasal movement, Agree, and partial wh-movement in a single
language: Malay . . . . . . . . . . . . . . . . . . . . . . . . . . . .
French: wh-phrasal movement and wh-feature movement in a single
language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CHAPTER 7 Wh-movement that forms a wh-phrase
7.1 Introduction . . . . . . . . . . . . . . . . . . .
7.2 Wh-in-situ languages . . . . . . . . . . . . . .
7.3 Overt wh-movement languages . . . . . . . . .
7.3.1 Basque . . . . . . . . . . . . . . . . . .
7.3.2 Tlingit . . . . . . . . . . . . . . . . . .
7.3.3 English . . . . . . . . . . . . . . . . . .
7.4 Conclusion . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
215
221
222
231
238
. 243
. 252
. 259
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
262
262
263
284
284
288
292
294
CHAPTER 8 Multiple wh-constructions . . . . . . . . . .
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . .
8.2 Agree languages . . . . . . . . . . . . . . . . . . . .
8.3 Wh-feature movement followed by Agree: Japanese
8.4 Wh-phrasal movement followed by Agree: English .
8.5 Previous analyses of additional wh-effects . . . . .
8.6 Multiple wh-feature movement . . . . . . . . . . . .
8.7 Languages that disallow multiple wh-constructions .
8.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
295
295
296
298
302
305
309
314
316
CHAPTER 9 The weird behavior of why . . . . . . . . . . . . . .
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.2 Why is base generated in [Spec, TypP] . . . . . . . . . . . .
9.2.1 Japanese, Korean, and Persian . . . . . . . . . . . . .
9.2.2 Mandarin . . . . . . . . . . . . . . . . . . . . . . . .
9.3 Further evidence that why must appear in [Spec, TypP] . . .
9.4 Feature movement may proceed from why in some languages
9.5 Why and island effects . . . . . . . . . . . . . . . . . . . . .
9.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
319
319
322
323
327
329
333
339
342
9
TABLE OF CONTENTS – Continued
CHAPTER 10 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
10
ABSTRACT
There has been a great deal of work examining the structures of yes/no and whconstructions that has led to many important developments in linguistic theory. In
this dissertation, I extend this work by developing a theory that explains how the
behavior of Qu-morphemes (question morphemes) and wh-phrases in interrogative
constructions in several different languages is influenced by ‘interrogative features.’
The interrogative features are 1) a Qu-feature, which is responsible for typing a
clause as an interrogative, 2) a wh-feature, which is responsible for giving a whphrase scope, and 3) a Focus-feature, which is responsible for focusing certain relevant phrases. The main focus of this work is on explaining the influence of these
interrogative features on the positions of question morphemes and wh-phrases. In
the first part of this work, I examine the behavior of Qu-morphemes. I account for
why a Qu-morpheme must appear in the clause periphery in certain languages, such
as Japanese, whereas it can appear in a non-clause-peripheral position in other languages, such as Sinhala. I explain how a Qu-feature associated with a Qu-morpheme
types a clause and why there is variation in the positions of Qu-morphemes. The
second part of this work focuses on the behavior of wh-phrases. I account for why whconstructions can be formed with an in-situ wh-phrase (for example, in Japanese),
with movement of a wh-phrase to a scope position (for example, in English), or with
movement of a wh-phrase to a non-scopal position (for example, in some dialects
of German). I also examine other phenomena involving wh-phrases. I show how
wh-feature movement can influence well-formedness of a wh-construction. I explain
why, in certain cases, what would normally be an ill-formed construction can be
repaired via the addition of a wh-phrase. I examine why some languages, but not
others, allow multiple wh-constructions. Lastly, I investigate the odd behavior of
the wh-phrase why, which behaves differently from other wh-phrases. This work is
11
novel in that it provides a unified analysis of cross-linguistic and language internal
variation in the structures of yes/no and wh-constructions.
12
ABBREVIATIONS
2p = 2nd person pronoun
3p = 3rd person pronoun
acc = accusative marker
asp = aspect marker
comp = complementizer
conc = conclusive
conj = conjunct inflection
cl = classifier
D = Determiner
dat = Dative
decl = Declarative
dir = direct voice
e = empty category
emph = emphatic
erg = Ergative marker
foc = focus
fut = future tense
gen = genitive
hon = honorific
ind = indicative
inf = infinitive
loc = locative
neg = negation
nom = nominative marker
past = past tense
13
perf = perfective
pl = plural
pres = present tense
Qu = question
S = sentence
t = trace
tl = title prefix
top = topic marker
Notes: Throughout this work, for the sake of consistency, I have made slight changes
to example sentences from various sources. Unless otherwise indicated, English and
Japanese examples are my own.
14
CHAPTER 1
Introduction
The purpose of this dissertation is to account for why there is cross linguistic, as
well as language internal, variation in the formation of interrogative constructions. I
use ‘interrogative constructions’ to refer to yes/no and wh-questions, which request
responses of a listener, and also to embedded yes/no and wh-constructions, which
do not request a response.1 Working within the framework of the Minimalist Program (cf. Chomsky 1995b), I attempt to answer the following questions. 1) Why
can interrogative constructions be formed with overt question morphemes (for example, in Sinhala and Japanese) or with no overt question morpheme (for example,
in English and Persian)? 2) Why is there cross-linguistic and language internal variation in the positions of question morphemes? For example, in Japanese a question
morpheme must appear in a clause-peripheral position, but in Sinhala a question
morpheme can appear in a non-peripheral clause-internal position. 3) Why can
wh-constructions be formed with an in-situ wh-phrase (for example, in Japanese),
with movement of a wh-phrase to a scope position (for example, in English), or with
movement of a wh-phrase to a non-scopal position (for example, in some dialects of
German, and in Malay)? 4) Why do some languages, but not others, allow multiple
wh-constructions? While finding the answers to these questions, the goal of the
work presented here is to provide a unified analysis of yes/no and wh-constructions
that demonstrates the following.
• what yes/no and wh-constructions have in common.
1
Although these embedded constructions are not questions, I consider them to be interrogatives
since they share many similarities in form with matrix yes/no and wh-constructions.
15
• why there is cross-linguistic and language internal variation in the structures
of yes/no constructions.
• why there is cross-linguistic and language internal variation in the structures
of wh-constructions.
I set about achieving these explanatory goals by examining the roles of interrogative
features.
Interrogative features are elements that play a role in the formation of interrogative constructions. I focus on two distinct interrogative features: wh-features
and Qu-features (question features).2 A third type of feature, a Focus-feature, also
plays an important role in the formation of certain interrogative constructions. I
discuss evidence for the existence of these features in various languages and I attempt to show how these features influence the syntactic structures of interrogative
constructions.
Following work by Katz & Postal (1964), Aoun & Li (1993) argue that the
presence or absence of wh- and Qu-features in a clausal typing projection results in
different types of clauses, as shown in (1).
(1)
(a) [+Qu, +wh] = wh-question
(b) [+Qu, -wh] = yes/no question
(c) [-Qu, -wh] = statement
(d) [-Qu, +wh] = exclamatory statements (e.g., How good he is! )
(Aoun & Li 1993:232-233)
In this work, I focus on the types of constructions described by (1a) and (1b). When
a single Qu-feature occurs in a clausal typing projection, which I refer to as a TypP,3
2
Question features and question morphemes are often referred to in the literature with ‘Q’
instead of ‘Qu.’ In order to avoid confusion with quantificational elements, which sometimes are
also referred to with a ‘Q,’ I borrow the term “Qu” from Aoun & Li (1993) and Denham (2000).
3
See chapter 2.5 for discussion of TypP.
16
and there is no wh-feature, then a yes/no construction results. When a Qu-feature
and a wh-feature co-occur in TypP, then a wh-construction results. A Qu-feature is
responsible for a construction becoming an interrogative. A wh-feature is responsible
for a wh-phrase having scope.
The first part of this work focuses on the properties and behavior of Qumorphemes in yes/no constructions.4 Typical yes/no constructions are formed with
one of the three types of Qu-morphemes given in (2).
(2)
Overt
Affix
Qu1
X
∅
Qu2
∅
X
Qu3
∅
∅
Qu1 is an overt Qu-morpheme that is an independent lexical item, and occurs in languages such as Japanese. Qu2 is a null Qu-morpheme that is an affix,
and occurs in languages such as English. Qu3 is a null Qu-morpheme that is not an
affix. This type of Qu-morpheme results in a yes/no construction that is formed via
intonation alone, as in languages such as Hopi. The Qu-morpheme, depending on
the language, either 1) appears in the clause periphery (e.g., Japanese), or 2) in a
lower clause-internal position (e.g., Sinhala). When in the clause periphery, I argue
that the Qu-morpheme is in a clausal typing projection that I refer to as TypP.
Variation in the position of the Qu-morpheme depends on whether a Qu-morpheme
contains only a Qu-feature as in (3a), or both a Qu-feature and a Focus-feature, as
in (3b). The Qu-morpheme in (3a) occurs in a clause-peripheral position and the
Qu-morpheme in (3b) occurs in a non-peripheral position within a clause.
4
I examine yes/no constructions that have a mono-clausal structure. I do not discuss yes/no
constructions which may have a bi-clausal structure, such as Mandarin A-not-A questions (cf.
Huang 1982, Hagstrom 2006, among others).
17
(3)
(a) Qu[FQu ]
(b) Qu[FQu ,FF oc]
These two different types of Qu-morphemes differ with respect to their semantic
effects. The Qu-morpheme in (3a) serves the purpose of typing a clause as an
interrogative, whereas that in (3b) both types a clause as an interrogative and
focuses a phrase within the clause.
The second part of this dissertation focuses on wh-constructions.
A wh-
construction is formed with the same types of Qu-morphemes used in yes/no constructions. The Qu-morpheme surfaces as an overt element, a null affix, or intonation, as in (2) above. In addition, a wh-construction crucially requires the presence
of a wh-phrase that either appears in-situ (in the position in which it receives its
theta-role), or undergoes movement. For example, in Japanese, a wh-phrase may
remain in-situ, whereas in English a single wh-phrase must move. In some cases,
such as in some dialects of German, a wh-phrase may even move to a non-scopal
position and have its scope marked by another element.
I begin my analysis of wh-constructions by examining the position of a Qumorpheme in a wh-construction. As in yes/no constructions, a Qu-morpheme can
occur in a clause-peripheral position. There are also languages that allow (and
sometimes require) a Qu-morpheme to occur in a non-peripheral position. I argue
that this distinction is dependent on whether the Qu-morpheme has a Focus-feature,
as shown in (3b) above, or lacks a Focus-feature, as shown in (3a). A Focus-feature
forces a Qu-morpheme to appear in a position adjacent to a focused in-situ whphrase. Otherwise, if it lacks a Focus-feature, it appears in the clause-periphery.
Next, I examine the role of a wh-feature in a wh-construction. I argue that
the means by which a wh-feature establishes a relationship with the clausal typing
projection TypP accounts for whether or not the wh-phrase remains in-situ or moves.
The wh-feature can undergo an Agree relation with the head of TypP. In this case,
there is no movement of the wh-feature or wh-phrase to TypP. The other option
18
is for the wh-feature to move to TypP. When this movement occurs, either the
wh-feature moves alone, or it moves together with its associated wh-phrase. This
movement is argued to be a different operation from Agree. These three options are
shown below.
(4)
Operation Description
Agree
No movement of wh-feature or wh-phrase
Move
Wh-feature moves alone
Move
Wh-phrase moves
In the following chapters, I discuss the formation of yes/no and wh-constructions
and the roles that Qu-, wh-, and Focus-features play in their formation. I attempt to
provide a unified analysis that accounts for what yes/no and wh-constructions have
in common, why some yes/no constructions have overt Qu-morphemes and others
do not, why there is variation in the positions of Qu-morphemes, and why there is
variation in the positions of wh-phrases.
The organization of this dissertation is as follows. In chapter 2, I explain the
basic theoretical background that I utilize in this work. I begin by providing a
brief overview of important assumptions of the Minimalist Program about how an
utterance is formed. Then I explain the basic clause structure that I utilize. This
is followed by discussion of syntactic phenomena that play an important role in
my analysis. In chapter 3, I explain what a Qu-feature is and why it surfaces in
yes/no and wh-constructions. In chapter 4, I examine the role of the Qu-morpheme
in the formation of yes/no constructions. I argue that features contained within
a Qu-morpheme account for cross-linguistic differences in the surface structure of
yes/no constructions. These features determine whether or not a Qu-morpheme is
pronounced and whether or not it is an affix. I next account for the fact that a Qumorpheme must occur in a clause-peripheral position in some languages, but can
19
occur TP-internally in other languages. In chapter 5, I account for cross-linguistic
variation in the positions of Qu-morphemes in wh-constructions. I examine languages in which a Qu-morpheme must appear in the clause-periphery in a whconstruction, and I examine languages in which a Qu-morpheme may either appear
in the clause-periphery or in a non-peripheral position adjacent to an in-situ whphrase. In chapter 6, I argue that the means by which a wh-feature values a probe
in the head of a clausal typing projection, TypP, determines whether or not there
is movement of a wh-element, and thus influences the form of a wh-construction. I
argue that there are three ways in which a wh-feature can form a relationship with
a probe in TypP. It can form an Agree relation, it can move to Typ, or there can
be movement of a wh-phrase to the specifier of TypP, in which case the wh-feature
values the probe in Typ via a Spec-head relation. In chapter 7, I discuss how, in
certain cases, a wh-feature can turn a larger phrase into a wh-phrase. I examine
how this wh-feature movement can circumvent potential island effects. In chapter
8, I discuss the roles of interrogative features in the formation of certain multiple
wh-constructions. In particular, I examine how multiple wh-features establish a
relationship with a probe in the clausal typing projection TypP. In chapter 9, I
examine the behavior of the wh-phrase why, which behaves differently from other
wh-phrases in a number of languages. Chapter 10 is the conclusion.
20
CHAPTER 2
Theoretical background
2.1 Introduction
This work is written within the Principles and Parameters framework, in which
there is claimed to be an innate module of the brain that contains a universal set
of principles (a Universal Grammar) that humans use to produce and understand
language. One of the main ideas within the Principles and Parameters program is
that languages do not differ very much underlyingly. The apparent differences in
languages result from minor variations in the settings of principles contained within
the language module.
The Minimalist Program (Chomsky 1995b) is a modern approach to Generative
Grammar. According to Chomsky (1999:1), the Minimalist Program has the goal
of answering the following question: “to what extent is the faculty of language
FL an optimal solution to minimal design specifications, conditions that must be
satisfied for language to be usable at all?”. The Minimalist Program relies on the
assumptions that the properties of the faculty of language are minimalistic; they are
as simple as they can be in order to carry out their functions.
In this chapter, I discuss the basic assumptions about language that I rely on in
this work. Section 2.2 explains, from the perspective of the Minimalist Program, the
operations that are utilized by the human mind to construct phrases. Section 2.3
discusses the structures of phrases and clauses. Section 2.4 explains Rizzi’s (1997,
2001) view of the elaborated left-periphery of a clause; the part of a clause where
clausal typing occurs and where a wh-phrase obtains scope. Section 2.5 explains
the clause structure that I utilize in this work. Section 2.6 discusses what island
21
effects are and how they have been accounted for in syntactic theory, and section 2.7
does the same for intervention effects. Section 2.8 briefly discusses head movement.
Section 2.9 is the conclusion.
2.2 Operations of the faculty of language
The innate faculty of language contains a set of features that are combined to form
lexical items and phrases. I use the term ‘feature’ to refer to something akin to an
‘atom’ of language that, when combined with other features, is used to construct
human language. Chomsky (1999:7) writes:
FL [the faculty of language] specifies the features F that are available to
fix each particular language . . . We adopt the conventional (usually tacit)
assumption that L [a language] makes a one-time selection [FL ] from F.
These are the features that enter into L; others can be disregarded in
use of L.
A subset of the universal set of features is utilized in a particular language, and the
other features are essentially ‘set aside’ and not used in that language.
There are several operations, given in (1), that I assume are used to construct
utterances from these features. These operations are built into the language faculty;
they are part of what makes the language faculty what it is.
(1)
(a) Form lexical items: Combine features into bundles that become lexical
items at Spell-Out.
(b) Form a numeration: Construct a numeration from lexical items.
(c) Select: Select lexical items from the numeration.
(d) Merge: Combine two objects together.
These operations (1a-d) always occur to form all utterances.1
1
Chomsky (2006) claims that a single word utterance, such as No!, does not involve Merge.
However, it could be that even in a single word utterance, the single lexical item Merges with
22
2.2.1 Lexical items
First of all, a language contains a set of lexical items. A lexical item is essentially
a bundle of features that is pronounced as dog, cat, etc., when an utterance is
pronounced (at Spell-Out). Chomsky (1999:7) writes:
L [a language] assembles [FL ] [the features that have been selected for
use in a particular language] to lexical items LI of a lexicon Lex, the LIs
then entering into computations as units.
These features that make up a lexical item determine its properties, such as its
interpretation, where it can occur in an utterance, how it is pronounced, etc.
2.2.2 Numeration
Once lexical items are formed, a numeration is constructed. The language faculty
selects lexical items (actually, the feature bundles that correspond to lexical items
at Spell-Out) that it needs in order to produce a particular utterance that expresses
something akin to what the speaker desires to express. This collection of lexical
items is referred to by Chomsky (1999, 2004) as a “lexical array” if each lexical item
only occurs once, or as a “numeration” if one or more lexical items are selected from
the lexicon more than once. For example, the sentence I am hungry is formed from
the lexical array {I, am, hungry} since each lexical item only occurs once, but the
sentence She ate the food on the table is formed from the numeration {she, ate, the2 ,
food, on, table}, where the subscript 2 refers to two occurrences of the, because the
occurs twice.2 For the sake of simplicity I use the term numeration regardless of
whether or not a lexical item is selected more than once.
another element, but the other element is null. For example, perhaps no merges with a null Tense
element.
2
A lexical array is a set; it only only contains distinct lexical items. A numeration, as I
understand it, is not a set because it contains one or more lexical items that occur more than
once. A numeration, though, could be a set if each occurrence of a single lexical item is distinct;
23
2.2.3 Select
The lexical items in the numeration of an utterance are Selected for insertion into
a derivation. This is an operation that Chomsky (1995b:226) refers to as Select. A
lexical item that is Selected is inserted into a derivation via the operation Merge.
2.2.4 Merge, Move, and Agree
Merge, at least on the surface, is fairly simple: two elements combine. Chomsky
(1999:2) describes Merge as an operation that:
. . . takes two syntactic objects α and β and forms the new object γ =
{α, β}. We assume further that γ is of some determinate type: it has
label LB (γ). In the best case, LB (γ) = LB (α) or LB (β), determined
by general algorithim.
Basically, when two elements are combined via Merge, the Merged structure has the
label of one of the elements that has been combined. This label refers to what the
phrase type is. For example, when a verb combines with a Determiner Phrase (DP),
the resulting structure is a verb phrase (VP), not a DP, and so the phrase has the
label of the verb.
Merge can apply in two situations. The first type of Merge is when an element
is selected from a Numeration and Merged to another element in the process of a
derivation. Chomsky (2004, 2006) refers to this as ‘external Merge.’ The second
type of Merge is when an element that has already been Merged into a derivation
is Merged again in a new position. Chomsky (2004, 2006) refers to this as ‘internal
Merge.’ The term ‘internal’ refers to the fact that internal Merge is an operation
in which an element Merges onto a syntactic object that it is already a part of, i.e.,
internal to, at least in its base position. Chomsky (2004:110) writes that when two
for example, if the1 and the2 are distinct elements. If this view is correct, then there is no need to
make a distinction between a lexical array and a numeration.
24
elements α and β are Merged, “[u]nder external Merge, α and β are separate objects,
under internal Merge, one is part of the other, and Merge yields the property of
“displacement”.” Crucially, internal Merge is an operation that has generally been
referred to as ‘movement’ in the literature. In this work, I refer to ‘movement’
instead of to ‘internal Merge,’ as the notion of movement clearly describes the fact
that internal Merge results in displacement. Note that movement involves Merge of
an element that has already been Merged, a view captured by Epstein et al. (1998),
who refer to this process as ‘Remerge.’ In current work in the Minimalist Program,
movement is generally thought to involve copying of an element from one position
and then re-Merging it to a new position (Chomsky 1993), (also see Cover & Nunes
2007).
Once an element has been Merged into a derivation, it can form an Agree relation
with another element in a syntactic structure. Chomsky (1999:3) refers to Agree as a
relation that holds ‘between α and β, where α has interpretable inflectional features
and β has uninterpretable ones, which delete under Agree.” It has been claimed, for
example by Chomsky (1999, chomsky00) that Agree is subject to locality effects (i.e.,
Agree can be blocked by intervening elements) and, that movement is triggered by
Agree. I take a different position in this work, arguing that Agree is not subject to
locality effects and that movement need not be triggered by Agree. Specifically, there
are certain configurations in which movement is blocked; movement can be blocked
by intervening elements, and movement out of certain types of clauses is barred.
Agree, on the other hand, is not subject to these blocking effects. Furthermore,
movement is a distinct operation from Agree. I give an overview of the two types of
important locality effects discussed in this work; island and intervention effects in
sections 2.6-2.7. Movement is discussed in chapter 4, and all subsequent chapters.
The distinction between Agree alone versus, Agree followed by Move is discussed in
detail in chapter 6, and this distinction plays an important role in the subsequent
chapters.
25
The operations of Merge, Agree, and Move differ in terms of Economy, where
Economy refers to how much work the faculty of language must put out for any
given operation. The less work an operation involves, the more economical it is.
Chomsky defines the Principle of Economy as follows, where convergence refers to
a derivation that is successful (i.e., well-formed).
(2)
Principle of Economy
At each stage of a derivation, apply the most economical operation that
leads to convergence. (Chomsky 1995b:367)
Merge (external Merge) is the most economical operation, Move is the least economical, and Agree falls in-between. Merge of an element from a numeration into a
derivation must occur for any element to appear in a derivation and so it is the simplest operation. According to Chomsky (2001b:3), Merge “comes free.” Agree, on
the other hand, involves Merging an element followed by an Agree relation to check
an uninterpretable feature. Movement is the most complex operation. It involves
several steps; an element α is Merged into a derivation, an element β attracts alpha
and forces it to be Merged again in a new position. A single element may undergo
this Move operation several times, when moving through different clauses, etc. If
it is the case that the most economical operations are preferred, then Merge alone
is preferable to Agree or Move, and Agree is preferable to Move. Agree and Move
occur only when they have to. Due to its complexity, Chomsky (2000:102) refers to
Move as a “last resort.”
2.3 Phrase structure
The operations Select, Merge, Agree, and Move construct phrases. I represent
phrases using a typical X-notation such as that in (3) (cf. Chomsky 1967, Jackendoff
26
1974, Jackendoff 1977, among others) as a convenience to the reader.3
(3)
XP
YP
X’
ZP
X’
X
WP
Probably the most important phrase structural relationship is that of c-command
(cf. Klima 1964, Langacker 1969, Lasnik 1976, Reinhart 1976, Stowell 1981, Aoun &
Sportiche 1981, among others), and the notion of c-command relies on the concept
of domination. A node dominates any nodes that are contained within its branches.
For example, in (3), the XP node branches into YP and X’. Therefore, it dominates
these nodes. Furthermore, XP dominates any nodes that these nodes dominate.
YP has no branches and does not dominate anything. But the higher X’ node
dominates ZP and the lower X’, as well as everything within X’. In this manner, XP
actually dominates all of the nodes in the tree, except for possibly itself (depending
on how domination is defined). The structural relationship of c-command is defined
by Chomsky & Lasnik (1995:35) as follows.
(4)
c-command
α c-commands β if α does not dominate β and every γ that dominates α
dominates β. (Chomsky & Lasnik 1995:35)
According to this definition, in (3), YP c-commands all of the nodes in the tree
except for XP. XP is the only node that dominates YP, and XP also dominates all
of the nodes in the tree. Therefore, every node that dominates YP dominates all of
3
Within the Minimalist Program, phrase structure is generally thought to have a structure in
accord with the ideas developed in Chomsky (1995a).
27
the other nodes, and YP then c-commands all of the other nodes, except for XP. 4,5
Another important concept involving phrase structure is that of the Extended
Projection Principle (EPP). The EPP was originally formulated as the requirement
that all clauses have a subject (Chomsky 1981:27). Chomsky & Lasnik (1995:55)
write that “The Extended Projection Principle (EPP) states that [Spec, IP] is obligatory, perhaps as a morphological property of I or by virtue of the predicational
character of VP.” The notion that the EPP is associated with IP (or TP) has been
extended to include other categories, which may or may not have an EPP feature
(Chomsky 2000). This view is summed up by Miyagawa (2001:295) as follows: “the
EPP-feature is associated with “core functional categories,” which are C, T, and v.”
Thus, the EPP feature is viewed as a requirement that certain functional heads have
a filled specifier. Furthermore, Alexiadou & Anagnostopoulou (1998) argue that in
some cases, the EPP feature can be a requirement that a head position, and not
necessarily a specifier, be filled (also see Miyagawa 2001). For example, a C head
can have an EPP feature that is satisfied by movement of an element to C rather
than to [Spec, CP]. This EPP feature will play an important role in my analyses of
wh-constructions.
I next turn to clause structure. Within the Minimalist Program, a clause is generally assumed to contain the following basic projections: Complementizer Phrase
(CP), Tense Phrase (TP), verb phrase (v P), and Verb Phrase (VP). These are shown
in (5).
4
5
See Epstein et al. (1998) for arguments that c-command results from Merge and Move.
See Chomsky (2006) for discussion of eliminating the notion of c-command.
28
(5)
CP
C’
C
TP
DP1
T’
T
vP
t1
v’
v
VP
V’
V
DP2
In (5), there is a split-VP shell (cf. Larson 1988, Hale & Keyser 1993, and Chomsky
1995b, among others). Also, a Determiner Phrase (DP) has raised from [Spec,
v P] to [Spec, TP], in accord with the VP-internal subject hypothesis (Koopman &
Sportiche 1991).
According to Phase Theory (Chomsky 1999, 2000), a clause is derived via phases
that are constructed separately. Lexical items are Selected and are grouped into subnumerations (or subarrays). A sub-numeration is part of a numeration that consists
of just the lexical items in a particular phase. Chomsky (1999:9) defines phases
as “verbal phrases with full argument structure and CP with force indicators.”
Chomsky (2006:143) writes that CP “is shorthand for the region that Rizzi (1997)
calls the “left-periphery” (see section 2.4). Furthermore, Chomsky notes, following
work by Svenonious (2004) and Hiraiwa (2005), that DP might also be a phase,
although he does not take a position on this issue.
29
The lexical items in a sub-numeration are Merged into a derivation to form a
phase. The faculty of language continues to construct phases from sub-numerations
until the numeration is emptied; all lexical items have been Selected. Once a phase
is completed, only the elements located in the head and specifier positions of the
highest projection of a phase can be accessed, as stated in the Phase-Impenetrability
Condition given in (6), where α is a phase. The head H refers to the head of
the highest projection in the phase and the edge refers to the specifier(s) of this
projection (Chomsky 2000:108).
(6)
Phase-Impenetrability Condition
In phase α with head H, the domain of H is not accessible to operations
outside α, only H and its edge are accessible to such operations. (Chomsky
2000:108)
Chomsky (2000:108) further writes:
[O]perations cannot “look into” a phase below its head H. H itself must
be visible for selection and head movement; hence, its Specs must be as
well.
Once a derivation is complete, then Spell-Out applies. Chomsky (1995b:189) refers
to Spell-Out as an operation “which switches to the PF [Phonological Form] component.” At this stage, the phases that have been produced are pronounced. Interestingly, they are pronounced in the reverse of the order that they were formed;
i.e., the phase that produced last is pronounced first. Beyond Spell-Out there is the
level of Logical Form (LF) which is where meaning is interpreted.
The notions of probes and goals are also of importance within Phase Theory.
A probe is a feature inside of a head that ‘probes for’ a ““matching” feature[ ]
that establish[es] agreement (Chomsky 2000:122).” The matching feature is a goal.
When the probe and the goal match, the probe is erased. When there is an Agree
relation, an uninterpretable feature on a probe is eliminated without any movement.
30
In this work, I will also take the position that movement of a goal, without an Agree
relation, can eliminate an uninterpretable feature on a probe.
2.4 The left periphery
The phrase structure that I utilize in this work follows closely that proposed by
Rizzi (1997, 2001). In this section, I discuss Rizzi’s proposed ‘left-periphery’ of a
clause, and in the next section, I explain the clause structure that I utilize.
Rizzi (1997, 2001) argues for the existence of an elaborated CP that consists of a
number of separate projections. Rizzi’s proposal hinges on the idea that a single CP
projection cannot account for the structure of the part of the clause that he refers to
as “the left periphery” (the top part) of a clause. According to Rizzi, a CP projection
contains specifications for Force and Finiteness. Force is associated with clause
typing; whether a clause is a statement, interrogative, etc. Finiteness determines
whether a clause is finite or non-finite. These Force and Finiteness specifications may
occur together on a single C head, in which case there is no elaborated CP projection,
or they may split into different projections when necessary. When a focused and/or
topicalized phrase occurs, then CP splits into several projections. Similarly, when
there are separate Force and Finite elements, then CP must take on an elaborated
structure. At the highest end of this split CP structure is a Force Phrase (ForceP).
At the lowest position is a Finite Phrase (FinP). Between these two projections is
one or more Topic Phrases (TopP). There may also be a single Focus Phrase (FocP),
which can be sandwiched between TopP projections. Rizzi’s proposed left-periphery
clause structure (Rizzi 1997:297) is shown below. The Kleene star after each TopP
represents the possibility that there may be multiple TopP projections.
31
(7)
ForceP
TopP*
FocP
TopP*
FinP
IP
Under Rizzi’s analysis, an elaborated CP occurs when there is a topicalized
and/or focused phrase that is fronted to clause-initial position in languages such
as Italian. A topicalized element presents old information and a focused element
presents new information. Rizzi (1997:285) writes:
The topic is a preposed element characteristically set off from the rest of
the clause by “comma intonation” and normally expressing old information, somehow available and salient in previous discourse. . . [A] preposed
element, bearing focal stress, introduces new information. . .
In (8), your book is a topic, presumably occuring in [Spec, TopP], and presents old
information. The rest of the construction presents new information.
(8)
Your book1 , you should give t1 to Paul (not to Bill). (Rizzi 1997:285)
In (9), the focused element YOUR BOOK, which occurs in [Spec, FocP], presents
new information, and the rest of the construction is “. . . contextually given information, knowledge that the speaker presupposes to be shared with the hearer (285).”
32
(9)
YOUR BOOK you should give t to Paul (not mine). (Rizzi 1997:285)
Rizzi points out that in Italian, elements in TopP and FocP show different syn-
tactic behavior, which is further evidence that they are different projections. First
of all, a topicalized phrase may occur with a coreferential resumptive clitic, but
a focused phrase may not. Secondly, a topicalized phrase is not subject to weakcrossover effects, but a focused phrase is. And thirdly, quantificational elements
can be focused, but not topicalized. Also, a clause may contain multiple topicalized phrases, but only a single focused phrase. Rizzi argues that these facts are a
result of focused elements being quantificational, whereas topicalized elements are
not quantificational.
Rizzi argues that the existence of a ForceP and FinP at opposite ends of an
articulated CP projection accounts for the assymetric behavior of complementizers
in Italian. For example, he argues that the finite complementizer che occurs in
ForceP and the non-finite complementizer di occurs in FinP. Example (10a) contains
the finite complementizer che and (10b) contains the non-finite complementizer di.
(10) Italian:
(a) Credo [che loro apprezzerebbero molto il tuo libro].
‘I believe that they would appreciate your book very much.’
(b) Credo [di apprezzare molto il tuo libro].
‘I believe ‘of’ to appreciate your book very much.’ (Rizzi 1997:288)
In each of these examples the complementizer occurs at the beginning of the embedded clause. However, when there is a Clitic Left Dislocated (CLLD) argument,
the distribution of the complementizers differs. The finite complementizer che must
precede a topicalized phrase and the non-finite complementizer di must follow a topicalized phrase. In (11a), che precedes and in (11b), di follows, the CLLD phrase il
tuo libro ‘your book.’
33
(11) Italian:
(a) Credo [che il tuo libro, loro lo apprezzerebbero molto].
‘I believe that your book, they would appreciate it a lot.’
(b) Credo [il tuo libro di apprezzarlo molto].
‘I believe, your book, ‘of’ to appreciate it a lot.’ (Rizzi 1997:288)
The examples in (12) below show that ill-formedness results when the finite che
follows and when the non-finite di precede the CLLD argument.
(12) Italian:
(a) *Credo [il tuo libro che loro lo apprezzerebbero molto].
‘I believe, your book, that they would appreciate it a lot.’
(b) *Credo [di il tuo libro apprezzarlo molto].
‘I believe ‘of’ your book to appreciate it a lot.’ (Rizzi 1997:288)
Rizzi accounts for these data by arguing that the finite che occurs in the ForceP
head and the non-finite di occurs in the FinP head. The finite che must precede
a topicalized phrase because TopP is in a lower position than ForceP, which is
the highest projection in a clause. The non-finite di may follow a topicalized phrase
because it is in the lowest position of the left periphery, below any TopP projections.
I next turn to Rizzi’s view of the position of a wh-phrase. When there is an
elaborated CP, a wh-phrase clearly cannot occur in [Spec, CP], as there is no unique
CP projection. Rather, a wh-phrase must occur in some position within the leftperiphery. Rizzi argues that this position is [Spec, FocP]. Evidence from Italian
shows that a wh-phrase may co-occur with a topicalized phrase, but not with a
focused phrase. For example, in (13a), the topicalized phrase A Gianni ‘to Gianni’
precedes the wh-phrase che ‘what,’ and the result is well-formed. On the other
hand, (13b) shows that the focused phrase A GIANNI ‘to Gianni’ cannot co-occur
with a wh-phrase regardless of whether the wh-phrase precedes or follows it.
34
(13) Italian:
(a) A Gianni, che cosa gli hai detto
‘To Gianni, what did you tell him?’
(b) (*A GIANNI) che cosa(*A GIANNI) hai detto (, non a Perio)
‘(TO GIANNI) what (TO GIANNI) did you tell(, not to Pero)? (Rizzi
1997:291)
These facts can be accounted for if a wh-phrase moves to [Spec, FocP]. Since there
can only be one FocP projection and it has a single specifier, a wh-phrase and a
focused phrase cannot co-occur. A wh-phrase and a topicalized phrase occur in
separate projections, and so they may co-occur.
Rizzi revises this analysis in Rizzi (2001) by arguing for the existence of a unique
interrogative projection (Int) that is located between ForceP and FocP. This Int
projection may be sandwiched between TopP projections, as shown in (14).
(14) Force (Top*) Int (Top*) Foc (Top*) Fin IP (Rizzi 2001:289)
Following Rizzi (2001), (7) can then be revised as in (15) below, with an added IntP
projection.
35
(15)
ForceP
TopP*
IntP
TopP*
FocP
TopP*
FinP
IP
Evidence for IntP comes from the appearance of an interrogative yes/no particle
se in embedded clauses in Italian. This particle may be followed by the focused
phrase questo ‘this,’ as shown in (16a) but it may not be preceded by questo ‘this,’
as shown in (16b).
(16) Italian:
(a) Mi domando se QUESTO gli volessero dire (non qualcos’ altro).
‘I wonder if THIS they wanted to say to him, not something else.’
(b) *Mi domando QUESTO se gli volessero dire (non qualcos’ altro).
‘I wonder THIS if they wanted to say to him, not something else.’
(Rizzi 2001:289)
Because se can only be followed, and not preceded, by a focused phrase, Rizzi argues
that it must be higher than FocP. Furthermore, se can be preceded and followed by
a topic, an indication that it can occur between TopP projections, as shown below.
36
(17) Italian:
(a) Non so se, a Gianni, avrebbero potuto dirgli la verità.
‘I don’t know if to Gianni, they could have said the truth.’
(b) Non so, a Gianni, se avrebbero potuto dirgli la verità.
‘I don’t know, to Gianni, if they could have said the truth.’ (Rizzi
2001:289)
Rizzi provides evidence that se must occur in a projection separate from ForceP.
The particle se can be preceded by a topic as shown in (17b) above, but an element
in Force cannot. For example, the declarative complementizer che, which Rizzi
argues is in Force, cannot be preceded by a topic. Example (18a) is fine because
the topic a Gianni ‘to Gianni’ follows che. However, example (18b) is ill-formed
because the topic a Gianni ‘to Gianni’ precedes che.
(18) Italian:
(a) Credo che a Gianni, avrebbero dovuto dirgli la veritá.
‘I believe that to Gianni, they should have said the truth to him’
(b) *Credo, a Gianni, che avrebbero dovuto dirgli la veritá.
‘I believe, to Gianni, that they should have said the truth to him’ (Rizzi
2001:289)
If se were in Force, then it should not be able to follow a topic. Further evidence
that IntP is separate from ForceP is, according to Rizzi, that Spanish contains
constructions in which both the heads of ForceP and IntP are filled. In (19), que
‘that’ is in Force and si ‘if’ is in Int.
37
(19) Spanish:
Maria decia que si debièramos dejarlas en paz.
‘Maria was saying that if we shouldn’t leave them in peace.’ (Plann
1982:300, per Rizzi 2001:290)
As discussed above, Rizzi (1997) argues that most wh-phrases occur in [Spec,
FocP], as they are not compatible with a focused phrase. In Rizzi (2001) this view is
altered to the idea that, at least in Italian, wh-arguments occur in [Spec, FocP] but
wh-adverbials occur in [Spec, IntP]. Whereas wh-arguments cannot co-occur with
focused phrases, as shown in (13) above, a wh-adverbial can occur with a focused
phrase. For example, in (20a), the wh-adverbial perchè ‘why’ precedes the focused
phrase questo ‘this.’ (20b) shows that the wh-adverbial cannot follow the focused
phrase. If the wh-adverbial is in IntP, then the IntP projection must be in a position
above, and separate from, FocP.
(20) Italian:
(a) Perché QUESTO avremmo dovuto dirgli, non qualcos’ altro.
‘Why THIS we should have said to him, not something else?’
(b) *QUESTO perché avremmo dovuto dirgli, non qualcos’ altro.
‘THIS is why we should have said to him, not something else?’ (Rizzi
2001:294)
In summary, Rizzi (1997) argues that a CP can split into a ForceP (the locus
of clause typing), a FinP (the locus of finiteness), one or more TopP projections,
as well as at most one FocP projection. The later analysis in Rizzi (2001) adds an
IntP projection to the left-periphery that can house an interrogative particle in its
head position and a wh-adverbial in its specifier position. This IntP occurs between
ForceP and FocP, and it may be sandwiched between TopP projections.
I next turn to a discussion of the basic clause structure that I utilize, which is
similar to that developed by Rizzi.
38
2.5 Basic clause structure
The clause structure that I utilize in this work follows closely that of Rizzi (2001)
except that I use what I refer to as a Type Phrase (TypP) instead of Rizzi’s IntP.6
I use TypP,7 since I believe that this projection is where clausal typing elements,
and not just interrogative complementizers, occur. Rizzi argues that ForceP, not
IntP, is where clausal typing occurs. My view differs in that I see a Force head as
containing an element that indicates that a clause is embedded (and possibly also
matrix),8 whereas a Typ head contains an element that indicates whether a clause
is a statement, interrogative, etc. The basic clause structure that I assume in this
work is shown in (21).
(21)
ForceP
TopP*
TypP
TopP*
FocP
TopP*
FinP
TP
6
Also, in accord with recent work in the Minimalist Program, I use TP instead of IP.
I am essentially borrowing Denham’s (2000) clausal typing projection TyP.
8
It is not clear to me whether or not matrix clauses have a ForceP. I briefly discuss this issue
7
below.
39
The separate ForceP and TypP projections account for the fact that in some
languages, a complementizer can precede a question morpheme (Qu-morpheme), as
in (19) above. Under the proposed phrase structure in (21) above, the embedded
clause has the following structure in which que ‘that’ is in Force and si ‘if’ is in
Typ. Under Rizzi’s proposal, si ‘if’ would be in Int.
(22)
ForceP
Force’
Force
que
that
TypP
Typ’
Typ
si
if
TP
debièramos dejarlas en paz
shouldn’t leave them in peace
Kishimoto (2005) gives the following example from Sinhala9 in which the embedded
clause contains a Qu-morpheme [email protected] followed by the complementizer [email protected] ‘that.’
9
This is an Indo-Aryan language that is spoken in Sri Lanka (Gair & Sumangala 1991:93).
40
(23) Sinhala:
Ranjit [kauru aawa
[email protected] [email protected]] [email protected]
Ranjit who
came.A Qu that
know.A
‘Ranjit knows who came.’ (Kishimoto 2005:6)
The embedded clause has the following structure, in which [email protected] is in Typ and [email protected]
‘that’ is in Force.
(24)
ForceP
Force’
TypP
Typ’
TP
kauru aawa
who came-A
Force
[email protected]
that
Typ
[email protected]
Qu
Japanese also can contain a Qu-morpheme followed by a complementizer, as in (25),
in which kadooka ‘Qu’ is followed by to ‘COMP.’
(25) Japanese:
Watashi-wa [kare-ga sore-o
yatta kadooka to]
tazunemashita.
I-top
he-nom that-acc did
Qu
comp asked
‘I asked if he did that.’
In this case, kadooka ‘Qu’ is in Typ and to is in Force.
41
(26)
ForceP
Force’
TypP
Typ’
TP
kare-ga sore-o yatta
he-nom that-acc did
Force
to
comp
Typ
kadooka
Qu
In each of these examples, the embedded complementizer co-occurs with a Qumorpheme. This fact can be accounted for if the complementizer is in Force and the
Qu-morpheme is in Typ.
Note that in the examples (19), (23), and (25), evidence for a Force head comes
from embedded clauses. In the matrix clauses of these constructions, there is clearly
no overt Force head, and it is not clear to me whether or not there is a Force
head in a matrix clause, although if ForceP exists in an embedded clause, it would
not be surprising if it exists in matrix clauses. However, since all clauses have
some particular type (statement, interrogative, etc.), I think that it is reasonable to
assume that if a matrix clause has no overt Typ head, there is a null Typ head.
I use TypP instead of Rizzi’s IntP because there is evidence that the element that
occurs in this projection does not necessarily have to be an interrogative complementizer, and the label ‘IntP’ suggests a phrase that is only for interrogative elements.
Rather, the head of TypP is an element that is associated with typing a clause as
a statement, interrogative, exclamative, etc. This can be seen clearly in Sinhala,
in which the Qu-morpheme is in complementary distribution with various clausal
typing morphemes. For example, the morphemes tamay ‘certainty,’ lu ‘reportative,’
nee ‘tag-Q,’ and yae ‘dubitative’ (Sumangala 1992:131) are all in complementary
42
distribution with the interrogative Qu-morpheme [email protected], as shown in (27a-e).
(27) Sinhala:
(a) [email protected]@ [email protected]
gaalu [email protected] [email protected]?
Gunapaala tomorrow Galle go.A
Qu
‘Is Gunapala going to Galle tomorrow?’
(b) [email protected]@ [email protected]
gaalu [email protected] tamay.
Gunapaala tomorrow Galle go.A
certainty
‘It is for sure that Gunapala is going to Galle tomorrow.’
(c) [email protected]@ [email protected]
gaalu [email protected] lu.
Gunapaala tomorrow Galle go.A
reportative
‘They say that Gunapala is going to Galle tomorrow.’
(d) [email protected]@ [email protected]
gaalu [email protected] nee?
Gunapaala tomorrow Galle go.A
tag-QU
‘Gunapala is going to Galle tomorrow, isn’t that so?’
(e) [email protected]@ [email protected]
gaalu [email protected] yae!
Gunapaala tomorrow Galle go.A
dubitative
‘It is doubtful that Gunapala is going to Galle tomorrow!’ (Sumangala
1992:131)
If all of these morphemes occur in the same projection, then the label IntP is not
accurate, as these are morphemes associated with clausal typing. I therefore use the
label TypP to indicate the position of a clausal typing morpheme. For example, the
clausal typing morphemes in (27a-e) occur in Typ, as shown in (28). Since there is
no overt Force element, I leave out ForceP.
(28)
TypP
Typ’
TP
...
Typ
[email protected]/tamay/lu/nee/yae
43
One remaining issue is that of the position of a fronted wh-phrase in a whconstruction. Rizzi argues that in Italian, a fronted wh-argument is in [Spec, FocP]
and a fronted wh-adjunct is in [Spec, IntP]. In this work, I take the position that
wh-movement is to [Spec, TypP], not to [Spec, FocP]. Thus, a fronted wh-phrase in
a language such as English is in [Spec, TypP]. In some languages, a wh-phrase can
move to [Spec, FocP] if movement is driven by a Focus-feature. For example, in some
languages a wh-phrase can either remain in-situ or move to a clause-peripheral focusposition. I assume that this is not wh-movement. I discuss wh-phrasal movement
in chapters 6 through 8.
2.6 Island effects
The presence and absence of island effects involving wh-phrases plays an important
role in the analyses in the following chapters. In this section, I briefly discuss what
islands are and some reasons why they may occur.
An island is essentially a phrase out of which another phrase cannot be extracted;
it is like an island from which something cannot escape. The study of ‘islands’
originated with Ross (1967) and has led to an extensive amount of further research.
Some types of islands for extraction of wh-elements are shown in (29a-e).10
(29)
(a) Wh-island:
*What1 do you wonder [who bought t1 ]? (Huang 1982:456)
(b) Whether/If island
??What1 do you wonder [whether/if John saw t1 ]? (adapted from
Lasnik & Saito 1992:11)
10
Note that I have labeled (29b) as a whether/if island, although it is generally referred to as
a wh-island in the literature (e.g., see Lasnik & Saito (1992)). I think that the label wh-island is
inappropriate since the embedded clause is not a wh-construction, unlike (29a) which contains an
embedded wh-construction. I return to this issue in the next section.
44
(c) Complex NP/DP-island:
??What1 did you read [a report [that John bought t1 ]]? (Lasnik &
Saito 1992:12)
(d) Adjunct island:
*Which book1 did John go to class [after he read t1 ]? (Lasnik & Saito
1992:12)
(e) Subject island:
?*Who1 did [a story about t1 ] amuse you? (Lasnik & Saito 1992:42)
These examples contain phrases that function as islands from which a wh-phrase
cannot escape. The exact cause(s) of island effects has been the subject of much
research and much debate, and the issue is far from resolved. Below I discuss a few
of these accounts.
2.6.1 Government and Binding Theory accounts
The notion of Subjacency, which was originally proposed by Chomsky (1973), and
further developed in Chomsky (1977, 1981, 1986) and Lasnik and Saito (1984, 1992),
among others, is one of the clearest proposals that accounts for island effects. Hornstein et al. (2007:150) give the following definition of Subjacency, which is a revised
version of a definition given by Chomsky (1977:73).
(30) A cyclic rule cannot move a phrase from position Y to position X (or
conversly) in (i), where α and β are cyclic nodes (NP and S).
(i) . . . X . . . [α . . . [β . . . Y . . . ] . . . ] . . .
Lasnik & Saito (1992:11) describe Subjacency as “one general constraint prohibiting
one step of movement from crossing two bounding nodes, where the bounding nodes
are S and NP.”
Subjacency effects can be seen in (31a-e), which are identical to (29a-e) above
except for the more detailed phrase structure labels.
45
(31)
(a) Wh-island:
*What1 do [S
you wonder [who2 [S t2 bought t1 ]]]? (Huang
1982:456)
(b) Whether/If island
??What1 do [S
you wonder [whether/if [S
John saw t1 ]]]? (adapted
from Lasnik & Saito 1992:11)
(c) Complex NP/DP-island:
??What1 did you read [N P
a report [t1 that [S
John bought t1 ]]]?
(Lasnik & Saito 1992:12)
(d) Adjunct island:
*Which book1 did [S
John go to class [after t1 [S
he read t1 ]]]?
(Lasnik & Saito 1992:12)
(e) Subject island:
?*Who1 did [S
[N P
a story about t1 ] amuse you]? (Lasnik & Saito
1992:42)
In the wh-island and whether/if island constructions (31a-b), the wh-phrase crosses
two S bounding nodes. In the complex-NP/DP island construction (31c), the whphrase crosses S and an NP bounding node. In the adjunct island (31d), the whphrase crosses two S bounding nodes and in the subject island (31e), the wh-phrase
crosses an NP and an S bounding node.
Interestingly, the specific bounding nodes appear to be subject to cross-linguistic
variation. In (32), from Spanish, the wh-phrase que libro ‘what book’ moves over
two S nodes but the result is well-formed.
46
han
sabes [S ′ [por que]2 [S te
(32) [S ′ [Que libro]1 [S no
why
to-you have
what book
neg (you) know
regalado t1 t2 ?
given
‘[What book]1 don’t you know [why they have given to you t1 ]?’ (Lasnik &
Saito 1992:12)
The wh-phrase, however, only moves over one S’ node. Thus if S’ is a bounding
node instead of S, then the well-formedness of (32) is accounted for. Therefore,
in languages such as Spanish and Italian, the lack of wh-island effects has been
attributed to S’ (not S) and NP being bounding nodes (cf. Rizzi 1980, Sportiche
1981/82, Torrego 1984, Lasnik & Saito 1992, among others).
Although the Subjacency Condition can account for adjunct and subject island
effects, as in (31d-e), another constraint, Huang’s (1982) Condition on Extraction
Domain, given in (33), has traditionally been utilized for these effects.
(33) Condition on Extraction Domain
A phrase A may be extracted out of a domain B only if B is properly
governed. (Huang 1982:506)
The requirement that extraction can only occur if a phrase is properly governed
results in a “condition that allows extraction out of maximal projections that are
complements, but prohibits movement out of noncomplements (i.e., subjects and
adjuncts) (Lasnik & Saito 1992:12).” Subject clauses and adjunct clauses are clauses
that are not complements. Thus, they cannot be properly governed (cf. Chomsky
1981)11 and wh-phrases cannot be extracted out of them.
11
Government is a complex notion that has been defined in various ways. Below is a typical
definition of Goverment from (Lasnik & Saito 1992:14) that follows the works cited within the
definition.
(i) α properly governs β iff α governs β and
(a) α is a lexical category X0 (lexical government), or
47
2.6.2 Minimalist Program accounts
Subjacency relies on the notion that there are bounding nodes and the Condition
on Extraction Domain relies on the notion of Government. Bounding nodes and
Government, however, have been eliminated from the framework of the Minimalist
Program (cf. Chomsky 1995b), as being unnecessary. How to account for island
effects in the Minimalist Program is problematic. Uriagereka (1999a) sums up recent
attempts.
The Minimalist Program has no general account of islands. In part, this
is because the system is designed in such a streamlined fashion - and
with the assumption that computational mechanisms exist to meet the
requirements of external interfaces - that little room is left for the apparently ad hoc considerations involved in formulating island conditions.
(Uriagereka 1999a:404)
In the Minimalist Program, however, there are several theories that can account for
certain island effects.
Following Rizzi’s (1990) work on Relatived Minimality, Chomsky (1995b) proposes the Minimal Link Condition (MLC), given in (34).
(34) Minimal Link Condition (MLC)
K attracts α if there is no β, β closer to K than α, such that K attracts β.
(Chomsky 1995b:311)
The MLC requires attraction of the closest element of the relevant type. The MLC
straightforwardly accounts for the wh-island effect in (29a), repeated below.
(b) α is coindexed with β (antecedent goverment).
(ii) α governs β iff every maximal projection dominating α also dominates β and conversely.
(Aoun & Sportiche 1981)
48
(29)
(a) Wh-island
What1 do you wonder [who bought t1 ]? (Huang 1982:456)
In (29a) an attractor in the matrix CP12 attracts the closest wh-phrase, who, which
prevents what from moving to the matrix specifier of CP. Thus if what moves, there
is a violation of the MLC.
The MLC, however, does not appear to account for constructions in which a
whether/if clause functions as an island, as in (29b), repeated below.
(29)
(b) Whether/If island
??What1 do you wonder [whether/if John saw t1 ]? (adapted from
Lasnik & Saito 1992:11)
Although the embedded clause of this type of construction is generally considered
a wh-island (at least when it begins with whether ), I am not sure if the label whisland is appropriate. This is because whereas in a true wh-island construction such
as (29a), a wh-phrase is extracted out of an embedded wh-construction, in (29b) a
wh-phrase is extracted out of an embedded yes/no construction. The lexical item
if clearly is not a wh-phrase, although the facts are not so clear with respect to
whether (see discussion below). If if or whether are complementizers that occur in
C of the embedded clause, they should not block movement of what. In other words,
if if or whether are not specifiers, then there is nothing in (29b) to block movement
of what and result in an MLC violation. The ill-formedness of (29b) then requires
another explanation.
There is evidence suggesting that whether actually might be an XP element
in a specifier position, [Spec, CP]. There are certain instances in which if and
whether cannot occur in the same environment; whether can be followed by PRO
(cf. Chomsky 1981 for discussion of PRO), but if cannot, as shown in (35a-b).
12
In this section I use the term CP, as it is what is generally used in the literature discussed
here. However, as noted above, in this work, CP corresponds to what I refer to as TypP.
49
(35)
(a) He doesn’t know [whether PRO to go to the movies].
(b) *He doesn’t know [if PRO to go to the movies]. (Kayne 1991:665)
Kayne (1991), following work by Katz & Postal (1964) and Larson (1985), argues
that whether is a wh-phrase that occurs in [Spec, CP], and he proposes that if,
unlike whether, is a head that occurs in C. Kayne accounts for the fact that if
cannot directly precede PRO, as in (35b), within the framework of Government
and Binding Theory (Chomsky 1981). PRO must be ungoverned, in accord with
Chomsky (1981). When PRO is preceded by the complementizer if in C, it is
governed, therefore resulting in ungrammaticality. In (35a), PRO is preceded by
whether in [Spec, CP]. PRO is not lexically governed by whether, because whether
is not a head, nor is it antecedent governed because whether is not its antecedent.
In this way, whether is not an appropriate governor of PRO. Within the framework
of the Minimalist Program, Kayne’s proposal can no longer be upheld without some
modification, since the notion of Government no longer exists. Importantly though,
if and whether behave differently and this fact could be an indication that if is a
head and whether is a specifier. If this is the case, then the presence of whether
in the embedded clause of (29b) does result in an MLC violation. Still, the facts
surrounding (29b) with if remain the same; there is no MLC violation because if is
a head.
Interestingly, there is evidence that suggests that both whether and if really are
heads. There is, in my opinion, a fairly clear well-formedness distinction between a
wh-island construction such as (29a) and a whether/if -island construction such as
(29b), repeated below.
(29)
(a) Wh-island:
*What1 do you wonder [who bought t1 ]? (Huang 1982:456)
50
(b) Whether/If island
??What1 do you wonder [whether/if John saw t1 ]? (adapted from
Lasnik & Saito 1992:11)
Both (29a) and (29b) are ill-formed, but (29a) is much worse than (29b), regardless
of whether or not whether or if appears in the embedded clause of (29b). This
distinction supports the idea that a wh-island differs from a whether/if island, and
it might be an indication that Kayne’s proposal that whether is a specifier is not
correct; i.e., it is a head in the same way that if is; although, then the fact that
if cannot co-occur with PRO requires explanation. If this is the case, then there is
no MLC violation in (29b), and ill-formedness must have another cause. I return to
this issue below.
Overall, the MLC is useful for accounting for wh-island effects as in (29a), as
well as for other properties of languages, such as Superiority Effects (Chomsky 1973)
and intervention effects (see section 2.7), but it does not necessarily account for
whether/if island effects. Nor does it account for island effects involving complexDPs, adjunct-clauses, and subject-clauses, as in these cases, there is no clear element
in a specifier position that blocks movement of the wh-phrase and leads to an MLC
violation. Other proposals, however, can account for these latter types of island
effects.
Uriagereka’s (1999b) theory of Multiple Spell-Out (MSO) can account for Condition on Extraction Domain effects (adjunct and subject islands). This MSO theory
relies on the notion that subjects and adjuncts are formed separately from other elements in a clause. Essentially, subjects and adjuncts are complete syntactic objects
before they are Merged with other elements in a syntactic structure. Thus, nothing
can be extracted out of them.
The idea behind MSO is that a phrase consists of an unordered set of elements
that at a certain point in a derivation is linearized, along the lines of Kayne’s
(1994) Linear Correspondence Axiom (LCA), and sent to Spell-Out. According
51
to the LCA, linear order is established by asymmetric c-command. Asymmetric ccommand refers to a situation in which α c-commands β but β does not c-command
α. Below is a definition of the LCA that is utilized by Nunes & Uriagereka (2000:23).
(36) Linear Correspondence Axiom
A lexical item α precedes a lexical item β iff α asymmetrically c-commands
β.
The term ‘linearized’ refers to the elements of an unordered set becoming ordered.
Uriagereka (1999b:257) writes that “if a phrase marker {α, {L,K}} collapses through
Spell-Out, the rest is {α, <L, K>}.” The brackets {} represent an unordered set,
and the brackets <> signify that the elements of the set have become ordered. The
element α represents the label of the phrase. Uriagereka (1999b:251) describes the
process of linearization as follows.
A visual image to keep in mind is a mobile by Calder. The hanging
pieces relate in a fixed way, but are not linearly ordered with respect to
one another; one way to linearize the mobile (e.g., so as to measure it)
is to lay it on the ground. The substantive part of Kayne’s axiom does
for the complex linguistic object what the ground does for the mobile:
it tells us how to map the unordered set of terms into a sequence of PF
slots.
Thus an unordered set of lexical items becomes ordered at Spell-Out.
The MSO account forces a non-complement to be sent to Spell-Out before it
is Merged with another element. In (37), from Nunes & Uriagereka (2000:22), if
the DP subject were not sent to Spell-Out before it Merged with the V’, then the
lexical items contained within the DP would not asymmetrically c-command the
other elements of the clause; the c-commands man, and vice-versa, but neither
lexical item asymmetrically c-commands the other elements in the verb phrase.
52
(37)
VP
DP
the man
V’
V’
PP
V
remained
AP
after the fact
proud of her
If the elements of the DP are linearized and sent to Spell-Out, then they form a
single DP, and the DP label asymmetrically c-commands the other elements of the
clause, thereby deriving the appropriate word order.
The facts regarding the PP adjunct after the fact in (37) are virtually identical,
although with the added complexity that the PP be Merged to the left of its sister V’. The lexical items contained within the PP are linearized and the PP label
asymmetrically c-commands the lexical items in its V’ sister. Since the PP asymmetrically c-commands the V’, it must precede the V’ in the underlying structure,
or else there will be a violation of the LCA (if α asymmetrically c-commands β then
α must precede β). A diagram is shown in (38).
(38)
V’
PP
after the fact
V’
V
remained
AP
proud of her
To obtain the appropriate word order, the V’ must move to a position above the PP
after the fact. The MSO account thus requires that a non-complement be Merged
in a position to the left of whatever it Merges with.
53
The MSO requires that a non-complement be sent to Spell-Out before Merging
with another element, thus accounting for subject and adjunct island effects. A noncomplement that has been linearized is essentially treated as a lexical item. Nunes
& Uriagereka (2000:24) write that once an element is linearized, “the computational
system treats it as a lexical item.” Once a non-complement, such as a subject or
adjunct is linearized, then nothing can move out of it. For example, in (29e),
repeated below, the clausal subject a story about who is constructed via Merge and
sent to Spell-Out before it is Merged with amuse you. Therefore, extraction of the
wh-phrase who cannot occur.
(29)
(e) Subject island:
?*Who1 did [a story about t1 ] amuse you? (Lasnik & Saito 1992:42)
Under the MSO approach, a complement is not an island because it is able to
establish an asymmetric c-command relation without being linearized. For example,
in (37) above, the verb remained asymmetrically c-commands all of the elements of
the AP, and thus the AP complement does not need to be sent to Spell-Out until
after Merging with the verb.
One disadvantage of the MSO account, from my perspective, is that it relies
on the LCA. In order to maintain the simple notion that linear order follows from
asymmetric c-command, a great deal of extra complexity is required to account
for surface word order. In a language that (at least on the surface) is head-final,
there must be a tremendous amount of leftward movement of consituents, and even
in the head-initial English, there are instances in which leftward movement must
occur. For example, in (37) above, the PP must be base-generated to the left of
the element that it Merges with, and then its sister must move to a higher position.
Thus, I find the LCA, and MSO which relies on the LCA, to be problematic.
Johnson (2002) has another account of subject and adjunct island effects, which
is very similar to Uriagereka’s MSO account. However, it has the advantage of not
relying on the LCA. Johnson uses the term ‘renumeration’ to refer to the process of
54
putting an assembled constituent back into a numeration. Renumeration, like MSO,
relies on the notion that a non-complement is linearized separately from the element
that it is Merged with. Under the MSO account, a linearized phrase is sent to SpellOut, whereas under the renumeration account, a linearized phrase is placed back
into the numeration. Subjects and adjuncts are, according to Johnson (2002:13), a
natural class because “they are the phrases that are required to renumerate.”
Johnson relies on the following definition of an adjunct, which groups together
non-complements (adjuncts and subjects).13
(39) An adjunct is a phrase whose sister is also a phrase and whose mother is not
its projection. (Johnson 2002:1)
For example, in (40), from (Johnson 2002:2), the subject DP an advocate and the
PP before the discussion are adjuncts; they are each sisters to a phrase and their
mothers are different projections.14
(40)
vP
vP
DP
an advocate
PP
v P before the discussion
v
VP
V
spoke
PP
to Betsy
An adjunct is an island, as stated in (41).
13
14
This is essentially the view of adjuncts proposed by Chomsky (1995a).
If one were to represent this bare phrase structure with X’ notation, then a phrase that is a
sister to to a specifier would be a bar-level projection. For example, in (40), the v P sister of the
subject DP would be a v ’.
55
(41) When a phrase’s underlying position in a phrase marker is such that it is a
sister to another phrase but doesn’t project, it is an island for extraction
(Johnson 2002:3).
In (40), the subject DP an advocate and the PP before the discussion are islands
because they must be constructed in their own separate derivational workspaces and
linearized before they can be Merged with the other elements of the numeration.
Their islandhood results from their being built in a separate workspace and being
renumerated. Johnson argues that an adjunct is an island because if it were not
renumerated, then it would violate the following constraint, which requires whatever
merges with a head to be its complement.
(42) If an X0 merges with a YP, then YP must be its argument. (Johnson 2002:5)
(42) prevents a non-complement from Merging with a head and requires a complement to Merge with a head. A complement, at least structurally, is defined as a
sister to a head, a configuration that allows the head to assign a theta-role to the
complement.
Example (43) demonstrates what happens if the non-complement PP before the
discussion, from (40), is not renumerated. (43a) shows the underlying numeration.
(43b) shows the DP that is formed when the and discussion are selected from the
numeration and Merged. (43b) also shows what remains in the numeration after the
DP is formed. (43c) shows formation of the PP before the discussion. The derivation
falls apart at (43d). The PP before the discussion is Merged with the noun Betsy.
This operation violates (42), because the X0 noun Betsy Merges with a phrase, a
PP, that is not its complement.
(43)
(a) N = {an, advocate, v , spoke, to Betsy, before, the, discussion}
(b) N = {an, advocate, v , spoke, to Betsy, before}
56
DP
D
the
N
discussion
(c) N = {an, advocate, v , spoke, to, Betsy}
PP
P
before
DP
D
the
N
discussion
(d) N = {an, advocate, v , spoke, to}
NP
N
Betsy
PP
P
before
DP
D
the
N
discussion
Similarly, if the subject an advocate is not constructed separately from the rest
of the v P and renumerated, the derivation crashes. (44a) is the underlying numeration.15 In (44b), the PP is formed. In (44c) the v P is formed. In (44d), the VP
merges with v to form a v P. At (44e), the derivation crashes because the noun, an
X0 element, Merges with a phrase, in this case a v P that is not its complement,
thereby violating (42).
(44)
(a) N = {an, advocate, v , spoke, to, Betsy}
(b) N = {an, advocate, v , spoke}
15
I have left out the lexical items forming the adjunct PP for the sake of simplicity.
57
PP
P
to
N
Betsy
(c) N = {an, advocate, v }
VP
V
spoke
PP
P
to
N
Betsy
(d) N = {an, advocate}
vP
v
VP
V
spoke
PP
P
to
N
Betsy
(e) N = {an}
NP
N
advocate
vP
v
VP
V
spoke
PP
P
to
N
Betsy
When adjuncts are renumerated, then a derivation can proceed succesfully, as
I demonstrate below for example (40). The PP before the discussion is formed via
58
Merge, as shown in (45a-c) below, and then it is renumerated, as shown in (45d). It
is reinserted into the numeration as a linearized object that essentially corresponds
to a single lexical item.
(45)
(a) N = {an, advocate, v , spoke, to, Betsy, before, the, discussion}
(b) N = {an, advocate, v , spoke, to, Betsy, before}
DP
D
the
N
discussion
(c) N = {an, advocate, v , spoke, to, Betsy}
PP
P
before
DP
D
the
N
discussion
(d) N = {an, advocate, v , spoke, to, Betsy, [P P before the discussion]}
In a separate derivational space, the DP an advocate is formed, as shown in (46ab).16 Then it is renumerated, as shown in (46c).
(46)
(a) N = {an, advocate, v , spoke, to, Betsy, [P P before the discussion]}
(b) N = {v , spoke, to, Betsy, [P P before the discussion]}
DP
D
an
N
advocate
(c) N = {[DP an advocate], v , spoke, to, Betsy, [P P before the discussion]}
16
Although the PP appears as a linearized element in (46a), it is not necessarily the case that it
is formed and renumerated before the DP is formed.
59
The full v P then is formed as shown in (47a-f). The PP complement to Betsy is
formed and Merged with the verb and v . Then the renumerated subject and adjunct
Merge with the v P.
(47)
(a) N = {[DP an advocate], v , spoke, to, Betsy, [P P before the discussion]}
(b) N = {[DP an advocate], v , spoke, [P P before the discussion]}
PP
P
to
N
Betsy
(c) N = {[DP an advocate], v , [P P before the discussion]}
VP
V
spoke
PP
P
to
N
Betsy
(d) N = {[DP an advocate], [P P before the discussion]}
vP
v
VP
V
spoke
PP
P
to
N
Betsy
60
(e) N = {[P P before the discussion]}
vP
DP
an advocate
vP
v
VP
V
spoke
PP
P
to
N
Betsy
(f) N = { }
vP
vP
DP
an advocate
PP
v P before the discussion
v
VP
V
spoke
PP
to Betsy
In this manner, renumeration requires a non-complement to be linearized and
reinserted into a numeration. As with the MSO account, a linearized element can be
manipulated as a whole, but the lexical items contained within it are not accessible
to further operations. Therefore, a wh-phrase contained within a renumerated noncomplement cannot move from within it, and subject and adjunct island effects are
accounted for.
Certain complex-DP island effects create a problem for the MSO and renumeration accounts. In (29c), repeated below, if the subordinate clause is an adjunct, then
61
these accounts can explain the island effect. The subordinate clause must be sent
to Spell-Out or renumerated. However, if the subordinate clause is a complement,
then it should not function as an island, and extraction of a wh-phrase should be
possible.
(29)
(c) Complex NP/DP-island:
??What1 did you read [a report [that John bought t1 ]]? (Lasnik &
Saito 1992:12)
There is evidence suggesting that a subordinate clause in certain complex-DPs is a
complement. For example, the subordinate clause within the complex-DP of (29c)
is quite similar to the object clause of the verb report in (48) which clearly is a
complement.
(48) He reported that John bought a book.
Similarly, the subordinate clause in (49a) corresponds to the complement of the verb
claim in (49b).
(49)
(a) *What1 do you believe [the claim [that John bought t1 ]]? (Lasnik &
Saito 1992:22)
(b) He claimed that John bought the book.
Nichols (2003) provides an explanation for these complex-DP effects by arguing
that the subordinate clause in this type of complex-DP really is an adjunct, despite
appearances. Nichols (2003:156) writes the following.
. . . the standard argument that (deverbal) nouns may (optionally) assign
theta roles is in actuality based on the stipulation that the complement
theta role is carried over from the verb. There is no direct evidence for
this position in the case of subordinate clauses, and, moreover . . . there
are other ways in which instances of apparent theta assignment by nouns
may be explained.
62
Nichols points out that, in constructions such as (29c), the evidence that the embedded clause is a complement to the noun comes from the resemblence between
the noun and its verbal form. However, she notes that the nouns story, idea and
fact behave in the same way as claim, even though they do not have verbal forms.
For example, (50a) demonstrates that these nouns which do not have a verbal form
can be followed by a subordinate clause. (50b) shows that extraction of a wh-phrase
from this clause results in a complex-DP island effect.
(50)
(a) I believe [the claim/report/story/idea/fact [that John bought the
book]]?
(b) ??What1 do you believe [the claim/report/story/idea/fact [that John
bought t1 ]]? (Lasnik & Saito 1992:22)
Nichols claims that because these nouns do not have transitive verbal forms, it is
not clear that their subordinate clauses, as in (50a-b) are complements. Without
going into more details (see Nichols (2003) for further arguments that this type of
subordinate clause is an adjunct), I will assume that Nichols is correct and that this
type of subordinate clause is an adjunct.
With the assumption that the subordinate clause is an adjunct, under the renumeration account, the complex-DP island effect in (29c) is accounted for as shown in
(51a-e). Beginning with the numeration in (51a), the CP adjunct clause that John
bought what is assembled and renumerated as shown in (51b-c). The DP a report is
assembled and the renumerated subordinate clause is Merged with it, as shown in
(51d-e).
(51)
(a) N = {a, report, that, John, bought, what}
(b) N = {a, report}
CP
that John bought what
63
(c) N = {a, report, [CP that John bought what]}
(d) N = {[CP that John bought what]}
DP
a report
(e) N = { }
DP
DP
CP
a report
that John bought what
Since the subordinate clause is renumerated, nothing can be extracted from it,
thereby accounting for why the wh-phrase cannot move out of the complex-DP. A
derivation in which there is no renumeration is blocked because then the CP would
Merge with a head, the noun report. As shown in (52), at the point in which the
noun report is Merged with the CP, an X0 element is a sister to an XP that is not
its complement, thus violating (42).
(52)
NP
N
report
CP
that John bought what
Lastly, I note that whether/if island effects as in (29b), repeated below, which
as discussed above pose a problem for the MLC account, also pose a problem for a
renumeration/MSO account.
(29)
(b) Whether/If island
??What1 do you wonder [whether/if John saw t1 ]? (adapted from
Lasnik & Saito 1992:11)
64
If the whether/if clause is a complement of wonder, then it should not be linearized
and sent to Spell-Out or renumerated. On the other hand, if it is an adjunct then
the ill-formedness of this construction is accounted for; the whether/if clause is
renumerated. At this point, I do not know of any clear evidence that a whether/if
clause is an adjunct, and thus the cause of this type of island effect requires further
examination.
In summary, the Minimalist Program requires an eclectic approach to island
effects because ‘an island’ is not a uniform element. An island is a phrase from which
an element cannot be extracted, but the cause of an island effect differs depending
on the type of island. The MLC accounts nicely for wh-island effects. MSO and
renumeration account for subject and adjunct island effects. They also can apply to
complex-DP island effects, assuming that a subordinate clause within a complex-DP
is an adjunct. A whether/if island effect, however, remains problematic; I briefly
return to this issue in chapter 6. In the remainder of this dissertation, I will rely on
the MLC and renumeration to account for relevant island effects. I use renumeration
instead of MSO because of its advantage, as noted above, of not relying on the LCA.
2.6.3 The absence of island-effects
It is well-known that there are languages in which one can observe a lack of one or
more of the various types of island effects given in (29a-e) above. As pointed out
by Huang (1982), this lack of island effects is generally found in instances in which
a wh-phrase remains in-situ. (53a-e) are well-formed, despite corresponding to the
ill-formed (29a-e) above.
65
(53) Mandarin:
(a) No wh-island effect:
Ni xiang-zhidao [shei mai-le
sheme]?
you wonder
who buy-asp what
‘What1 do you wonder [who bought t1 ]? (Huang 1982:479)
(b) No whether/if island effect
Ni xiang-xhidao [ta xi-bu-xihuan
shei]?
you wonder
he whether-or-not like who
‘Who1 do you wonder [whether he likes t1 ]?’ (Lasnik & Saito 1992:32)
(c) No complex-DP island effect:
Ni xihuan [DP shei xie
de
shu]?
you like
who write gen book
‘Who1 do you like [the book t1 wrote]?’
(d) No adjunct island effect:
Ta [T ypP yinwei ni shuo shenme hua] hen shengqi?
he
because you say what
word very angry
‘What1 was he angry [because you said t1 words]?’(Aoun & Li 1993:203)
(e) No subject island:
[DP shei xie
de
shu] zui
youqu?
who wrote gen book most interesting
‘Who1 are [books that t1 wrote] most interesting?’ (Lasnik & Saito
1992:122)
These types of instances in which a wh-phrase is contained within a clause that
functions as an island in certain languages, but not in others, will be of importance
in the following analysis, in which I examine the presence and absence of island
effects to determine if movement of interrogative features has occurred.
66
2.7 Intervention effects
The island effects discussed in the previous section occur when a wh-phrase cannot
move out of a particular type of clause. Another type of locality effect in which
movement of a wh-element is blocked is an ‘intervention effect.’ This phenomenon
also plays an important role in my analyses.
I use the term ‘intervention effect’ in accord with Beck (1996) and Beck & Kim
(1997) to refer to ill-formedness that results when a particular scope bearing phrase,
which is a quantificational and/or focus element, c-commands a wh-phrase. Pesetsky
(2000:67) summarizes intervention effects as follows.
(54) Intervention effect (universal characterization)
A semantic restriction on a quantifier (including wh) may not be separated
from that quantifier by a scope-bearing element.
Beck (1996:38) proposes the generalization that quantifiers block LF movement
based on German constructions such as those in (55a-b). In (55a), the negative
element niemanden ‘nobody,’ c-commands the wh-phrase wo ‘where.’ When the
wh-phrase appears in a position preceding niemanden ‘nobody’ in (55b), the result
is well-formed.
(55) German:
(a) ??Wer hat niemanden wo
angetroffen?
who
has nobody
where met
‘Who didn’t meet anybody where?’
(b) Wer hat wo
niemandem angetroffen?
who has where nobody
met
‘Who didn’t meet anybody where?’ (Beck 1996:6)
Below are intervention effects from Japanese. In (56a), the Negative Polarity Item
sika ‘only’ c-commands the wh-phrase nani-o ‘what-acc.’ In (56b), the wh-phrase
precedes sika ‘only,’ and the result is well-formed.
67
(56) Japanese:
(a) ?*Taroo-sika nani-o
yoma-nai no?
Taroo-only
what-acc read-NEG Qu
‘What did only Taro read?’ (Tanaka 1997:159)
(b) Nani-o1 Taroo-sika t1 yoma-nai no?
what-acc Taroo-only
read-NEG Qu
‘What did only Taro read?’ (Tanaka 1997:162)
Interveners were originally classified as quantificational elements, including negation, by Beck (1996) for German. However, Beck & Kim (1997:370-371) point out
that some focus phrases, which are not obviously quantificational, such as man
‘only’ and to ‘also’ in Korean cause intervention effects, while some quantificational phrases such as taepupun ‘most,’ hangsang ‘always,’ and chachu ‘often’ do
not cause intervention effects. Tomioka (2007:1574) similarly points out that the
Japanese subete-no/zenbu-no-NP ‘all (the) NP’ are quantificational, yet they do not
cause intervention effects, whereas some DPs that are not quantificational, such as
the Japanese NP-mo/NP-to ‘NP also’ are interveners. Kim (2002, 2006) deals with
this apparently nonuniform class of interveners by taking the position that interveners in Korean are focused phrases (a class that crucially includes Negative Polarity
Items). This view is further elaborated on by Beck (2006). Tomioka (2007) refers to
the class of interveners as ‘Anti-Topic Items’ since they consist of elements that he
argues cannot be topicalized in Japanese and Korean. As Tomioka notes, this view
is compatible with the notion that intervenors are focused elements. For the sake of
simplicity, in this work I generally refer to interveners as scope bearing elements.
Karimi & Taleghani (2007) argue that the Minimal Link Condition (MLC) (see
(34) above) accounts for intervention effects. This makes sense if either a wh-phrase
undergoes covert movement over an intervening scope bearing element, or if an
element associated with the wh-phrase undergoes movement of some sort, and this
movement is blocked by the intervening element; for example, supposing K attracts
68
an element associated with a wh-phrase α, but there is a closer scope bearing element
β to K that blocks this movement.
Note that under this proposal, there is a parallelism between certain wh-island
effects and intervention effects in that they both result from violations of the MLC.
When there is a wh-island, a higher wh-phrase blocks a lower wh-phrase from obtaining scope. When there is an intervention effect, a scope bearing element blocks
a wh-phrase from obtaining scope. In each case, a scope relation is blocked by an
intervening element.
In order to investigate movement of interrogative features, I examine intervention
effects later in this work. Of importance is the fact that intervention effects appear
to indicate that movement has occurred.
2.8 Head movement
Head movement plays an important role in the analyses presented in this work
in that I claim that movement of X0 category elements can occur in interrogative
constructions. Head movement, as discussed in this work, refers to movement of an
element from the head position of one maximal projection to the head position of
another, as shown in (57).
(57)
XP
X’
X
α1
YP
Y’
Y
t1
ZP
69
The nature of head movement within the Minimalist Program is controversial.
Under the view that a moved head adjoins to another head, there is a violation of
the Extension Condition (Chomsky 1995b), which is the notion that movement, via
Merge, must be to the top of a tree (thereby extending it). For example, if α is a
head that adjoins to another head X, as shown in (58), movement of alpha does not
extend the tree. Rather, movement is to a position within the already constructed
spine of the tree.
(58)
XP
X
α1
YP
X
Y’
t1
ZP
Furthermore, if a head adjoins to another head, as in (58), then the moved head
does not c-command its trace (cf. Matsushansky 2006).
There have been proposals that deal with these problems. Notably, Chomsky
(1995b) proposes that head movement occurs at PF. Harley (2004) proposes that
head movement results from the passing of phonological features of a head up a tree
(notably, before PF), without actual movement of the head. Since these proposals do not involve movement of a head, they get around the problematic issues of
the Extension Condition and c-command. However, both Chomsky’s and Harley’s
proposals rely on movement of phonological features alone, and thus predict that
head movement should not influence the semantics of a construction. Matsushansky
(2006), on the other hand, proposes that a head adjoins to a maximal projection
and that there is a morphological operation whereby the head of the maximal projection and the moved head merge. This latter analysis involves actual movement
of a head, and thus does not necessarily rule out semantic effects of head movement
(see Matsushansky 2006 for discussion of this issue). In the following chapters, the
70
head movement that I propose has semantic effects and thus this latter analysis is
most compatible with my work.
In this work, I represent head movement as shown in (57) above, without examining whether or not head movement involves adjunction, etc. There is much that
can be examined with respect to head movement and the proposals presented in the
following chapters. However, I leave aside these important issues (as my focus is
on other matters) and simply assume that head movement of some form can exist
whereby either a head or a feature contained within a head moves to another head
position.
2.9 Conclusion
In this chapter, I have discussed the basic assumptions about language that will be
of importance in the following chapters. I explained the operations that the faculty
of language uses to construct lexical items and phrases. I discussed the importance
of the clausal-periphery for interrogative constructions and clausal typing, and I
described the basic phrase structure that I assume in this work. I then discussed two
types of phenomena; island- and intervention-effects, which I utilize as diagnostics
of movement in the following chapters. Lastly, I noted some problematic issues
concerning head movement. I next turn to analyses of interrogative constructions.
71
CHAPTER 3
Qu-features and Qu-morphemes
3.1 Introduction
I next turn to a discussion of Qu-features and the Qu-morphemes that house them.
The existence of a Qu (question) element in interrogative constructions has been
argued for by Katz & Postal (1964), Aoun & Li (1993), Chomsky (1995b), and
Denham (2000), among others. In this chapter, I support these views by presenting
evidence that there is a Qu-element, specifically, a Qu-morpheme that contains a
Qu-feature in all yes/no and wh-constructions. I argue that this Qu-feature is crucial
to the formation of yes/no and wh-constructions because it is responsible for typing
a clause as an interrogative.
The organization of this chapter is as follows. Section 3.2 discusses evidence
that there is a Qu-feature that is present in yes/no and wh-constructions. Section
3.3 shows that a single language can have multiple Qu-morphemes which differ in
certain semantic and syntactic ways, but which still serve the same purpose of typing
a clause as an interrogative. Section 3.4 is the conclusion.
3.2 Evidence for a Qu-feature
Evidence for the existence of a Qu-feature is seen most clearly in languages that have
overt Qu-morphemes. For example, in some languages, such as Japanese, Korean,
Yavapai, and Luiseño, a matrix yes/no construction is formed simply by adding
a Qu-morpheme to a corresponding non-interrogative construction. Furthermore,
an identical Qu-morpheme in the same position may occur in wh-constructions. I
take the appearance of an identical Qu-morpheme in yes/no and wh-constructions
72
as evidence that a Qu-feature contained within a Qu-morpheme is responsible for
typing a clause as an interrogative.
The Japanese Qu-morpheme ka is one of several Qu-morphemes in Japanese that
occurs in both yes/no and wh-constructions. For example, the statement in (1a)
becomes a yes/no construction when the Qu-morpheme ka is added to clause-final
position, as shown in (1b).
(1)
Japanese:
(a) Kare-wa ie-ni
kaerimashita.
He-top home-dat went
‘He went home.’
(b) Kare-wa ie-ni
kaerimashita ka?
He-top home-dat went
Qu
‘Did he go home?’
This same Qu-morpheme ka occurs in wh-constructions, as can be seen in (2). In
this example, the wh-phrase doko ‘where’ occurs together with ka ‘Qu’ to form a
wh-construction. Note that ka occurs in sentence-final position, just as in the yes/no
construction (1b) above.
(2)
Japanese:
Kare-wa doko-ni
ikimashita ka?
He-top where-dat went
Qu
‘Where did he go?’
The Korean Qu-morpheme ni also occurs in both yes/no and wh-constructions,
and like the Japanese ka, it occurs in clause-final position. When the Qu-morpheme
ni replaces the declarative morpheme ta found in (3a), a yes/no construction results,
as shown in (3b).
73
(3)
Korean:
(a) Mary ka
o
ass
ta.
Mary nom come past decl
‘Mary has come.’
(b) Mary ka
o
ass
ni?
Mary nom come past Qu
‘Has Mary come?’ (Jang 1999:849)
When this Qu-morpheme ni co-occurs with a wh-phrase, such as nwukwu ‘who,’ as
in (4), a wh-construction results, as shown below.
(4)
Korean:
Nwukwu i ni?
who
be Qu
‘Who is it?’ (Jang 1999:851)
Yavapai1 is another language that contains a Qu-morpheme, in this case the mor-
pheme e:, which occurs in clause-final position of both yes/no and wh-constructions.
Replacing the incompletive aspect morpheme km in (5a) with the Qu-morpheme e:
in (5b) results in a yes/no construction.
(5)
Yavapai:
(a) Ma:-c m-cky at-km.
2-pl
2-chop-incompletive
‘You are/were chopping it.’
1
This is a Yuman language spoken in Arizona.
74
(b) Ma:-c m-cky at-e:?
2-pl
2-chop-Qu
‘Are/were you chopping it?’ (Kendall 1976:107)
Similarly, the Qu-morpheme e: occurs in clause-final position in wh-constructions.
For example, in (6), e: occurs in a clause along with the wh-phrase kavyu ‘why’ to
form a wh-question.
(6)
Yavapai:
Kavyu-m
Tivo:y m-qaly-e:?
why-AllocentricReferent onion 2p-dislike-Qu
‘Why don’t you like onions? (Kendall 1976:109)
The Luiseño2 Qu-morpheme /s u occurs after the initial lexical item of a clause
in both yes/no and wh-constructions. Compare (7a) and its corresponding yes/no
counterpart in (7b).3
(7)
Luiseño
(a) Xwaan ‘aamoq.
Juan
hunting
‘Juan is hunting.’
(b) Xwaan /su ‘aamoq?
Juan
Qu hunting
‘Is Juan hunting?’ (Hyde 1971:20)
In (8), the Qu-morpheme occurs after the wh-phrase hax ‘who,’ thereby forming a
wh-construction.
2
3
This is a Uto-Aztecan language spoken in Southern California.
The examples given in Hyde (1971) do not contain word-for-word glosses, so I have provided
glosses based on other examples from the same source with identical lexical items.
75
(8)
Luiseño
Hax /su ‘owo‘aq?
who Qu working
‘Who is working?’ (Hyde 1971:23)
The above examples show that interrogatives in some languages are formed
with a Qu-morpheme that shows up in identical positions in both yes/no and whconstructions. In Japanese, Korean, and Yavapai, a Qu-morpheme occurs in clausefinal position. In Luiseño, a Qu-morpheme occurs in clause-second position. I take
these facts to be evidence that a Qu-feature, contained within a Qu-morpheme, occurs in yes/no and wh-constructions in these, and possibly all, languages. In those
languages that lack an overt Qu-morpheme, a null Qu-morpheme is present and is
responsible for an interrogative interpretation.4
3.3 Multiple Qu-morphemes in a single language
A Qu-morpheme is not identical to a Qu-feature. Rather, it is a morpheme that
contains a Qu-feature. If a Qu-morpheme were simply the overt representation of a
Qu-feature, then a single language would probably not be able to have more than
one Qu-morpheme, or if it had more than one, all of the Qu-morphemes would
behave identically. This is not the case, as a single language may have multiple
4
This view is in agreement with Katz & Postal (1964), Aoun & Li (1993), and Denham
(2000), who argue that English, Mandarin, and Babine Witsuwit’en, respectively, contain null
Qu-morphemes. Cheng (1997) argues that a language either contains a wh-particle, which is essentially equivalent to a Qu-morpheme, or it has overt wh-movement, but it does not have both.
Therefore, a wh-in-situ language has a wh-particle and a wh-movement language does not. My
view, contra Cheng, is that a Qu-morpheme is always present in a yes/no or wh-construction,
regardless of whether or not a language forms wh-constructions via a wh-in-situ strategy. I base
this assumption on the existence of languages that have both overt Qu-morphemes and wh-phrasal
movement. See chapter 6 for further discussion.
76
Qu-morphemes that differ in their semantics and in their syntactic behavior, as I
show in the following discussion. These differences are a result of Qu-morphemes
containing other features in addition to a Qu-feature.
Thai is a language with several Qu-morphemes that differ semantically and syntactically. I briefly illustrate a few of these differences.5 The Qu-morpheme máy “is
used to ask about information that belongs exclusively to the addressee (Iwasaki &
Ingkaphirom 2005:280)” including questions about emotions, as shown in (9).
(9)
Thai:
Dii-cay máy?
glad
Qu
‘Are/were you glad?’ (Iwasaki & Ingkaphirom 2005:280)
According to Iwasaki & Ingkaphirom (2005), this Qu-morpheme cannot be used
with a nominal predicate. However, the Qu-morpheme rú-plàaw can be used with
a nominal predicate, as shown in (10), in which the Qu-morpheme co-occurs with
the nominal predicate fEEn khun-àphı́châat ‘girlfriend tl-Apichat.’
(10) Thai:
Kháw pen fEEn
khun-àphı́châat r ú-plàaw?
Qu
3p
is
girlfriend tl-Apichat
‘Is she Mr. Apichat’s girlfriend?’ (Iwasaki & Ingkaphirom 2005:283)
This morpheme rú-plàaw is used when the speaker considers information “to be
more public (Iwasaki & Ingkaphirom 2005:283).” Note that this is contrary to máy,
shown in (9) above, which is used when the speaker “considers the information
to belong solely to the addressee (Iwasaki & Ingkaphirom 2005:283).” Thai also
contains the Qu-morpheme rú-yaN, which can be used to indicate the perfective
aspect.
5
See Iwasaki & Ingkaphirom (2005) for a more extensive discussion of the differences among
these morphemes.
77
(11) Thai:
Kin r ú-yaN?
eat Qu
‘Have you eaten yet?’ (Iwasaki & Ingkaphirom 2005:284)
The Qu-morpheme lˇ@ is used when the speaker “. . . has intense curiousity regarding
the information he has at hand (Iwasaki & Ingkaphirom 2005:285).”
(12) Thai:
Nǐi kháN lĚEk lˇ
@?
this time first Qu
‘Is this your first time (experience of an earthquake)?’ (Iwasaki &
Ingkaphirom 2005:285)
All of these Qu-morphemes are used to form yes/no constructions, yet they differ in
their semantics and in the types of constructions they can occur in. These differences
can be accounted for if each morpheme has its own distinct set of features, in addition
to a Qu-feature.
Korean contains several Qu-morphemes that differ in terms of politeness. Examples (13a-c) all correspond to Is it raining? in English.
(13) Korean:
(a) Pi ka
o-ni?
rain nom come-Quplain
(b) Pi ka
o-na?
rain nom come-Quf amiliar
78
(c) Pi ka
o-nya?
rain nom come-Quneutral (Sohn 1999:269-270)
Although the meanings of examples (13a-c) are essentially the same, they each differ
in formality to the addressee. The Qu-morphemes ni, na, and nya show what Sohn
(1999) refers to as plain, familiar, and neutral levels of formality. These differences
can be accounted for if each of these Qu-morphemes has a different feature for
politeness. More specifically, ni, na, and nya have different features that give them
the qualities that Sohn (1999) refers to as plain, familiar, and neutral.
Japanese also has Qu-morphemes that differ in their syntactic behavior and in
their semantics.
Japanese Qu-morphemes differ with respect to whether they can occur in only a
yes/no construction, only a wh-construction, or both a yes/no and wh-construction.
The Qu-morphemes ka and no can both occur in either yes/no or wh-constructions.
Examples (14a-b) show these two morphemes in yes/no constructions.
(14) Japanese:
(a) Gohan-o tabeta ka?
food-acc ate
Qu
‘Did you eat a meal?’
(b) Gohan-o tabeta no?
food-acc ate
Qu
‘Did you eat a meal?’
Examples (15a-b) show these morphemes in wh-constructions.
(15) Japanese:
(a) Nani-o
tabeta ka?
what-acc ate
Qu
‘What did you eat?’
79
(b) Nani-o
tabeta no?
what-acc ate
Qu
‘What did you eat?’
In contrast, the Qu-morpheme kai can only occur in yes/no constructions. The
yes/no construction ending with kai in (16a) is fine, but the wh-construction ending
in kai in (16b) is ill-formed.
(16) Japanese:
(a) Kare-ga ie-ni
kaetta kai?
He-nom home-dat went Qu
‘Did he go home?’
(a) *Dare-ga ie-ni
kaetta kai?
who-nom home-dat went Qu
Intended: ‘Who went home?’ (Junko Ginsburg, p.c.)
The Qu-morpheme ndai can only occur in wh-constructions. This morpheme is fine
at the end of the wh-construction in (17a), but is ill-formed at the end of the yes/no
construction in (17b).
(17) Japanese:
(a) Dare-ga kuru ndai?
who-nom come Qu
‘Who will come?’ (Hagstrom 1998:16)
(b) *Kare-ga kuru ndai?
he-nom come Qu
Intended: ‘Is he coming?’ (Junko Ginsburg, p.c.)
Japanese Qu-morphemes also differ with respect to whether or not they can
occur in a matrix or embedded clause. The above examples in (14-17) show the
morphemes ka, no, kai, and ndai all in matrix clauses. Of these Qu-morphemes,
80
only ka can also occur in an embedded clause. Example (18a) below is fine with ka
at the edge of the embedded clause. However, (18b-d) with no, kai, and ndai at the
edge of embedded clauses are all ill-formed.
(18) Japanese:
(a) Watashi-wa [kare-ga nani-o
tabeta ka] wakaranai.
I-top
he-nom what-acc ate
Qu don’t know
‘I don’t know what he ate.’
(b) *Watashi-wa [kare-ga nani-o
tabeta no] wakaranai.
I-top
he-nom what-acc ate
Qu don’t know
Intended: ‘I don’t know what he ate.’
(c) *Watashi-wa [kare-ga nani-o
tabeta kai] wakaranai.
I-top
he-nom what-acc ate
Qu don’t know
Intended: ‘I don’t know what he ate.’
(d) *Watashi-wa [dare-ga kuru ndai] wakaranai.
I-top
who-nom come Qu
don’t know
Intended: ‘I don’t know what he ate.’
Japanese also contains a Qu-morpheme kadooka that can only occur in an embedded
yes/no construction. (19a) is fine with kadooka at the edge of the embedded yes/no
construction, but (19b) is ill-formed because the embedded clause contains a whphrase, and (19c) is ill-formed because kadooka cannot occur in a matrix clause.
(19) Japanese:
(a) Watashi-wa [kare-ga sore-o
tabeta kadooka] wakaranai.
I-top
he-nom that-acc ate
Qu
don’t know
‘I don’t know if he ate that.’
(b) *Watashi-wa [kare-ga nani-o
tabeta kadooka] wakaranai.
I-top
he-nom what-acc ate
Qu
don’t know
Intended: ‘I don’t know what he ate.’
81
(c) *Kare-ga sore-o
tabeta kadooka?
he-nom that-acc ate
Qu
Intended: ‘Did he eat that?’ (Junko Ginsburg, p.c.)
These Japanese Qu-morphemes also differ in terms of politeness, with ka being
more polite then no, and no being more polite then kai or ndai. The embedded
clause Qu-morpheme kadooka appears not to be specified for politeness. I base this
assumption on native speaker intuition that kadooka is neither polite nor impolite
(Junko Ginsburg, p.c.).
The differences in these Qu-morphemes are summarized in (20). Ka and no
occur in both yes/no and wh-constructions. Kai and kadooka occur in only yes/no
constructions and ndai only occurs in wh-constructions. Only ka can occur in both
an embedded and matrix clause. All of the other Qu-morphemes except for kadooka
must occur in matrix clauses. Kadooka can only occur in an embedded clause.
Furthermore, ka is more polite than no, and no is more polite then kai or ndai.
In the chart, these three levels or politeness are referred to as polite, plain, and
informal. I have not specified kadooka for politeness.
(20)
Morpheme
yes/no
wh
ka
X
X
X
no
X
X
X
kai
X
ndai
kadooka
X
X
matrix embedded
X
polite plain
informal
X
X
X
X
X
X
X
I assume that each Japanese Qu-morpheme differs in its syntactic behavior and
semantics because it contains a unique set of features. Despite these differences,
these Qu-morphemes all contain a Qu-feature.
82
The Qu-morpheme data from Thai, Korean, and Japanese show that a single
language may have more than a single Qu-morpheme, and that these Qu-morphemes
are not identical to each other, since they show up in different environments. These
facts are evidence that a Qu-morpheme contains more than just a Qu-feature. Each
Qu-morpheme has its own unique set of features that differs in some way from the
features that make up other Qu-morphemes.
3.4 Conclusion
If the claims made here are correct, then the languages discussed in this chapter
have Qu-morphemes; i.e., a morpheme that contains a Qu-feature, as well as other
features. A yes/no or wh-construction requires the presence of a Qu-feature, and
a Qu-feature can only enter a derivation via attachment to a Qu-morpheme. In
the following chapters, I will argue that the Qu-feature, contained within either an
overt or covert Qu-morpheme, plays an important role in the formation of yes/no
and wh-constructions.
83
CHAPTER 4
Yes/no constructions
4.1 Introduction
In this chapter, I examine the formation of yes/no constructions. I argue that crosslinguistic variation in yes/no constructions results primarily from properties of Qumorphemes. I begin by demonstrating that there are various types of Qu-morphemes
which differ in terms of whether or not they are pronounced, and whether or not
they are free morphemes. Next, I examine why some Qu-morphemes appear in the
clause periphery and others in a TP-internal position. I argue that a Qu-morpheme
appears in a TP-internal position if it has a Focus-feature.
This chapter is organized as follows. In section 4.2, I propose that certain features
determine whether a Qu-morpheme is pronounced, and whether or not it is an affix.
Section 4.3 discusses languages in which a Qu-morpheme must appear in the clauseperiphery. I argue that these Qu-morphemes are Merged directly in Typ. Section
4.4 discusses languages in which a Qu-morpheme can either appear TP-internally
or at the clause-periphery. Section 4.5 is the conclusion.
4.2 Types of Qu-morphemes
In this section, I propose that certain features contained within Qu-morphemes
account for cross-linguistic variation in yes/no construction formation. Specifically,
I account for why languages form yes/no constructions in the following ways: 1)
with an overt Qu-morpheme that is an independent lexical item, 2) with an overt
Qu-morpheme that affixes onto another lexical item, 3) with movement of a lexical
item but no overt Qu-morpheme, and 4) with only intonation.
84
There are several types of Qu-morphemes.
Some languages use overt Qu-
morphemes in yes/no constructions and others do not, and a single language may
even allow yes/no constructions to be formed either with or without an overt Qumorpheme. As discussed in chapter 3, I assume that a Qu-morpheme is present in
all yes/no constructions. When it is not pronounced, it is null. Some languages then
have an overt Qu-morpheme, some have a null Qu-morpheme, and others have both
types of Qu-morpheme, one that is overt and another that is null. A Qu-morpheme
may also be an affix or an independent lexical item. I argue that the source of
these differences in Qu-morphemes lies in the [±] values of [OVERT] and [AFFIX]
features.
The table in (1) lists four possible types of Qu-morphemes in terms of the features
that they contain.
(1)
FEATURES
(a)
[+OVERT], [+AFFIX]
(b)
[+OVERT], [-AFFIX]
(c)
[-OVERT], [+AFFIX]
(d)
[-OVERT], [-AFFIX]
(1a) describes a Qu-morpheme that is an overtly pronounced affix, (1b) an
overt free morpheme, (1c) a null affix, and (1d) a Qu-morpheme that is neither
overt nor an affix, but an intonation pattern.
In summary, a Qu-feature surfaces inside of a Qu-morpheme that is overt or
null, and that is an affix, an independent lexical item, or a particular intonation
pattern. These properties of Qu-morphemes result from the features they have. In
the following subsections, I explain how these Qu-morpheme features account for
cross-linguistic variation in the forms of yes/no constructions.
85
4.2.1 Overt Qu-morphemes
There are several different types of overt Qu-morphemes. Tohono O’odham and
Haida have overt Qu-morphemes that are affixes, whereas Japanese and English
have overt Qu-morphemes that are free morphemes. Example (2) shows these Qumorphemes in terms of the features that they have.
(2)
LANGUAGE
FEATURES
Tohono O’odham, Haida [+OVERT], [+AFFIX]
Japanese, English
[+OVERT], [-AFFIX]
I discuss these Qu-morphemes below.
Yes/no constructions in Tohono O’odham1 are formed with an overt Qumorpheme that is an affix. For example, in (3a), the subject mistol ‘cat’ occurs
in clause-initial position followed by what is generally considered to be an auxiliary
(AUX) ’o (cf. Zepeda 1983, Hale 1983, Hale & Selkirk 1987, Payne 1987, Hale
2002). In the yes/no construction (3b), the subject no longer occurs in sentence
initial position. Instead, this position is filled with the Qu-morpheme n followed by
the AUX o.2
(3)
Tohono O’odham:
(a) Mistol ’o
ko:s..
cat
AUX sleeping
‘The cat is/was sleeping.’
1
This is a Uto-Aztecan language spoken in southern Arizona and northern Mexico. I use the
term Tohono O’odham in accord with Hale (2002) and the conventions adopted by the tribe.
Tohono O’odham is also referred to in the linguistics literature as Papago (Hale 2002).
2
The subject mistol ‘cat’ is preceded by a determiner in (3b), but not in (3a) because “the g
determiner never occurs at the beginning of a sentence (Zepeda 1983:8).”
86
(b) N-o
g
mistol ko:s.?
Qu-AUX DET cat
sleeping
‘Is/was the cat sleeping?’ (Zepeda 1983:13)
The Tohono O’odham Qu-morpheme n is pronounced and so it clearly has a
[+OVERT] feature. It also appears to be an affix.
Evidence that n ‘Qu’ is an affix is given by Zepeda (1983:14), who writes that in a
yes/no question, “since the auxiliary combines with the n-, it loses its initial glottal
stop.” In (3a), the ’ at the beginning of ’o ‘AUX’ represents a glottal stop. This
glottal stop shows that the AUX is an independent phonological word. When the
AUX combines with the Qu-morpheme in an interrogative construction, as in (3b),
the glottal stop disappears, indicating that the AUX is no longer an independent
phonological word, but has combined with the Qu-morpheme. The AUX combines
with the Qu-morpheme because the Qu-morpheme has a [+AFFIX] feature.
The Tohono O’odham Qu-morpheme is an affix, and therefore, it requires an
element to host it. The [+AFFIX] feature forces the Qu-morpheme to attract the
AUX which is the closest head, forcing the AUX to suffix onto it. In (3b), if the
Qu-morpheme n is in Typ, then the AUX raises from T to Typ, a shown in (4).
(4)
TypP
Typ’
Typ
n-o1
Qu-AUX DP
mistol
cat2
TP
T’
T
t1
vP
t2 ko:s
sleeping
87
The fact that the Tohono O’odham Qu-morpheme is a prefix is dependent on
the [+AFFIX] feature of the Qu-morpheme and a requirement that the AUX appear
in second position in a clause. This phenomenon whereby a clitic appears in second
position is known as Wackernagel’s law, which Hale (2002:303) describes as “the
principle of Second Position Placement of certain elements.”3 Importantly, if the
Qu-morpheme were to suffix onto the AUX, then the AUX would end up in clauseinitial position, which is not allowed.
Haida4 also contains a Qu-morpheme that is an overt affix. Enrico (1986:103)
writes that yes/no constructions “require the clitic particle gu after the first constituent.” In other words, the Qu-morpheme gu must appear after the clause-initial
element, unlike the Tohono O’odham Qu-morpheme which must precede the clauseinitial element. Compare the statement in (5) with the yes/no construction in (6),
which is formed by adding the Qu-morpheme gu to the end of the initial lexical
item.
(5)
Haida:
Bill ’la qing-gan.
Bill 3p see-Past
‘She saw Bill/Bill saw her.’ (Enrico 2003:151)
(6)
Haida:
Daa-gu 0 gudang?
you-Qu 3p understand
‘Do you understand it?’ (Enrico 2003:132)
3
4
For discussion of why Wackernagel’s law exists, see Anderson (1993), among others.
This is a language spoken in the Queen Charlotte Islands of British Columbia, and Dall and
the southern Prince of Wales Islands in Alaska (Enrico 2003).
88
The Haida Qu-morpheme attaches onto whatever is in clause-initial position, and
notably, there is a lot of freedom as to what can appear in clause-initial position. In
(7a), the Qu-morpheme gu attaches onto a pronoun, which is ambiguous between
the subject or object. In (7b), the Qu-morpheme attaches onto a locative, and in
(7c) it attaches onto a pronominal subject.
(7)
(a) 7laa-gu
dang skudaa-yaa?
3rdper-Qu you punch-nw
‘Did you punch him?’ or ‘Did he punch you?’
(b) 7anaa-gu Bill 7is?
inside-Qu Bill copula
‘Is Bill inside?’
(c) daa-gu king?
you-Qu see
‘Do you see it?’ (Enrico 1986:104)
One limitation is that the clause initial element cannot be a verb (Enrico 1986).
Enrico (2003:131) considers the Haida Qu-morpheme to be an affix, as he glosses
it as an affix and describes it as an “interrogative clitic.” There is syntactic evidence
to support this idea. When there is no non-verbal lexical item for the Qu-morpheme
to affix onto in clause-initial position, then the dummy element huu is inserted to
serve as a host. Enrico (1986:105) writes that huu is an adverbial that has “the
demonstrative meaning ‘there,’” but when used with the Qu-morpheme “huu . . . is
meaningless and it is gu that is in some sense basic.” In (8), note that gu ‘Qu’
follows this dummy huu.
(8)
Haida:
Huu-gu
tajuu?
Dummy-Qu be blowing/windy
‘Is it windy?’ (Enrico 1986:105)
89
These facts indicate that the Haida Qu-morpheme has a [+AFFIX] feature.
Note that the Haida Qu-morpheme functions as a suffix.
Like in Tohono
O’odham, I assume that this is due to facts that are external to the Qu-morpheme.
Specifically, the clause-initial element precedes the Qu-morpheme. This is likey because it is in a higher projection, such as TopP, than where the Qu-morpheme is.
The clause initial element, like the Qu-morpheme, is in the periphery of a clause,
whereas the elements following the Qu-morpheme are located within the TP. Therefore, it may be that the Qu-morpheme suffixes onto the clause-initial element simply
because it is closest.
I next turn to the Japanese Qu-morpheme ka, which is also overt, but unlike
the Tohono O’odham and Haida Qu-morphemes, it is not an affix. The Japanese
Qu-morpheme thus contains [+OVERT] and [-AFFIX] features. Example (1) from
chapter 3, repeated below as (9), demonstrates that ka is a Qu-morpheme. The
statement in (9a) becomes a yes/no construction when the Qu-morpheme ka is
added to clause-final position as shown in (9b).
(9)
Japanese:
(a) Kare-wa ie-ni
kaerimashita.
He-top home-dat went
‘He went home.’
(b) Kare-wa ie-ni
kaerimashita ka?
He-top home-dat went
Qu
‘Did he go home?’
Evidence that ka ‘Qu’ is an independent lexical item is that it is not possible
to pause before a suffix, but it is possible to pause before ka ‘Qu.’ Examples (10ab) show two verbs with the present tense suffix -masu. The verb iki-masu ‘go’ is
composed of iki ‘go’ and the suffix -masu ‘pres,’ and tabe-masu ‘eat’ consists of
tabe ‘eat’ and -masu ‘pres.’ The verbs in (10a-b) are single phonological words that
each consists of two morphemes.
90
(10) Japanese:
(a) iki-masu
go-pres
‘go’
(b) tabe-masu
eat-pres
‘eat’
It is ill-formed to pause when speaking between the verb roots and -masu, as shown
in (11a-b) because -masu is an affix. It is possible to pause before ka ‘Qu’ as shown
in (11c). Although it may be a bit odd to pause before ka ‘Qu,’ it is not nearly as
bizarre as pausing before -masu. This is then evidence that ka ‘Qu’ is not an affix.
(11) Japanese:
(a) *iki-PAUSE-masu
go-PAUSE-pres
‘go’
(b) *tabe-PAUSE-masu
eat-PAUSE-pres
‘eat’
(c) ?(Anata-wa) ikimasu PAUSE ka?
(You-top)
go
PAUSE Qu
‘Are you going?’
English contains an [+OVERT] [-AFFIX] Qu-morpheme if that appears in embedded yes/no constructions.5 It cannot appear in wh-constructions, nor in matrix
interrogatives. Example (12a) shows a matrix statement. In its yes/no counterpart,
(12b), the auxiliary should is fronted to clause-initial position where a null matrix
Qu-morpheme attaches onto it (see the next subsection for a discussion of why the
5
See Baker (1970) for arguments that if is the overt realization of a Qu-morpheme.
91
auxiliary moves and attaches onto the Qu-morpheme). When this yes/no construction is placed in an embedded clause, as in (12c-d), the auxiliary cannot front to
clause-initial position; instead the overt Qu-morpheme if is required.6 I hypothesize
that the auxiliary cannot front because an embedded clause cannot contain a null
Qu-morpheme in English. Example (12c) in which the embedded clause contains
a fronted auxiliary is ill-formed.7 The embedded clause functions as a well-formed
embedded yes/no construction when it contains the overt Qu-morpheme if, as shown
in (12d).
(12)
(a) She should go.
(b) Should1 she t1 go?
(c) *I wonder [should1 she t1 go].
(d) I wonder [if she should go].
This Qu-morpheme if contains a set of features that requires it to only appear in
an embedded clause.
The Qu-morpheme if is not an affix. Just as it is possible to pause before the
Japanese Qu-morpheme ka, it is possible to pause before, as well as after, if. For
example, in (13), it is possible to pause between wonder and if, and between if and
it’s. Even with a pause, this construction is understandable.
(13) I wonder (PAUSE) [if (PAUSE) it’s true].
Compare this example with a word that contains a clear affix, such as cat-s. It is
extremely odd to pause between cat and the plural morpheme -s, because -s is an
affix that attaches onto cat.
6
Whether may also occur in this position. Whether differs in its behavior from if and thus it
is not clear to me whether or not it is a wh-phrase or a Qu-morpheme, although I suspect that it
is a Qu-morpheme. See Katz & Postal (1964), Larson (1985), and Kayne (1991), as well as the
discussion of whether and if in chapter 2.6.
7
This constructions is ill-formed in ‘standard’ English. In some dialects of English, an auxiliary
can appear in the initial position of an embedded clause. See section 4.2.3.
92
4.2.2 Covert Qu-morphemes
I next discuss several different types of null Qu-morphemes; morphemes with a
[-OVERT] feature. English and Spanish have Qu-morphemes that are null affixes.
Dholuo, Hopi, Swati, Mandarin Chinese, and Persian have Qu-morphemes that show
up as intonation. These Qu-morphemes and their relevant features are given in (14).
(14)
LANGUAGE
FEATURES
English, Spanish
[-OVERT], [+AFFIX]
Dholuo, Hopi, Swati, Mandarin, Persian
[-OVERT], [-AFFIX]
English does not have an overt Qu-morpheme in matrix clauses, and so its matrix
Qu-morpheme clearly must be null, indicating it has a [-OVERT] feature. Furthermore, the Qu-morpheme causes the element in Tense to raise to Typ. This is because
the Qu-morpheme is an affix. In the statement in (15a) below, the Tense head contains the auxiliary could. In the yes/no version (15b), could has raised to Typ to
serve as a host for the Qu-morpheme. The Qu-morpheme’s affix property forces the
element in T, in this case could, to raise to Typ.
(15)
(a) He [T could drink the wine] .
(b) [T ypP [T yp Could1 -∅Qu he [T P [T t1 drink the wine]]]]?
When there is no element in T, then a form of do appears. Compare (16a-b). In
a statement such as (16a), the Tense head, with a value of [past] is pronounced as
part of the verb ate. In (16b), the Tense head raises to Typ where it attaches onto
the null Qu-morpheme. Since it has moved away from the verb, it can no longer be
pronounced as part of the verb, and English resorts to a form of the dummy element
do.
(16)
(a) He [T ∅P AST ate the food].
93
(b) [T ypP [T yp Did1 -∅Qu [T P he [T tP AST 1 eat the food]]]]?
Spanish also has a null Qu-morpheme that is an affix. The statements in (17-18a)
become yes/no constructions, as shown in the corresponding (b) examples, when the
verbal elements raise to clause-initial position.
(17) Spanish:
(a) Usted es americano.
you
are American
‘You are American.’
(b) Es1 usted t1 americano?
are you
American
‘Are you American?’ (Stein 2005:184)
(18) Spanish:
(a) Linda viene
tarde.
Linda is coming late
‘Linda is coming late.’
(b) Viene1 Linda t1 tarde?
is coming Linda
late
‘Is Linda coming late?’
(Stein 2005:184)
In Spanish yes/no constructions, as in English, the Tense morpheme raises to Typ
where it affixes onto a null Qu-morpheme. However, unlike in English, Tense can
raise to Typ along with a main verb that it is affixed to. In (18b), the main verb
viene ‘is coming’ raises to Typ.
Some languages can form yes/no constructions merely with intonation.8 I propose that these languages contain a Qu-morpheme with [-OVERT] and [-AFFIX]
8
Some languages of this type can also form yes/no constructions with Qu-morphemes. For
example, Japanese has overt Qu-morphemes, as discussed above, but yes/no constructions can
94
features. The Qu-morpheme is neither overt, nor is it an affix, as there is no apparent movement of lexical items to indicate affixation. These Qu-morphemes have a
particular intonation pattern that influences the pronunciation of a construction. I
do not examine the exact intonation patterns here, but if Greenberg (1966) is correct, then the question intonation pattern always, or usually, shows up at the end
of a clause. Greenberg (1966:80) writes that the following is a language universal.
When a yes-no question is differentiated from the corresponding assertion
by an intonational pattern, the distinctive intonational features of each
of these patterns are reckoned from the end of the sentence rather than
from the beginning.
Intonation is used to form interrogatives in a number of languages. For example,
in Dholuo,9 according to Omondi (1982:137), “[yes/no] questions are formed from
statements or indicative sentences by a change in intonation. Whereas statements
end in a final fall, questions end in a final rise.”
(19) Dholuo:
(a) Nyathi madho chak. [Final fall]
‘[The/a] baby is drinking milk.’
(b) Nyathi madho chak. [Final rise]
‘Is [the/a] baby drinking milk?’ (Omondi 1982:138)
also be formed with intonation alone. Similarly, a yes/no construction in English can be formed
with intonation, without movement of an auxiliary. For example, ‘You’re sick?’ can function as
a yes/no construction if said with the appropriate intonation, although it is not clear to me that
this is the same type of question as one with auxiliary movement, such as ‘Are you sick?’. ‘You’re
sick?’ is more likely to be used as an echo question asking for confirmation about information
previously heard than as a standard yes/no construction asking for new information.
9
This is a language spoken in Western Kenya (Omondi 1982).
95
Kalectaca (1978:102) writes that “[u]se of special intonation is the normal way in
Hopi10 of forming questions answerable by ‘yes’ or ‘no.’” For example, (20) means
The blanket is red. or Is the blanket red? depending on the intonation.
(20) Hopi:
Taapalo paalangpu./?
blanket red
‘The blanket is red./Is the blanket red?’ (Kalectaca 1978:102)
Swati11 is another language that uses intonation to form yes/no constructions. Ziervogel & Mabuza (1976:219) write that a yes/no construction is formed via “a change
of intonation in the interrogative clause” version of the following example.
(21) Swati:
Usaphila./?
‘He is still well.’/‘Is he still well?’ (Ziervogel & Mabuza 1976:219)
Mandarin Chinese also can use intonation alone to form a yes/no question. Tiee &
Lance (1986:56) write that “[a] simple “yes/no” question can be formed merely by
raising the pitch of one’s voice at the end of a statement.” For example, (22) could
be a statement or yes/no construction, depending on the intonation.
(22) Mandarin:
Nǐ mǎi shū./?
You buy book
‘You buy books./Did you buy books?’ (Tiee & Lance 1986:56)
Persian also forms yes/no constructions with intonation, as shown below. Rising
intonation signals a yes/no construction.
10
11
This is a Uto-Aztecan language spoken in northern Arizona (Kalectaca 1978).
This is a language spoken in Swaziland and South Africa (Ziervogel & Mabuza 1976).
96
(23) Persian:
(a) Kimea madres a-st. [Falling intonation]
Kimea school is
‘Kimea is at school.’
(b) Kimea madres a-st? [Rising intonation]
Kimea school is
‘Is Kimea at school?’ (Simin Karimi, p.c.)
In examples (19-23), a statement and a yes/no construction are identical except
for a change in intonation. This intonation change may influence the way that
phonemes are pronounced, but it does not change the actual string of lexical items
that occur in a statement and a corresponding yes/no construction. Note that there
is no movement of a lexical item. I take this to be evidence that a Qu-morpheme in
these languages is not an affix.
4.2.3 An overt or covert Qu-morpheme
Notably, there are some English dialects that allow either an overt or a covert Qumorpheme to appear in an embedded clause.
Hiberno English has a null Qu-morpheme that occurs in embedded clauses. In
examples (24a-d), the element in Tense raises to adjoin to a null Qu-morpheme.
(24) Hiberno English:
(a) Ask your father [T ypP [T yp does1 -∅Qu [T P he [T tP RES 1 want his
dinner]]]].
(b) I was wondering [T ypP [T yp would1 -∅Qu [T P he [T t1 come home for the
Christmas]]]].
97
(c) Do you remember [T ypP [T yp did1 -∅Qu [T P they [T tpast1 live in
Rosemount]]]]?
(d) I’ve never found out [T ypP [T yp would1 -∅Qu [T P he [T t1 really have
come with me]]]]. (McCloskey 1991:294)
Furthermore, Hiberno English also contains the overt Qu-morpheme if. (25a)
shows that Tense can raise to Typ and attach onto a null Qu-morpheme, and (25b)
shows that the overt Qu-morpheme if can appear. Both Qu-morphemes cannot
co-occur, as shown in (25c), since a single head cannot be filled with two Qumorphemes.
(25) Hiberno English:
(a) Ask your father [T ypP [T yp does1 -∅Qu [T P he [T tpres1 want his dinner]]]].
(b) Ask your father [T ypP [T yp ifQu he wants his dinner]].
(c) *Ask your father [T ypP [T yp ifQu does1 -∅Qu he tpres1 want his dinner]].
(McCloskey 1991:295)
The fact that both a null Qu-morpheme and an overt Qu-morpheme can appear in an
embedded clause is evidence that Hiberno English contains two types of embedded
clause Qu-morphemes which differ in terms of their featural makeup. Specifically,
if is an overt independent lexical item and the other Qu-morpheme is a null affix.
Ozark English also allows both an overt and a covert Qu-morpheme in embedded
yes/no constructions, as in (26a-b), although note that, according to Rusty Barrett
(p.c.), (26b) is not as natural as (26a).
(26)
(a) I don’t know [T ypP [T yp is1 -∅Qu [T P she [T t1 going]]]].
(b) I don’t know [T ypP [T yp ifQu she is going]]. (Barrett 2008)
98
As in Hiberno English, both the overt and covert Qu-morphemes cannot co-ccur,
as indicated by the ill-formedness of (27) below. This is to be expected since Typ
cannot simultaneously contain two Qu-morphemes.
(27) * I don’t know [T ypP [T yp ifQu is1 -∅Qu she tpres1 going]]. (Barrett 2008)
One factor that determines the use of the two possible Qu-morphemes in an
embedded yes/no construction in Hiberno and Ozark English likely involves the
semantics of the Qu-morpheme. Rusty Barrett (p.c.) notes that when a yes/no
construction follows unless, as in (28a-b), there is a clear semantic difference. (28a),
which I assume contains a [+AFFIX] Qu-morpheme, emphasizes doubt that she will
go, whereas (28b) does not emphasize doubt.
(28)
(a) I won’t go unless [is she going too]. (emphasizing doubt that she will go)
(b) I won’t go unless [she is going too]. (Rusty Barrett, p.c.)
More examination of the semantic differences between overt and covert Qumorphemes in Hiberno and Ozark English is required, but the evidence from Ozark
English is suggestive that a speaker uses the overt Qu-morpheme in a different
situation than the covert Qu-morpheme. This is to be expected, since if a language
has multiple Qu-morphemes they are unlikely to be completely identical. If they
were identical, then why would a language need multiple ones? Also, as discussed in
chapter 3.3, a number of languages have multiple Qu-morphemes which, although
identical in that they serve the function of typing a clause as an interrogative, differ
in their semantic and syntactic behavior.
4.2.4 Summary
In this section, I have argued that the features that a Qu-morpheme has play important roles in the formation of yes/no constructions. A Qu-morpheme has [±]
valued [OVERT] and [AFFIX] features. A [±OVERT] feature determines whether
99
or not a Qu-morpheme is overt, and a [±AFFIX] feature determines whether it
shows up as an affix or an independent lexical item. Notably, a Qu-morpheme that
is an affix can force movement of a tense element, as is the case in English and
Spanish. If a Qu-morpheme has [-] values of both of these features, then it shows
up as a particular intonation pattern that indicates a yes/no construction. Thus
the particular features contained within a Qu-morpheme influence the structure of
a yes/no construction.
4.3 Qu-morphemes in Typ
Cross-linguistically, in yes/no constructions, Qu-morphemes frequently appear in a
clause-peripheral position. As discussed in chapter 2.5, I claim that TypP is associated with clause typing. Typing of a clause as a yes/no construction occurs when
the Qu-feature of a Qu-morpheme checks an uninterpretable feature in Typ. Since
TypP occurs in a clause peripheral position, a clause-pheripheral Qu-morpheme is
likely Merged directly in Typ. All of the languages discussed in the previous section
appear to contain clause-peripheral Qu-morphemes, assuming that a Qu-morpheme
that surfaces as intonation alone is also in a clause-peripheral position. In this section I demonstrate that a Qu-morpheme in Modern Japanese and English occurs
directly in Typ. The purpose of this discussion is to show that in general (although
not always, as I discuss in section 4.4), a Qu-morpheme is Merged directly in Typ.
4.3.1 Modern Japanese
In Modern Japanese, a Qu-morpheme must appear at the end of a clause. In the
following Japanese example, the Qu-morpheme ka occurs in clause-final position. If
it is in any other position, the result is an ill-formed yes/no construction.
(29) Kare-wa (*ka) ie-ni
(*ka) kaerimashita ka?
He-top
home-dat
went
Qu
‘Did he go home?’
100
This example can be accounted for as shown in (30), with the Qu-morpheme Merged
directly in Typ. The FQu subscript on the Qu-morpheme represents a Qu-feature.
(30)
TypP
Typ’
TP
Typ
kaFQu
T’
kare-wa ie-ni kaerimashita
‘he-top home-dat went’
Furthermore, the Japanese Qu-morpheme can only take scope in the clause in
which it appears. Example (31a) shows a yes/no construction with an embedded
clause. The Qu-morpheme ka is in the matrix clause, which it types as a yes/no
construction. Example (31b) with a Qu-morpheme in the embedded clause is illformed since the Qu-morpheme cannot take matrix scope.12
(31)
(a) Anata-wa [kare-ga gohan-o suki datta to]
omottimasu ka?
you-top he-nom food-acc like did
comp think
Qu
‘Do you think he liked the food?’
(b) *Anata-wa [kare-ga gohan-o suki-datta ka to]
omottimasu?
you-top
he-nom food-acc like-past Qu comp think
‘Do you think he liked the food?’
Example (32) below contains an embedded yes/no construction. The Qu-morpheme
ka appears at the end of the embedded clause and it cannot take matrix scope. The
embedded clause can only be interpreted as a yes/no construction, and the matrix
clause can only be interpreted as a non-interrogative.13
12
The embedded clause requires that there be a complementizer to. I assume that to is in the
head of ForceP. See chapter 2.5 for discussion of clause structure.
13
If the matrix clause ends in rising intonation, it can be construed as an interrogative yes/no
construction. In this case, the meaning is ‘Do you know if he came?’. But this interrogative
101
(32) Anata-wa [kare-ga kitta ka] shiteiru.
you-top he-nom came Qu know
(i) ‘You know if he came.’
(ii) ∗Do you know (that) he came?
These facts indicate that the Qu-morpheme is Merged directly in the Typ head in
which it has scope.
4.3.2 English
The English Qu-morpheme, like the Japanese Qu-morpheme, occurs in a fixed position, a fact that is accounted for if it is Merged directly in Typ of the clause in
which it has scope.
The English Qu-morpheme can only appear in a clause-peripheral position corresponding to Typ. In (33-34a) the auxiliary element in T has raised to clause-initial
position to affix onto a null Qu-morpheme, as shown in the corresponding (b) examples.
(33)
(a) Could he drink the wine?
(b) ∅Qu -Could1 he t1 drink the wine?
(34)
(a) Did he eat the food?
(b) ∅Qu -Did1 he tpast1 eat the food?
This auxiliary is confined to clause initial position. It cannot follow the subject as
in (35) and result in a yes/no construction.
interpretation must, at least under my analysis, result from the presence of a null Qu-morpheme in
the matrix clause that is separate from the Qu-morpheme of the embedded clause, as the embedded
clause must still be an interrogative.
102
(35)
(a) * ∅Qu - he did eat the food?
(b) * ∅Qu - he could drink the wine?
Similarly, in an embedded clause the overt Qu-morpheme if must occur at the
beginning of the clause, preceding the subject.14 Example (36a) is fine because if
precedes the subject he, but (36b) is ill-formed because if follows the subject.
(36)
(a) I wonder [if he ate the food].
(b) *I wonder [he if ate the food].
These facts suggest that the Qu-morpheme occurs in a clause peripheral TypP head.
It cannot be preceded by a subject, since a subject occurs below Typ in [Spec, TP].
Furthermore, a Qu-morpheme in an embedded clause cannot have matrix scope.
Example (37a) contains a statement with an embedded declarative clause. Example
(37b) shows that if the embedded clause contains the Qu-morpheme if, then if
cannot make the matrix clause into a yes/no construction. Note that the Qumorpheme can have embedded scope as shown in (36a) above. But in (37), the verb
think selects for a statement, and so the embedded clause cannot be an interrogative.
(37c) shows that a null Qu-morpheme in an embedded clause cannot have matrix
scope, although this is also ill-formed because a null Qu-morpheme simply cannot
appear in an embedded clause in Standard English. Example (37d) is the only way
to form a matrix yes/no construction in this case; there must be a Qu-morpheme in
the matrix clause.
(37)
(a) You think [that she should go].
(b) *You think [if she should go].
(c) *You think [∅Qu -should1 she t1 go].
14
See section 4.2.1 for arguments that if is a Qu-morpheme.
103
(d) ∅Qu -Do1 you tpast1 think [(that) she should go]?
The fact that the Qu-morpheme in the embedded clause cannot have matrix scope
is evidence that the Qu-morpheme must be in Typ of the clause in which it has
scope.
The English data then show that the English Qu-morpheme is simply Merged
directly in Typ where its Qu-feature types a clause as an interrogative.
4.4 TP-internal Qu-morphemes
As I discuss in chapter 2.2, it is less work to Merge a syntactic object into a derivation
then it is to Merge a syntactic object and then move it (i.e., remerge it in a new
position). Therefore, it is most economical for a Qu-morpheme to be Merged directly
in Typ from where its Qu-feature can type a clause. As discussed in section 4.3,
this option appears to be the only one in languages such as Japanese and English.
There are languages, however, that do not always require a Qu-morpheme to
be directly Merged in Typ in a yes/no construction. In these languages, a Qumorpheme can appear in a TP-internal position; i.e., a position below TP. In the
following examples from Sinhala, the Qu-morpheme [email protected] appears in TP-internal positions (see section 4.4.1 for evidence that the Qu-morpheme actually is in a TPinternal position).
(38) Sinhala:
(a) Chitra [ee [email protected]][email protected] kieuwe?
Chitra that book-Qu read.E
‘Was it that book that Chitra read?’ (Kishimoto 2005:11)
(b) Siri [email protected]@
waduwæ[email protected]
[email protected]@nne?
˙
˙
˙
Siri tomorrow-Qu
woodworking
do-PRES.E
‘Is it tomorrow that Siri does woodworking?’ (Gair & Sumangala
1991:96)
104
When the Qu-morpheme is Merged in a position below Typ as in (38a-b) then an
operation in addition to the initial Merge of the Qu-morpheme is required for the
Qu-feature to type a clause. This raises the following question: why can a Qumorpheme be Merged in a position other then Typ in languages such as Sinhala? I
claim that the answer involves a Focus-feature.
I propose that there are two types of Qu-morpheme, shown in (39a-b).
(39)
(a) Qu[FQu ]
(b) Qu[FF oc ,FQu ]
Languages such as Japanese and English contain only the Qu-morpheme in (39a),
which has a Qu-feature FQu and is Merged directly in Typ. This Qu-morpheme
merely serves the purpose of typing a clause as an interrogative. Other languages
contain the Qu-morpheme (39b), which has a Focus-feature FF oc in addition to a
Qu-feature FQu .
A Qu-morpheme that has both Qu- and Focus-features serves two functions; 1)
it types a clause as an interrogative and, 2) it focuses a particular phrase within
a clause. The Qu-feature is responsible for typing a clause and the Focus-feature
is responsible for focusing a phrase. I propose that this type of Qu-morpheme is
Merged in a position adjoined to a focused phrase, as shown in (40).
(40)
XP
XP
Qu[FQu ,FF oc ]
The Qu-morpheme’s Focus-feature moves to the head of a Focus Phrase (FocP),
where it focuses its associated phrase, and the Qu-feature moves to Typ where it
types a clause. Following Alexiadou & Anagnostopoulou’s (1998) view that an EPP
feature can motivate head movement (see chapter 2.3), I propose that EPP features
in FocP and TypP force movement of Qu- and Focus-features.
105
Evidence for this proposal that a Focus-feature forces a Qu-morpheme to be
Merged TP-internally comes from the interpretation of constructions with TPinternal Qu-morphemes. The phrase that the Qu-morpheme is adjacent to is emphasized in a manner consistent with the notion of ‘focus.’
Focus is an emphasized part of a sentence that presents new information (Rizzi
1997). By ‘emphasized’ I refer to an element that stands out from the rest of
the sentence via its syntactic position, pronunciation, or adjacency to a particular
morpheme that marks focus (this is the phenomenon examined in this section). A
Qu-morpheme with a Focus-feature emphasizes a particular phrase contained within
a yes/no construction, and furthermore, this type of focus appears to be what Kiss
(1998) refers to as “identificational focus.”
Kiss (1998) argues that there are two types of focus: identificational focus (which
is interpreted as contrastive focus in some languages) and information focus. According to Kiss, both types of focus refer to new information that is emphasized,15
but their primary differences are 1) that an identificational focus refers to an exhaustive set (all elements of a set), whereas information focus does not, and 2) an
identificational focus appears in the specifier of a Focus Phrase (FocP) projection
in the left periphery of a clause, whereas information focus is not associated within
any particular syntactic position. For example, an identificational focus occurs in a
preverbal position in Hungarian, and as a cleft in English, whereas an information
focus occurs in a postverbal position in Hungarian and in a non-clefted construction
in English.16
Examples (41a-b) demonstrate information focus.
15
Although the information is new, it can make reference to old information by referring to a
presupposed set. I thank Andrew Carnie for pointing this out to me.
16
See Kiss (1998) for further discussion of the differences between these types of focus.
106
(41) Hungarian:
(a) Mari ki nézett magának
egy kalapot és egy kabátot.
Mary out picked herself.dat a
hat.acc and a
coat.acc
‘Mary picked a hat and a coat for herself.’
(b) Mari ki nézett magának
egy kalapot.
Mary out picked herself.dat a
hat.acc
‘Mary picked a hat for herself.’ (Kiss 1998:250)
In (41a), a hat and a coat, in both Hungarian and English, is new information that
is emphasized, but it does not refer to an exhaustive set. Therefore, (41b) can be
a logical consequence of (41a). Note that the information focus does not occur in a
special position within a clause.
The examples in (42a-b) from Hungarian, along with their corresponding English
glosses, demonstrate identificational focus.
(42) Hungarian:
(a) Mari egy kalapot és egy kabátot nézett ki magának.
Mary a
hat.acc and a
coat.acc picked out herself.to
‘It was a hat and a coat that Mary picked for herself.’
(b) Mari egy kalapot nézett ki magának.
Mary a
hat.acc picked out herself.to
‘It was a hat that Mary picked for herself.’ (Kiss 1998:250)
In (42a), Kiss argues that a hat and a coat is an identifiational focus both in Hungarian and in the corresponding English gloss. Kiss points out that (42b) is not a
“logical consequence” of (42a), because a hat and a coat is a contrastive focus that
refers to an exhaustive set.17 The idea here is that if one says (42a), that person
expresses the idea that Mary must have picked both a hat and a coat, whereas (42b)
indicates that Mary only picked a hat, and so (42a) does not imply that (42b) is
17
Kiss attributes this type of test for exhaustivity to Szabolcsi (1981).
107
true. (42a) thus refers to an exhaustive set. In (42a), a hat and a coat in both Hungarian and English is argued to be in [Spec, FocP], which corresponds to a preverbal
position in Hungarian and a cleft in English.18 Thus (42a) has an element, a hat
and a coat that for semantic and syntactic reasons, is an identificational focus.
When a Qu-morpheme occurs in a TP-internal position, as I discuss in the following subsections, the resulting constructions correspond to clefted yes/no constructions in English, such as (43) below.
(43) Was it Bob’s book and Mary’s book that Harold read?
Assuming that Kiss’ claims are correct, then the clefted phrase Bob’s book and
Mary’s book in (43) is an identificational focus. For example, if one answers this
question with a yes, then the implication is that (44) is true, and the clefted phrase
here refers to an exhaustive set, as in (42a) above.
(44) It was Bob’s book and Mary’s book that Harold read.
My analysis differs from Kiss’ analysis with respect to the FocP. Kiss argues that
a phrase that is an identificational focus must appear in [Spec, FocP]. However, I
argue that this need not be the case. A phrase with an adjacent Qu-morpheme can
be an identificational focus without appearing in FocP. The focus interpretation is
obtained via movement of a Focus-feature. Specifically, a Focus-feature associated
with a Qu-morpheme that is adjoined to a salient phrase raises to the head of FocP
and gives the phrase an identificational focus interpretation; the phrase itself need
not appear in FocP. I elaborate on this proposal in the following subsection.
Below, I argue that Sinhala, Premodern Japanese, Okinawan, Tupı́, and Ewen
are languages that have the two types of Qu-morphemes in (39a-b), one with and
one without a Focus-feature. The Qu-morpheme without a Focus-feature (this is
essentially the ‘default’ type of Qu-morpheme) simply is Merged in Typ and types
18
See Kiss (1998) for further discussion of the clause structure of this type of construction.
108
a clause as an interrogative. The Qu-morpheme with a Focus-feature is Merged
TP-internally, and serves the dual function of typing the clause as an interrogative
and focusing an adjoined phrase.
Before I begin my investigation of TP-internal Qu-morphemes, I note that in
the following discussion I do not discuss the [OVERT] and [AFFIX] features of the
Qu-morphemes. All of the Qu-morphemes that I examine are overt. This is simply
because it is difficult to determine if a [-OVERT] Qu-morpheme is in a TP-internal
position.19 Also, although some of the Qu-morphemes in these languages may be
affixes, to the best of my knowledge an [AFFIX] feature does not force movement
of any element, and thus an [AFFIX] feature does not play a particularly important
role in determining the syntactic structure of these constructions.
4.4.1 Sinhala
Sinhala is a language with a Qu-morpheme [email protected] that can appear in clause-final position. For example, the statement in (45a) becomes a yes/no construction when [email protected]
‘Qu’ is added to the end of the clause, as shown in (45b). The fact that the addition
of [email protected] results in a yes/no construction is evidence that [email protected] is a Qu-morpheme. Also,
note that the verb ends in -a, glossed as ‘-A.’20 The importance of this fact will
become evident in the following discussion.
(45) Sinhala:
(a) [email protected]@[email protected] basekaka [email protected]
Colombo bus
be.A
‘There is a bus to Colombo.’
19
One way of determining if there is a [-OVERT] Qu-morpheme in a TP-internal position would
be to examine intonation patterns (Heidi Harley, p.c.), which is outside the scope of this work.
20
I follow Sumangala (1992) in glossing -a as ‘-A.’
109
(b) [email protected]@[email protected] basekaka [email protected] [email protected]?
Colombo bus
be.A
Qu
‘Is there a bus to Colombo.’ (Gair 1970:139)
Notably, the Sinhala Qu-morpheme is not confined to clause-final position; it
can occur in a TP-internal position as a suffix on a focused phrase. In (46a), the
Qu-morpheme appears in the typical clause-final position, but in (46b), it appears
as an affix on the nominal ee [email protected] ‘that book,’ and a yes/no construction results.
The semantics of (46a) and (46b) differ in that in (46b) the phrase ee [email protected] ‘that
book,’ to which the Qu-morpheme is affixed, is focused, as evidenced by the clefted
English gloss. Also, when the Qu-morpheme appears in a TP-internal position, the
verb occurs with an -e ending, which I gloss as ‘-E.’21 This verbal ending differs
from the ‘-A’ verbal suffix that occurs when a Qu-morpheme appears in clause-final
position, as in (45a) above.
(46) Sinhala:
(a) Chitra ee
[email protected] kieuwa [email protected]?
Chitra that book read.A Qu
‘Did Chitra read that book?’
(b) Chitra [ee [email protected]][email protected] kieuwe?
Chitra that book-Qu read.E
‘Was it that book that Chitra read?’ (Kishimoto 2005:11)
Example (47) below shows that the Qu-morpheme may even attach onto adjuncts
such as [email protected] ‘tomorrow.’
˙
(47) Sinhala:
Siri [email protected]@
waduwæ[email protected]
[email protected]@nne?
˙
˙
˙
Siri tomorrow-Qu
woodworking
do-pres.E
‘Is it tomorrow that Siri does woodworking?’ (Gair & Sumangala 1991:96)
21
I follow Sumangala (1992) in glossing -e as ‘-E.’
110
Three descriptive facts about these constructions are of importance for this analysis. First, the presence of the Qu-morpheme is responsible for a yes/no interrogative
interpretation. Second, when the Qu-morpheme attaches onto a TP-internal phrase,
as in (46b) and (47), this phrase becomes focused, as indicated by the clefted glosses.
As discussed in the previous section, this appears to be identificational focus. The
third fact about these constructions is that the verb ending differs depending on
where the Qu-morpheme is pronounced. When the Qu-morpheme occurs in clausefinal position, the verb has an ‘-A’ suffix. The verb also appears with this suffix
in non-interrogative constructions such as (45a) above. When the Qu-morpheme
appears in a TP-internal position, the verb of the clause that the Qu-morpheme is
associated with occurs with the ‘-E’ ending. I next turn to an explanation for these
facts.
When a Qu-morpheme occurs in a clause-peripheral position, it is most likely
directly Merged in Typ. First of all, direct Merge in Typ is the most economical
operation (see chapter 2.2.4). Secondly, the Typ head in Sinhala occurs clausefinally. Sinhala is an SOV language with head-final projections (Sumangala 1992).
For example, the noun minissu ‘people’ and the verb bivva ‘drink-past’ follow their
complements in (48a-b) respectively.22
(48) Sinhala:
(a) NP: gamee
minissu
village-gen people
‘people of the village’
(b) VP: tee bivva
tea drink-past
‘drank tea’ (Sumangala 1992:41)
22
Sinhala allows scrambling. When there is scrambling, a complement need not precede its head.
However, if one considers the underlying word order of scrambled constructions to be SOV, then
the head must follows its complement in the underlying structure.
111
Since the Qu-morpheme occurs at the end of the clause in constructions such as
(46a), a natural assumption would be that it occurs in Typ, as I claimed in section
4.3 was the case in Japanese and English.
Thus, example (46a) with a clause-final Qu-morpheme can be accounted for
as shown in (49). The Qu-morpheme is Merged directly in Typ, where the Qufeature values an uninterpretable clausal typing feature and types the clause as an
interrogative.
(49)
TypP
Typ’
TP
Typ
[email protected][FQu ]
Chitra ee [email protected] kieuwa
‘Chitra that book read-A’
When a Qu-morpheme does not appear at the clause-periphery, as in (46b)
repeated below, I have claimed that it is in a TP-internal position.
(46) Sinhala:
(b) Chitra [ee [email protected]][email protected] kieuwe?
Chitra that book-Qu read.E
‘Was it that book that Chitra read?’ (Kishimoto 2005:11)
As noted above, projections in Sinhala are head-final.
Therefore, if the Qu-
morpheme [email protected] in (46b) were in Typ, then everything preceding it would have to
be in a higher position, such as the specifier of [TypP]. But if this were the case,
then it is not clear where the verb kieuwe ‘read-E’ would occur, since Typ linearly
follows the v P projection housing the verb. The verb would have to undergo rightward movement, or be a remnant of a complex series of leftward movements. On
112
the other hand, if the Qu-morpheme occurs below the TP in a position adjoined to
the object, then the facts are accounted for straightforwardly.
In constructions such as (46b), I propose that the presence of a Focus-feature
within a Qu-morpheme forces it to be Merged TP-internally adjacent to a phrase
that is the focus of the interrogative. For example, in (50), the Qu-morpheme [email protected] is
adjoined to an XP that is contained in a TP-internal position in a clause.
(50)
TypP
Typ’
FocP
Typ
Foc’
TP
...
Foc
XP
XP
[email protected][FQu ,FF oc ]
...
In order to focus a phrase, the Focus-feature associated with the Qu-morpheme
must move to the head of a FocP projection. A Qu-morpheme with a Focus-feature
cannot be Merged directly in Typ because then its Focus-feature 1) would not be
associated with any phrase that needs to be focused, and 2) would not be in a
position to check the feature in a FocP head, assuming that TypP is above FocP
(see the discussion of clause structure in chapter 2.5).
Other clausal typing morphemes in Sinhala can also appear TP-internally and
focus an adjacent phrase. The following examples show that tamay ‘certainty’ can
either appear in the clause periphery (51a) or clause internally (51b). When it
113
occurs TP-internally in (51b), if focuses [email protected] ‘tomorrow’ in the same manner that
˙
the Qu-morpheme does in (47).
(51) Sinhala
(a) [email protected]@ [email protected]
gallu [email protected]?
Gunapaala tomorrow Galle go-A-CERTAINTY
‘It is for sure that Gunapala is going to Galle tomorrow.’
(b) [email protected]@ [email protected]
gallu yanne?
Gunapaala tomorrow-CERTAINTY Galle go.E
‘It is certainly tomorrow that Gunapala is going to Galle.’ (Sumangala
1992:131)
Also, lu ‘reportative,’ nee ‘tag question’ and yæ ‘dubitative’ can occur either clause
peripherally or clause internally (cf. Sumangala 1992:131-132). Sinhala clausal
typing morphemes, including the Qu-morpheme, thus are capable of turning a constituent into an identificational focus.
I am now in a position to account for constructions such as (46b), repeated
below, in which the Qu-morpheme occurs as an affix on the TP-internal phrase ee
[email protected] ‘that book.’
(46) Sinhala:
(b) Chitra [ee [email protected]][email protected] kieuwe?
Chitra that book-Q read.E
‘Was it that book that Chitra read?’ (Kishimoto 2005:11)
In (46b) the Qu-morpheme is Merged in a position adjoined to the object ee [email protected]
‘that book’ as shown in (52).
114
(52)
DP
DP
[email protected][FF oc ,FQu ]
ee [email protected]
‘that book’
When a Qu -morpheme is Merged TP-internally in a position below FocP, as
in (46b), its Focus-feature needs to check an uninterpretable feature in the head of
FocP to focus a phrase and the Qu-feature needs to check an uninterpretable feature
in Typ to type a clause. There are at least two possible ways for the features
to do this; one is via Agree between the features and Typ, and another is via
movement. I will take the position that movement is involved. This is because, as I
will discuss in chapter 5, Agree is not subject to island effects, whereas relationships
between the Qu-morpheme and Typ in Sinhala are subject to island effects. If
movement is the crucial operation, then either the Qu- and Focus-features must raise
along with the Qu-morpheme, or they must undergo feature movement (movement
without an associated lexical item). Since the Qu-morpheme clearly does not move
overtly to the clause-periphery, the most likely possibility is that its Qu- and Focusfeatures undergo feature movement. Motivation for movement of these features can
be attributed to EPP features that serve the purpose of forcing these features to
move.
Thus, if this analysis is correct, then in (46b), the Focus- and Qu-features associated with the Qu-morpheme undergo feature-movement that leaves the Qumorpheme in-situ. A diagram is shown in (53). The t[FF oc1,FQu 2] within the QuP
refers to the traces of the Focus- and Qu-features. I assume that the ‘-E’ verbal
ending is the overt pronunciation of a Focus-feature in Foc, and thus I have placed
the ‘-E’ in Foc (I discuss the verbal ending in more detail below). I also assume that
the verb raises to T (I return to this issue below). Note that there are EPP features
in Foc and Typ that force movement of the Focus- and Qu-features, respectively.
115
(53)
TypP
Typ’
FocP
TypEP P
FQu2
Foc’
TP
T’
FocEP P
-eFF oc1
-E
v’
T
kieuw3
‘read’
vP
DP
Chitra
DP
v
t3
DP
QuP
ee [email protected]
‘that book’
[email protected]+t[FF oc1 ,FQu2]
In (53), the Focus-feature and Qu-feature raise together as a feature-bundle to Foc,
where the Focus-feature remains, and then the Qu-feature excorporates and raises
on to Typ.
Verbal agreement in Sinhala provides further evidence for the proposal that a
Focus-features undergoes feature movement. Verbal endings are dependent on the
position of the Qu-morpheme, or other clausal-typing morpheme. As mentioned
above, in Sinhala yes/no constructions, when the Qu-morpheme occurs clausefinally, the verb occurs with the ‘-A’ ending. The ‘-A’ ending is something akin
to a default verbal ending, as suggested by Gair and Sumangala’s (1991:94) statement that the ‘-A’ affix “is the most common [verbal affix] in finite independent
116
sentences.” Below is an example of a declarative sentence with the default ‘-A’ ending on the verb.
(54) Sinhala:
Chitra [email protected] gatta.
Chitra book bought.A
‘Chitra bought the book.’ (Kishimoto 2005:4)
When the Qu-morpheme occurs TP-internally, the verb shows up with the ‘-E’
ending, which I assume is the pronunciation of the Focus-feature in Foc. Specifically,
movement of the Focus-feature to Foc triggers this ‘-E’ suffix. When there is no
Focus-feature in Foc, then the default ‘-A’ ending surfaces on the verb. Only tensed
verbs appear with these suffixes (Gair & Sumangala 1991). Therefore, a reasonable
assumption is that a tensed verb raises as high as T, as is claimed by Gair &
Sumangala (1991), and that the default ‘-A’ ending is in T. When the verb appears
with the Focus-suffix, either the verb remains in T or it raises to Foc. Which actually
occurs is not crucial to my analysis, thus I do not investigate this issue.
This feature movement analysis requires that features (specifically, Qu- and
Focus-features) raise either over or through intervening projections that are filled,
as can be seen in (53) above. The Focus- and Qu-features associated with the Qumorpheme raise over a v P head and a TP head (not shown in the diagram). This
movement, however, appears to be a violation of the Head Movement Constraint
(HMC) (Travis 1984), which states that “a head cannot be separated from its trace
by an intervening head (Lasnik 2003:70).” Within the Minimalist Program, if movement of an element is blocked by an intervening head, there is a violation of the
Minimal Link Condition (see (34) in chapter 2.6). Hagstrom (1998:61) accounts for
why this long head movement may be possible, as follows:
. . . feature attraction drives movement of the closest element with the
relevant feature. If a feature F is being attracted and a head H carries the
117
feature F, movement of H will only be blocked if there is an intervening
head which also carries the feature F. Any head which does not carry
this feature is irrelevant.
Hagstrom’s view is in accord with the notion of Relativized Minimality, proposed
by Rizzi (1990). In a configuration such as (55), Rizzi (1990:2) writes that “if Z is
a potential governor of some kind for Y, it will block only government of the same
kind from X.”
(55) . . . X . . . Z . . . Y . . . (Rizzi 1990:1)
If the notion of different types of government (Chomsky 1981) is replaced with the
notion of different types of features23 then it can be argued that Z will block a
relationship between X and Y only if Z has a particular type of feature that X and
Y have. For example, (56a) allows X and Y to establish a well-formed relationship
(such as movement of an element from X to Y) because Z does not contain the
feature F1 that is contained by X and Y. However, (56b) is ill-formed because Z
contains the same feature F1 that X and Y have, thereby blocking any relationship
between X and Y.
(56)
(a) . . . XF 1 . . . ZF 2 . . . YF 1 . . .
(b) . . . XF 1 . . . ZF 1 . . . YF 1 . . .
*
If these views about head-movement are on the right track, then Qu- and Focusfeatures in Sinhala can move through the heads of intervening projections, at least
as long as these heads do not carry the same type of feature. Movement of these
23
I note that comparing features with types of government is an overly simplistic way to view
things, as the notions of government developed by Chomsky (1981), and much other work, are
complex and cover a broad spectrum of syntactic phenomena.
118
features may proceed through v and T because these projections lack Qu- and
Focus-type features.24
This analysis also accounts for the fact that a Qu-morpheme in Sinhala can
take scope outside of the clause that it appears in. In (57), the Qu-morpheme [email protected]
is attached to [email protected] ‘Nimal-dat’ in the embedded clause and a matrix yes/no
construction results; the embedded clause is not a yes/no construction. In this case,
the verb of the embedded clause has the ‘-A’ ending and the matrix verb, which is
the verb of the clause in which the Qu-morpheme has scope, has the ‘-E’ ending; the
verb of the clause that is typed as an interrogative by the Qu-morpheme appears
with the ‘-E’ ending.
(57) Sinhala:
Gunee [Siri [email protected] [email protected]@ dunna
[email protected]] kivve?
Gune Siri that Nimal-dat-Q give-past.A that] say-past.E
‘Is it to Nimal that Gune said that Siri gave it?’ (Sumangala 1992:135)
Example (57) can be accounted for as follows. The Qu-morpheme adjoins to the
phrase [email protected] ‘Nimal-dat.’ The Focus- and Qu-features of the Qu-morpheme
raise to the matrix Foc and Typ projections, and the Qu-morpheme remains in-situ.
This allows [email protected] ‘Nimal-dat’ to obtain matrix scope as a focused element and
the Qu-morpheme to type the matrix clause. Also, the Focus-feature in the matrix
Foc shows up as the ‘-E’ ending on the matrix verb. A diagram is shown below.
For the sake of simplicity, I have indicated movement of the Qu- and Focus-features
as skipping over intervening heads. However, as discussed above, these features
actually move through intervening heads, but that this movement is simply not
blocked.
24
Note that I differ on this point from Rivero (1991) who argues that long head movement
can skip over intervening heads. Whether or not there is movement through or over intervening
heads is not that crucial to my analysis, however, it is an important issue that requires further
investigation than I have been able to give it here.
119
(58)
TypP
Typ’
FocP
TypEP P
FQu2
Foc’
TP
DP3
Gunee
T’
FocEP P
-eFF oc1
-E
v’
T
kivve3
said
vP
t3
TypP
v
t3
Siri [email protected] [email protected]@+t[FF oc1 ,FQu2] dunna [email protected]
‘Siri that Nimal-dat-Qu
give-past.A that’
Even though the Focus- and Qu-features move through the embedded clause in
(57), the ‘-E’ ending is not triggered on the embedded verb. The ‘-E’ ending only
occurs on the verb of the clause that contains the final landing cites of the Focusand Qu-features. This fact is accounted for straightforwardly if the ‘-E’ ending is
the Spell-out of a Focus-feature. The Focus-feature of the Qu-morpheme does not
sit in the embedded FocP, and thus it should not trigger the ‘-E’ ending on the
embedded verb.
Lastly, there is the issue of how the Qu- and Focus-features in the embedded
clause in (57) can be attracted by elements in the matrix clause. If the Phase
Theory view that only elements in the edge of a phase are accessible to higher
operations, and the embedded TypP, being at the clause periphery, is a phase, then
120
the Qu- and Focus-features must move to the edge of the embedded clause before
they can be accessible to further movement to the matrix clause. On this point, I
assume that EPP features in the Foc and Typ heads in the embedded clause attract
the Qu- and Foc-features. Since the Foc and Typ projections are at the periphery of
the embedded clause, they are at the phase edge and thus features in the embedded
Foc and Typ are accessible to attraction by EPP features in the matrix clause.
In summary, I have taken the position that the Sinhala [email protected] is a Qu-morpheme
that types a clause as an interrogative and that where it occurs in a construction
boils down to whether or not it contains a Focus-feature. Furthermore, my analysis
relies on the notion that there is a fixed FocP projection in the clausal periphery.
My analysis differs in several ways from the earlier analysis of Sumangala (1992).
Sumangala (1992) argues that the Sinhala Qu-morpheme, along with other
clausal-typing morphemes, is a “focus marker” that occurs in the head of a FocP
projection. This FocP notably is not confined to one particular clausal position.
It may occur in the clause-periphery or TP-internally. The focused phrase is the
complement of the Focus head and there is a null operator in the specifier of the
FocP. The null operator is required due to “The Focus-Criterion” given below.
(59) The Focus-Criterion
(a) A Focus operator must be in a Spec head configuration with an X0
[Focus].
(b) An X0 [Focus] must be in a Spec-head configuration with a focus
operator. (Sumangala 1992:162).
This focus operator triggers verbal agreement. Specifically, when the FocP is located
in a TP-internal position, its operator raises to [Spec, CP]. When it passes through
the specifier of IP (Sumangala uses IP instead of TP), it triggers the -e ending on
the verb.
121
Sumangala’s analysis works as follows. (27a), repeated below, is a yes/no construction with a clause-peripheral Qu-morpheme.
(27)
(a) [email protected]@ [email protected]
gaalu [email protected] [email protected]?
Gunapala tomorrow Galle go
Qu
‘Is Gunapala going to Galle tomorrow?’ (Sumangala 1992:131)
Sumangala argues that the structure of this construction is as shown in (60), adapted
from Sumangala (1992:165). The Qu-morpheme is base-generated in the head of a
FocP,25 which is in the clause-periphery. Its complement is the IP and there is a
null operator in its specifier position.
(60)
CP
C’
FocP
Op
C
Foc’
IP
[email protected]@ [email protected] gallu [email protected]
Gunapala tomorrow Galle go
Foc
[email protected]
‘Qu’
Sumangala argues that because the entire IP is the complement of Foc, the entire
IP is focused.
In a construction, such as (61), which is identical to (27a) except that the Qumorpheme appears TP-internally, Sumangala claims that the Qu-morpheme is also
base generated in the head of a FocP projection. However, in this case FocP is in a
TP-internal position.
25
Sumangala uses FP.
122
(61) [email protected]@ [email protected]
gaalu [email protected] yanne?
Gunapala tomorrow Galle Qu go.E
‘Is it to Galle that Gunapala is going tomorrow?’ (Sumangala 1992:163)
The underlying structure of (61) is shown in (62), adapted from Sumangala
(1992:165). The adverbial [email protected] ‘tomorrow’ is left out of the diagram, but presumably it is adjoined to the verb phrase. In this case, because the FocP projection
is TP-internal, Sumangala argues that the null operator in [Spec, FocP] raises to
[Spec, CP] to give the focused phrase scope. Furthermore, movement of the operator
through IP triggers the ‘-E’ ending on the verb.
(62)
CP
Op1
C’
IP
t1
C
I’
VP
[email protected]@
FocP
t1
gaalu
Galle
I
-e
V’
V
yannFoc’ go
Foc
[email protected]
Qu
Sumangala’s analysis differs from mine in several ways. Sumagala’s important
claims are that in Sinhala, the Qu-morpheme is always a Focus-morpheme, a FocP
can appear in various positions, and there is a null operator in the specifier of
123
FocP. First of all, I agree with respect to the Qu-morpheme essentially being a
Focus-morpheme when it occurs TP-internally. However, it is not clear that it is
a Focus-morpheme when it occurs in the clause-periphery, where it merely appears
to serve the function of typing a clause as an interrogative. Under my analysis,
the clause-peripheral Qu-morpheme lacks a Focus-feature. Secondly, Sumangala
argues that a FocP occurs wherever the Qu-morpheme is. However, a fixed FocP
projection in the clause periphery (see discussion of Rizzi (1997, 2001) in chapter
2.4) straightforwardly accounts for a focused phrase obtaining scope, and is in accord
with research on identificational focus (see discussion of Kiss (1998) in section 4.4).
Another aspect of Sumangala’s analysis is that it makes recourse to a null Focus
operator. The idea that movement of the Focus-operator triggers the verbal -E
ending works nicely, but under my analysis there is no need for movement of a
null operator. There is simply movement of a Focus-feature, and this Focus-feature
surfaces as the -E ending. Also, although Sumangala’s analysis accounts for focus
effects, it does not account for the fact that a Qu-morpheme types a clause as a
yes/no construction. Since in non-focus yes/no constructions, the Qu-morpheme
appears in the clause-periphery, and this clause-peripheral position appears to be
the default position for Qu-morphemes in a number of languages, I think that clausal
typing is associated with the clause periphery, and not a TP-internal FocP position.
I account for clausal typing as resulting from Qu-feature movement. Thus, the
advantages of my analysis are as follows: 1) it accounts for the fact that both
clausal typing and identificational focus are associated with a Qu-morpheme, 2)
there is only a single FocP projection in the clausal periphery, and 3) there is no
postulation of null Focus operator movement.
4.4.2 Okinawan
Okinawan is another language that has a TP-internal Qu-morpheme that co-occurs
with verbal agreement, similar to the Sinhala Qu-morpheme.
124
Ga is a Qu-morpheme that occurs in clause-final position in wh-constructions.
Example (63a) is a statement. When the subject John is replaced with the whphrase taa ‘who’ and ga ‘Qu’ is added to clause-final position, a wh-construction
results, as shown in (63b).26
(63) Okinawan:
(a) John-ga
ich-u-n.
John-nom go-pres-ind
‘John is going.’
(b) Taa-ga
ich-u-ga?
who-nom go-pres-Qu
‘Who is going?’ (Miyara 2001:27)
Although ga can appear in clause-final position in a wh-construction, it cannot
appear in this position in a yes/no construction. Instead, the Qu-morpheme -mi is
used. Compare (63a) above with (64). The indicative suffix n of (63a) has been
replaced with mi in (64), resulting in a yes/no construction.
(64) Okinawan:
John-ga
ich-u-mi?
John-nom go-pres-Qu
‘Is John going?’ (Miyara 2001:27)
Basically, the set of features that make-up ga are such that ga is not permitted to
appear clause-finally in a yes/no construction.27
The Qu-morpheme ga, however, can appear in a TP-internal position in a yes/no
construction. In this case, it behaves like the TP-internal Qu-morphemes of Sinhala
and Premodern Japanese. Examples (65a-b) are yes/no constructions in which the
26
27
The nominative case marker ga is a different morpheme from the Qu-morpheme ga.
See chapter 5 for discussion of the use of the Okinawan Qu-morpheme ga in wh-constructions.
125
TP-internal Qu-morpheme ga focuses the subject John-ga ‘John-nom’ and types
the clause as an interrogative. Miyara’s (2001) gloss of (65a) begins with I wonder,
but it ends in a question mark indicating that it is a question, but one with a
focused phrase. Since this is an interrogative construction and John is focused, the
gloss most likely corresponds to ‘Is it John who is going?’. Similarly, the focused
John in (65b) suggests that this construction corresponds to ‘Is it John who bought
a car?’. Also, note that the verbs end in -ra ‘-RA,’ which according to Miyara
(2001), is required when ga appears in a TP-internal position. I take this fact to be
evidence that ra is verbal agreement, similar to the verbal ending that occurs with
a TP-internal Qu-morpheme in Sinhala.28
(65) Okinawan:
(a) John-ga-ga
ich-u-ra?
John-nom-Qu go-pres-RA
‘I wonder if JOHN is going?’ (Miyara 2001:27)
(b) John-ga-ga kurumaa koo-ta-ra?
John-nom-Q car
buy-past-RA
‘Did JOHN buy a car?’ (Miyara 2001:42)
Miyara glosses the Qu-morpheme ga in (65a-b) as an emphasis marker (I gloss
it as ‘Qu’). I take this to be an indication that ga is responsible for focusing the
subjects in these examples. Also, ga is identical in form to the clause final Qumorpheme ga that appears in (63b), which suggests that it is a Qu-morpheme.
According to my analysis, the TP-internal ga ‘Qu’ contains a Focus-feature.
When ga is Merged in a TP-internal position, it focuses a phrase and its Focus- and
Qu-features raise to Foc and Typ, respectively. For example, (65a) has the structure
in (66). I assume that the present tense ending -u is in T. Also, although I leave the
28
Miyara glosses ga in this case as an emphasis marker and -ra as a Qu-morpheme. Since ga
appears to be responsible for typing a clause as a yes/no construction and -ra only appears when
-ga is in a TP-internal position, I assume that ga is a Qu-morpheme and -ra is verbal agreement.
126
verb in v , it is possible that it moves to T. Movement of the Qu- and Focus-features
is motivated by EPP features.
(66)
TypP
Typ’
FocP
TypEP P
FQu2
Foc’
TP
T’
FocEP P
-ra[FF oc1]
RA
v’
T
-u
-pres
vP
DP
DP
John-ga
John-nom
-ga+t[FF oc1,FQu2 ]
v
ich
go
The Qu-morpheme adjoins onto the subject within the [Spec, v P] position. The
Focus- and Qu-features feature raises to Foc, where the Focus-feature is pronounced
as the ‘-RA’ ending, and the Qu-feature moves to Typ, where it types the clause as
an interrogative.
When the Qu-morpheme in a yes/no construction lacks a Focus-feature, it simply
surfaces in clause-final position as mi. In this case, there is no special verbal ending.
The Qu-morpheme is Merged directly in Typ, as shown in the following simplified
diagram of (64).
127
(67)
TypP
Typ’
TP
Typ
mi[FQu ]
John-ga ich-u
John-nom go-pres
Okinawan thus is a language in which the form of the Qu-morpheme that occurs
in a yes/no construction differs depending on whether or not it has a Focus-feature.
The default Qu-morpheme is Merged directly in Typ and surfaces as mi. The
Qu-morpheme with a Focus-feature surfaces as ga and is Merged in a TP-internal
position adjoined to a focused phrase.
4.4.3 Premodern Japanese
In Premodern Japanese, unlike in Modern Japanese (see section 4.3.1), a Qumorpheme can also occur both in TP-internal and clause-peripheral positions. The
TP-internal Qu-morpheme, like the TP-internal Qu-morphemes of Sinhala and Okinawan, focuses an adjacent phrase. According to my analysis, Premodern Japanese
then contains two types of Qu-morphemes: one with a Focus-feature that is Merged
TP-internally and one without a Focus-feature that is Merged directly in Typ.
There are two Qu-morphemes in Premodern Japanese that can appear in a TPinternal position: ya and ka. In (68), the Qu-morpheme ya occurs in a TP-internal
position attached to the phrase ama-to ‘fisherman-as.’ In (68), the verb mira-mu
‘see’ appears with an adnominal ending, which occurs when the Qu-morpheme is
in a TP-internal position. This type of verbal ending is referred to as musubi in
traditional Japanese grammar, and I gloss it as ‘M’ following Hagstrom (1998).
128
(68) Shirotahe-no Fujie-no ura-ni
izari-suru ama-to-ya
epithet-gen Fujie-gen beach-LOC fishing-do fisherman-as-Qu
mira-mu tabi-yuku ware-wo?
see-M
travel
me-acc
‘Will people take me on a trip to be a fisherman who is fishing in the gulf or
Fujie?’ (Man’youshuu no. 3607, per Watanabe 2002:188)
I assume that ama-to ‘fisherman-as’ is focused based on Watanabe’s (2002) claim
that “the focus of yes/no questions is marked by ya (180).” Thus (68) corresponds
to a rather complex, and in English somewhat marginal, clefted yes/no construction
such as ‘Is it a fisherman who is fishing in the gulf or in Fujie that the people will
take me on a trip to be?’. In (69) below, the Qu-morpheme ya follows the TPinternal nominal phrase tatu no kubi no tama ‘dragon gen head gen gem’ and it
types the clause as an interrogative. The verb torite ‘take’ appears with the musubi
endings -i-te and the verb oFas-i-tar-u ‘come’ appears with the musubi endings -i
and -u.29
(69) Ofotomo no
Dainagon Fa [tatu
no
kubi no
tama ya]
Otomo gen Councillor top dragon gen head gen gem Qu
tor-i-te
oFas-i-tar-u?
take-M-M come.HON-M-perf-M
‘Did Otomo-no-Dianagon get the gems on the dragon’s head?’ (Taketori
Monogatari : 859, per Whitman 1997:161)
This construction presumably corresponds to ‘Was it the gems on the dragon’s head
that Otomo-no-Dainagon got?’. Similarly, the Premodern Japanese Qu-morpheme
ka, which unlike ya retains its use as a Qu-morpheme in Modern Japanese, can
also appear next to a focused phrase in a TP-internal position. In (70), ka directly
follows the TP-internal nominal phrase atamitaru tora ‘irritated tiger.’ Note that
Watanabe’s clefted gloss indicates that atamitaru tora ‘irritated tiger’ is focused.
Also, the verb hoyuru has a musubi ending ru.
29
It is interesting that the musubi endings appear on both verbs. Why exactly this would be
the case is not clear to me.
129
(70) . . .
atamitaru tora-ka hoyuru? . . .
irritated tiger-Q roar-M
‘Is it an irritated tiger that is roaring?’ (Man’youshuu #199, per Watanabe
2002:180)
These Qu-morphemes can also appear in clause-final position. In (71), ya occurs
in clause-final position where it follows the non-musubi conclusive form of the verb;
the verb sira-zu ‘know’ ends in the conclusive form zu.
(71) Ware wo ba
sira-zu
ya?
me
acc top know-not.conc Qu
‘Don’t you know me?’ (Ise Monogatari, per Whitman 1997:165)
(72) shows that ka can also occur in clause-final position to form a yes/no construction. Unlike ya, the Qu-morpheme ka follows a verb with a musubi ending;
Faradati-tamaF-er-u ‘squabble-HON-perf-M’ ends with the musubi ending -u.
(72) WaraFabe-to Faradati-tamaF-er-u
ka?
children-with squabble-HON-perf-M Qu
‘Have you been fighting with the children?’ (Genji Monogatari, per Whitman
1997:165)
The Premodern Japanese facts thus are as follows. There are two Qu-morphemes,
ka and ya, that can either appear in the clause-periphery or TP-internally. When
they occur TP-internally, they focus an associated phrase and the verb shows up
with a musubi ending. When ya appears clause-finally, it does not trigger the musubi
ending on the verb, but the clause-final ka does.
Just as in Sinhala, the clause-final Qu-morpheme marks a yes/no construction
and the TP-internal Qu-morpheme both marks a yes/no construction and focuses
a phrase. Thus, my proposal that the clause-final Qu-morpheme lacks a Focusfeature and the TP-internal Qu-morpheme contains a Focus-feature accounts for
the data. The clause-final Qu-morpheme is Merged directly in Typ, where its Qufeature types a clause as a yes/no construction and the TP-internal Qu-morpheme
130
adjoins to a focused phrase. Its Qu-feature raises to Typ where it types a clause and
its Focus-feature raises to Foc, where it focuses the relevant TP-internal phrase.
The facts regarding the musubi verbal endings though, are not so clear. When
both ya and ka occur TP-internally, they trigger the musubi endings on the verbs.
As I argued was the case in Sinhala, it is likely the case that movement of the
Focus-feature to Foc surfaces as the musubi ending. However, a problem for this
analysis is what occurs when these Qu-morphemes appear in clause-final position.
When ya occurs in clause-final position, it does not occur with the musubi ending.
This would be expected if the musubi ending is a Focus-feature, since the clausefinal Qu-morpheme lacks a Focus-feature. However, when ka occurs clause-finally,
unlike ya, it triggers the verbal ending. Thus, regardless of whether ka occurs TPinternally or clause-peripherally, it triggers the musubi ending. The fact that the
clause-peripheral Qu-morpheme triggers the musubi ending could be evidence that
this ending is not actually a Focus-feature. Another possibility is that the clause
peripheral ka actually originates in a TP-internal position adjoined to a focused
phrase, from where it undergoes movement to the clause-pheriphery, and in the
process leaves a Focus-feature in Foc that surfaces as the musubi ending.
In summary, the Premodern Japanese facts suggest the presence of two types of
Qu-morphemes. One type contains a Focus-feature and is Merged in a TP-internal
position adjacent to a focused phrase. The Focus- and Qu-features raise to Foc and
Typ, where they focus a phrase and type a clause. The other type of Qu-morpheme
is Merged directly in Typ and does not focus any TP-internal element. Although
there is a relationship between these Qu-morphemes and musubi verbal endings, it is
not clear whether or not feature movement triggers these endings, since a clause-final
ka can trigger them. This issue requires further analysis.
131
4.4.4 Tupı́
The Tupı́ languages30 of Sateré-Mawé and Assurinı́ have the same phenomenon as
Sinhala, Okinawan, and Premodern Japanese, whereby a Qu-morpheme can appear
adjacent to a TP-internal phrase. According to Brandon & Seki (1984:83), in yes/no
constructions in Sateré-Mawé and Assurinı́, “the question particle may be placed
after any major phrase (NP, VP, or PP) (83).”
The following examples in (73) are from Sateré-Mawé. In (73a-b), the Qumorpheme apo occurs in clause-final position after a subject and an adverb, respectively, and in (73c), it occurs TP-internally after a verb.
(73) Sateré-Mawé:
(a) i-kahu
rakat u’i
apo?
3-beautiful nom flour Qu
‘Is the flour good?’ (i.e., a good thing)
(b) Ere-ket kahato apo?
2s-sleep very
Qu
‘Do you sleep well?’
(c) Etu-nug kahu
teran apo ui-yat?
2s-make beautiful want Qu 1s-house
‘Do you want to fix my house?’ (Summer Institute of Linguistics 1978,
per Brandon & Seki 1984:83-84)
Similarly, the Qu-morpheme pa can appear in various positions in Assurinı́, as
shown in (74). In (74a-b) it occurs in clause-final position after a verb. In (74c-d)
it appears TP-internally, after a verb and an object, respectively.
(74) Assurinı́:
(a) Karoa o-ata a-ha pa?
Karoa 3-hunt 3-go Qu
‘Is Karoa hunting?’
30
These are languages spoken in Central and South America (Cable 2007:189).
132
(b) O-ma’esiroa oe-raha pa?
3-things
3-take Qu
‘Did she take her own things?’
(c) n-a-ha-ihi
pa ne-mena
ka’a-pe?
NEG-3-go-NEG Qu 2s-husband jungle-to
‘Didn’t your husband go to the jungle?’
(d) Itasoa pa ere-reke?
triangle Qu 2s-have
‘Do you have a triangle?’(Nicholson 1978, per Brandon & Seki 1984:84)
According to my analysis, in a TP-internal position, the Qu-morpheme contains
a Focus-feature that forces it to adjoin to a focused element. Note that it is not clear
from the discussion provided by Brandon & Seki (1984) whether or not the phrases
that the Qu-morpheme attaches onto are focused, but if my analysis is correct,
then I predict this to be the case. The TP-internal Qu-morpheme adjoins to a
focused-phrase and its Focus- and Qu-features raise to Foc and Typ, respectively.
The clause-final Qu-morpheme lacks a Focus-feature and is Merged directly in Typ.
I have not been able to confirm whether or not there is verbal agreement that cooccurs with the Qu-morpheme, but this is an issue that is worthy of further analysis.
4.4.5 Ewen
Ewen (Lamut)31 has a Qu-morpheme gu that can appear either in clause-final or
in TP-internal position. Whitman (1997:174) writes that gu ‘Qu’ “marks the focus
element in yes/no questions (Benzing 1955:111) internally to the clause; it also marks
the clause type in clause-final position.”
Example (75) shows gu ‘Qu’ in clause-final position. More specifically, there are
two Qu-morphemes, each associated with a separate clause. The first Qu-morpheme
is attached onto the verb hør-jin ‘go-fut’ and the second Qu-morpheme is attached
onto @[email protected] ‘not-fut.’
31
This is a Tungusic language spoken in Siberia (Vovin 2006:143).
133
(75) Ewen:
Aman-si timina
hør-jin-gu, @[email protected]?
father-2S tomorrow go-fut-Q not-fut-Qu
‘Is your father going tomorrow or not?’ (Robbek 1989:148, per Whitman
1997:174)
Just as not in the English gloss represents an elided clause corresponding to Is
your father not going tomorrow, the same is likely the case in (75). There is a biclausal structure with gu ‘Qu’ in the final position of each clause. This clause-final
Qu-morpheme is most likely Merged directly in Typ where it types a clause as an
interrogative.
In (76), the Qu-morpheme is attached onto the TP-internal phrase asal ‘women.’
(76) Ewen:
Asal-gu olra-w
[email protected]
women-Q fish-acc catch-3PL
‘Do the women catch fish?’ ((Benzing 1955:121, per Whitman 1997:173)
I assume that this phrase is focused based on Whitman’s statement that the Qumorpheme marks a “focus element (174),” in which case (76) actually corresponds to
‘Is it the women who catch fish?’ in English. The Qu-morpheme then is focusing a
phrase and typing a clause, indicating that the TP-internal Qu-morpheme contains
a Focus-feature.
Although the data are sparse, gu ‘Qu’ appears to have two versions, as in the
other languages discussed in this section. One form has a Focus-feature and occurs
TP-internally, and the other form lacks a Focus-feature and occurs in Typ. Unlike
in Sinhala, Premodern Japanese, and Okinawan, there is no clear verbal marking
that co-occurs with this Qu-morpheme (Whitman 1997:174). Feature movement in
Ewen then does not appear to trigger any special verbal endings.
134
4.4.6 Summary
In this section, I have examined languages which allow a Qu-morpheme to either
appear in a TP-internal position or a clause-final position. I have argued that a
clause-final Qu-morpheme is Merged directly in Typ. A TP-internal Qu-morpheme
contains a Focus-feature that forces it to be Merged in a TP-internal position adjoined to a focused phrase. From this position its Qu- and Focus-features undergo
movement to the clause periphery. Note that all of the languages discussed in this
section allow both a Qu-morpheme with and without a Focus-feature. The evidence
here suggests that if a language has a Qu-morpheme that contains a Focus-feature, it
also contains a Qu-morpheme that lacks a Focus-feature. On the other hand, there
could be languages in which a Qu-morpheme must always contain a Focus-feature.
Further investigation of this issue is required.
4.5 Conclusion
In this chapter, I have argued that cross-linguistic variation in yes/no constructions
results from properties of Qu-morphemes. Specifically, variation depends on whether
or not a Qu-morpheme is overt or covert, or an affix or an independent lexical
item. I argued that these differences result from [±OVERT] and [±AFFIX] features.
Furthermore, the formation of a yes/no construction is dependent on where a Qumorpheme is Merged into a derivation. By default, a Qu-morpheme is Merged
directly in Typ, since it is most economical for the Qu-morpheme to be Merged
directly in the position where its Qu-feature can type a clause, presumably by valuing
an uninterpretable clausal typing feature in Typ. However, some Qu-morphemes
in certain languages contain a Focus-feature which forces the Qu-morpheme to be
Merged in a TP-internal position, from where its Qu- and Focus-features undergo
movement.
In the next chapter, I examine the positions of Qu-morphemes in wh-
135
constructions. I show that the analysis of clause-peripheral and TP-internal Qumorphemes developed in this chapter can be applied straightforwardly to account
for the positions of Qu-morphemes in wh-constructions.
136
CHAPTER 5
Qu-morphemes in wh-constructions
5.1 Introduction
As discussed in chapter 4, in a yes/no construction, a Qu-morpheme either must
appear in a clause-peripheral position or in a TP-internal position, and I took the
position that TP-internal Merge is motivated by a Focus-feature contained within
a Qu-morpheme. In this chapter, I extend the analysis of the positions of Qumorphemes in yes/no constructions to account for the positions of Qu-morphemes
in wh-constructions. The essential facts regarding yes/no constructions are identical,
except that in a wh-construction, when a Qu-morpheme appears in a TP-internal
position, it is adjoined to a TP-internal wh-phrase. Specifically, a Qu-morpheme
with a Focus-feature adjoins to a wh-constituent. As in a yes/no construction, in
a wh-construction, a Qu-morpheme by default is Merged directly in Typ, unless a
Focus-feature forces it to be Merged TP-internally. In this chapter, I account for
the position of the Qu-morpheme as being determined by whether or not it contains
a Focus-feature.
The possible positions of a Qu-morpheme are represented in (1a-c), which represents a language in which there is a clausal typing projection that is head-final at
Spell-Out.
(1)
(a) [T ypP . . . wh-phrase Qu . . . ]
(b) [T ypP . . . [T ypP . . . wh-phrase . . . ] Qu . . . ]
(c) [T ypP . . . wh-phrase . . . Qu ]
When the Qu-morpheme is in a TP-internal position, it either appears adjacent to a
wh-phrase, as in (1a), or adjacent to an embedded clause containing a wh-phrase, as
137
in (1b). The other option is for the Qu-morpheme to appear in the clause periphery
as in (1c).
Hagstrom (1998) argues that in a wh-construction in wh-in-situ languages such as
Sinhala, Okinawan, Premodern Japanese, and Modern Japanese, a Qu-morpheme
is base generated in a position adjacent to a wh-phrase. From this position the
Qu-morpheme moves to C. This movement can be overt, in which case the Qumorpheme must appear in the clause periphery, as in (1a) above, or this movement
can be at LF, in which case the Qu-morpheme appears in a TP-internal position as
in (1b) and (1c). I follow Hagstrom’s view that when a Qu-morpheme appears in
a TP-internal position, there is movement from the position of the Qu-morpheme.
However, whereas Hagstrom argues that there is LF movement of a Qu-morpheme,
I argue that there is movement of Qu- and Focus-features. Furthermore, unlike
Hagstrom, I take the position that a clause-final Qu-morpheme is base-generated in
Typ, and does not move to the clause periphery.
It is generally assumed in the literature that a wh-phrase receives focus. Karimi
& Taleghani (2007:173), following work by Rochemont (1978, 1986), Horvath (1986),
Bresnan & Mchombo (1987), and Kiss (1998), among others, claim “that a whelement is inherently focused.” This inherent focus likely corresponds to Kiss’ notion
of an information focus (see chapter 4.4), since a wh-phrase is emphasized and it
refers to new information, as the value of the wh-phrase is new information. In the
following discussion, a wh-phrase is by default assumed to be an information focus.
However, when a wh-phrase appears with an adjacent Qu-morpheme, it receives ‘extra focus.’ In this case, I assume that it functions as an identificational focus, in the
same manner that I argued that a TP-internal clause with an adjacent Qu-morpheme
is an identificational focus in chapter 4.4. The wh-phrase receives a special focus
interpretation, comparable to a cleft in English, and it refers to an exhaustive set.
Therefore, it corresponds to Kiss’s (1998) definition of an identificational focus (see
chapter 4.4).
138
The organization of this chapter is as follows. In section 5.2, I discuss Mandarin
and Japanese, which are languages in which a Qu-morpheme must appear in a fixed
clause-peripheral position. In section 5.3, I discuss Qu-morphemes in English, which
also must appear in the clause periphery. I examine the interesting asymmetries in
English that exist between subject and non-subject matrix wh-constructions and
between matrix and embedded wh-constructions. I also account for some dialect
differences found in the formation of wh-constructions in English. In section 5.4, I
examine wh-constructions in which a Qu-morpheme can appear either in a clauseperipheral or a TP-internal position. Where the Qu-morpheme occurs is dependent
on whether or not it contains a Focus-feature. In section 5.5, I discuss the fact that
although a Qu-morpheme can appear in a TP-internal position in languages such as
Sinhala and Okinawan, when it occurs in certain embedded clauses, there are island
effects due to the blocking of movement of Qu- and Focus-features. In section 5.6, I
examine some problematic constructions in Sinhala in which a Qu-morpheme must
appear in a fixed TP-internal position. In section 5.7, I compare my analysis of the
facts concerning Qu-morphemes in wh-constructions with some previous analyses.
Section 5.8 is the conclusion.
5.2 Clause-peripheral Qu-morphemes
As with the formation of yes/no constructions, in wh-constructions, a Qu-morpheme
generally appears in a clause-peripheral position.
5.2.1 Mandarin
Mandarin contains a Qu-morpheme ne that can appear in a wh-construction.1 It is
optional in matrix wh-constructions, and it is absent in embedded wh-constructions
(Cheng 1997:26).2
1
2
There is a different yes/no Qu-morpheme ma.
I have added the parenthesis around ne to indicate that it is optional.
139
(2)
Mandarin:
Qiaofeng mai-le
shenme (ne)?
Qiaofeng buy-asp what
Qu
‘What did Qiaofeng buy?’ (Cheng 1997:22)
As in chapter 4, the position that I take is that when an overt Qu-morpheme is
absent, a null Qu-morpheme is present, and so when ne is absent, a null version
occurs.3
In Mandarin, the Qu-morpheme cannot occur in a TP-internal position. For
example, the Mandarin example (3) is ill-formed if ne does not appear in clausefinal position, as shown below.
(3)
Mandarin:
Qiaofeng (*ne) mai-le
(*ne) shenme (ne)?
Qiaofeng
buy-asp
what
Qu
‘What did Qiaofeng buy?’ (Sunjing Ji, p.c.)
These facts can be accounted for straightforwardly if the Qu-morpheme is Merged
directly in Typ, assuming that Typ appears in a clause-final position in Mandarin,
which is an SVO language that appears to otherwise have head-initial projections
(cf. Huang (1982), among others).
5.2.2 Japanese
The Japanese Qu-morpheme ka is also confined to clause-final position in whconstructions. It cannot appear TP-internally, as shown in (4). The parentheses
around ka indicate that it can be dropped; i.e. there is a null version of the Qumorpheme ka.
3
See Cheng (1997) for arguments supporting the claim that a null Qu-morpheme is present in
Mandarin wh-constructions.
140
(4)
Japanese:
Hideya-wa (*ka) nani-o
(*ka) kaimashita (ka)?
Hideya-top
what-acc
bought
Qu.
‘What did Hideya buy?’
Evidence in Japanese clearly shows that in a wh-construction, the Qu-morpheme
can only occur in the clause in which a wh-phrase has scope. Examples (5a-b)
demonstrate this.
(5)
Japanese:
[dare-ga erab-areta ka] tutaete-imasen.
(a) Sinbun-wa
newspapers-top who-nom elected-was Qu report-is not
‘Newspapers do not report who was elected.’
(b) Sinbun-wa
[dare-ga erabare-ta to] tutaete-imasu ka?
newspapers-top who-nom elected-was that report-is
Qu
‘Who1 do the newspapers report t1 was elected?’ (Nishigauchi 1990:9)
In (5a) the Qu-morpheme is in the embedded clause, which it types as an interrogative and the wh-phrase has embedded scope. In (5b), the Qu-morpheme types the
matrix clause as an interrogative, and therefore the wh-phrase must have matrix
scope. This latter example shows that a wh-phrase in an embedded clause can have
matrix scope as long as there is a Qu-morpheme at the edge of the matrix clause.
These facts are accounted for straightforwardly if the Qu-morpheme is Merged directly in Typ of the clause where it has scope.
Furthermore, note that whereas a matrix Qu-morpheme can be dropped in
Japanese, as in (4) above, an embedded clause Qu-morpheme cannot be dropped.
For example, the Qu-morpheme ka in the embedded clause in (5a) must be present.
In a matrix clause, when a Qu-morpheme is dropped, special intonation occurs; the
clause ends with a rise. One possibility is that this interrogative intonation is not
allowed in an embedded clause, and thus the Qu-morpheme is required.4
4
I thank Heidi Harley (p.c.) for suggesting this to me.
141
5.2.3 Summary
The positions of the Qu-morphemes discussed in this section have been straightforward. Languages such as Mandarin and Japanese contain a Qu-morpheme with
a Qu-feature that is simply Merged in the optimal position in Typ, where its Qufeature is able to directly type a clause.
5.3 English
The English Qu-morpheme also must appear in the periphery of a wh-construction.
In this section, I show that the English Qu-morpheme is Merged directly in Typ of
the clause in which it has scope. However, there is a certain degree of complexity
involved in the formation of wh-constructions in English. First of all, in a matrix
clause do-support can occur when there is movement of a non-subject wh-phrase, but
do-support cannot occur when there is movement of a subject wh-phrase. Also, in
a matrix clause a tense element must move, but this is not allowed in an embedded
clause in ‘standard English,’ although some other dialects of English allow tense
movement. I account for these facts in terms of the properties of the English Qumorpheme.
5.3.1 The standard American English facts
Below I discuss the facts regarding the formation of wh-constructions in standard
American English.5 These facts do not hold of all dialects of English, an issue that
I return to.
As with English yes/no constructions (see chapter 4.3.2), a tense element moves
in a wh-construction. This can be seen clearly when a non-subject wh-phrase undergoes wh-movement. Compare the statement in (6a) with the corresponding wh5
I use ‘standard’ to refer loosely to the prominent dialects of English spoken in the United
States.
142
construction in (6b).
(6)
(a) Reginald has eaten dinner.
(b) What has Reginald eaten?
Whereas the auxiliary has follows the subject Reginald in the statement in (6a), it
precedes the subject in the wh-construction (6b), which is evidence that the auxiliary
has moved. When there is no overt auxiliary a form of do is inserted, as in (7).
(7)
What did she eat?
The facts become more complex when a wh-construction is formed with a fronted
subject wh-phrase. Compare (8a) and (8b).
(8)
(a) John has eaten dinner.
(b) Who has eaten dinner?
Note that unlike in (6a-b), where it can be seen that the auxiliary moves, there is no
clear auxiliary movement in (8). In both the statement in (8a) and the corresponding
wh-construction (8b), the auxiliary follows the subject. One possibility, which I
argue is the case below, is that the auxiliary has moves, but that the wh-phrasal
subject moves to a position higher than the auxiliary.
Examples (9a-b) are well-formed wh-constructions with subject wh-phrases.
Compare these with (7) above, which contains did.
(9)
(a) Who ate dinner?
(b) What lives in forests?
Furthermore, a form of the dummy auxiliary do cannot occur. Although (10a-b)
are well-formed as emphatic wh-constructions when did receives special stress, they
are not well-formed as standard non-emphatic wh-constructions.6
6
I assume that the fact that the emphatic interpretations requires do-support, most likely in
Foc, is unrelated to the formation of a wh-construction.
143
(10)
(a) *Who did eat dinner?
(b) *What did live in forests?
In an embedded wh-construction, there does not appear to be auxiliary movement. For example, when ‘What did she eat?’ occurs in an embedded clause, it
surfaces as (11a) which lacks an auxiliary, and not as (11b).
(11)
(a) I wonder [what she ate].
(b) *I wonder [what did she eat].
I next turn to an analysis of these facts. In chapter 4, I argued that an English
matrix yes/no construction contains a null [+AFFIX] Qu-morpheme. The same is
true in a wh-construction. There is a null [+AFFIX] Qu-morpheme in Typ, which
is responsible for giving a clause an interrogative interpretation. The [+AFFIX]
feature of the Qu-morpheme attracts the tense element to Typ. Furthermore, I
assume that there is wh-phrasal movement to [Spec, TypP]. Here, I focus on the
Qu-morpheme, and although extremely important, I leave aside discussion of whmovement until chapter 6.
This analysis acccounts straightforwardly for constructions such as (6b) and (8b),
repeated below as (12a-b).
(12)
(a) What has Reginald eaten?
(b) Who has eaten dinner?
The null [+AFFIX] Qu-morpheme in Typ attracts the Tense element. In these
constructions, T contains the auxiliary has, which raises to Typ. Diagrams of (12ab) are shown below in (13) and (14), respectively.
144
(13)
TypP
DP1
what
Typ’
Typ
∅FQu +has2
TP
DP3
Reginald
T’
T
t2
vP
t3
v’
v
eaten
t1
(14)
TypP
DP1
who
Typ’
Typ
∅FQu +has2
TP
t1
T’
T
t2
vP
t1
v’
v
eaten
DP
dinner
145
As noted above, the facts concerning when do-support is allowed are complicated.
When a non-subject wh-phrase fronts to clause-initial position, a form of do must
occur if there is no other auxiliary element, as shown in (7), repeated below as (15a).
However, when a subject wh-phrase fronts to clause-initial position, a form of do
does not occur even if there is no auxiliary element, as in (9a), repeated below as
(15b).
(15)
(a) What did she eat?
(b) Who ate dinner?
This subject-object asymmetry, whereby a non-subject wh-phrase requires dosupport, but a subject wh-phrase does not thus requires an explanation.
5.3.2 Accounts of the subject/non-subject wh-construction assymmetry
There have been a number of analyses that have examined the asymmetry between
subject and non-subject wh-phrases in English. In this subsection, I briefly review
these analyses, before I present my own analysis in the following subsection.
Within the framework of Government and Binding Theory, Koopman (1983,
1984) proposes that the Empty Category Principle (ECP) accounts for the English wh-subject extraction facts. According to Koopman (1983), a wh-subject that
fronts to clause-initial position, [Spec, CP], must antecedent govern its trace. This
antecedent government is blocked if an overt tense element appears in C. An object
wh-phrase can be followed by an overt tense element in C because it need not be
antecedent governed, since it is lexically governed by a verb. One problem with
this proposal is that it requires parametrization. Specifically, a tense element does
not appear to block antecedent government in languages such as Dutch. A fronted
wh-subject can be directly followed by a moved auxiliary, as in (16) below.
146
(16) Dutch:
Wie1 heeft2 t1 hem/Jan gezien t2 ?
who has
him/John seen
‘Who saw him/John?’ (Koopman 1983:349)
Another problem with this analysis is that the notion of government and the ECP
do not exist within the framework of the Minimalist Program.
Agbayani (2000) takes a different view of the subject wh-extraction facts; he
relies on the notion that a subject wh-phrase remains in [Spec, TP]. Following
George (1980) and Chomsky (1986), Agbayani argues for the Vacuous Movement
Hypothesis, which is “the notion that in English local overt wh-movement takes place
except for subjects (703).” Agbayani’s proposal relies on the notion that feature
movement and phrasal (XP) movement (which he refers to as ‘category movement’)
procede separately. Agbayani (2000:707) writes the following.
I propose that Universal Grammar allows feature attraction and category
movement to apply separately. Attract F adjoins a set of formal features
(F) to an attracting head. A second operation, Move Cat(egory), raises
the category to a specifier position where it is in a local relation with its
formal features adjoined to the attracting head.
According to this proposal, a feature and its category have to be adjacent, where
adjacency is defined as follows.
(17) X and Y are adjacent if no elements that are visible at the interface
intervene between X and Y (Agbayani 2000:707)
Agbayani follows Chomsky’s (1995b) view that elements that are visible at the interface are “phonological features and XP- and X0 - level categories, but not segments
or X’-level categories (Agbayani 2000:707).”
147
With these assumptions, Agbayani argues that the following occurs in an English
wh-construction. A wh-feature of a non-subject wh-phrase adjoins to C. It is not
adjacent to its category (the wh-phrase that it originated in) because there is visible
material intervening between the feature and the wh-phrase. Therefore, the whphrase must raise to [Spec, CP]. When a wh-feature of a subject wh-phrase raises to
C, it is adjacent to its category in [Spec, TP]. There are no visible elements between
the feature and the subject. Thus, the subject need not move. It can remain in
[Spec, TP] and be adjacent to its feature.
This analysis accounts for the subject/non-subject asymmetry as resulting from
whether or not a wh-phrase moves to [Spec, CP]. However, as I understand it, it does
not account for T to C movement in non-subject wh-constructions. Wh-movement
in a matrix clause requires T to C movement, but why exactly it requires T to C
movement is not clear. It is also not clear why there is no T to C movement in an
embedded clause.
Pesetsky & Torrego (2000) provide an analysis that relies on the notion that the
English matrix C has an uninterpretable Tense feature, uT, which has an “EPP
property” (360). This uT feature needs to be deleted via movement of another uT
into C. Specifically, the uT on C forms an Agree relation with another uT feature
which then moves to C. Movement, they argue, is driven by the EPP property that
is contained by the uT feature on C. Under this analysis, the EPP is referred to as
a “subfeature of a feature” that “is a property of a feature of a head-not a property
of the head itself (359).” Pesetsky and Torrego assume that nominative case, which
a subject contains, is a uT feature that can delete a uT feature in C.
Pesetsky & Torrego argue that when a wh-phrasal subject moves to [Spec, CP]
to check an uninterpretable wh-feature, its nominative case feature deletes the uT
on C. For example, the structure of (18a) with a fronted subject wh-phrase is that
in (18b).
(18)
(a) Who bought the book?
148
(b) [C, uT, uWh] [T P [Who, uT] T [V P bought the book]] (Pesetsky &
Torrego 2000:363)
The wh-subject moves to [Spec, CP] to check the uWh feature. Also, the subject
contains a uT feature because it has nominative case. Therefore, the subject deletes
the uT feature on C. Since the uT feature on C is deleted, there is no need for
movement of a tense element to C.
When a non-subject wh-phrase moves into [Spec, CP], the uT feature on C
remains unchecked, since a non-subject lacks nominative case. T contains a uT
feature (nominative case) which it obtains from movement of a subject into its
specifier. T then must move to C to delete the uT feature. For example, the whconstruction (19a) with a fronted object wh-phrase has the underlying structure in
(19b).
(19)
(a) What did Mary buy?
(b) [C, uT, uWh] [T P [Mary, uT] T [V P bought what]] (Pesetsky & Torrego
2000:363)
The wh-phrase what moves to [Spec, CP] to check the uninterpretable wh-feature
uWH on C. Then T moves to C to check the uninterpretable Tense feature uT. In this
manner, when a non-subject wh-phrase undergoes wh-movement, “C must delete its
uWh and its uT in two separate operations (Pesetsky & Torrego 2000:363).”
Pesetsky and Torrego argue that in embedded wh-constructions, there is no T to
C movement because the uT feature on C does not contain an EPP feature. They
write that the“[s]tandard English embedded interrogative C lacks the EPP property
for uT (380).” For example, (20a) is well-formed because T has not been attracted
to C. (20b) is ill-formed because there is no EPP feature on C to attract the moved
tense element did.
(20)
(a) Bill asked what Mary bought.
149
(b) *Bill asked [what did Mary buy]. (Pesetsky & Torrego 2000:378)
In an embedded wh-construction, because the EPP feature does not attract a uT
feature, Pesetsky & Torrego claim that the uT feature is satisfied via an Agree
relation with T. In other dialects that allow T to C movement in an embedded
clause, such as Hiberno English, the embedded C can contain an EPP feature.
Pesetsky and Torrego’s analysis accounts nicely for the data, but there is a
great deal of complexity involved with the notion that C contains a uT feature that
requires nominative case. It is not clear why C would require nominative case, as
nominative case is generally associated with T.
Bobaljik (1994) provides an analysis that I find to be most appealing due to
its simplicity. This analysis relies on the notion that do-support is dependent on
whether or not tense and the verb are adjacent to each other. Bobaljik argues that
the do-support facts result from the need for tense to be adjacent to the verb in
cases in which there is no auxiliary that houses tense. According to Bobaljik, in
English, when tense and a verb are adjacent, even if they are in different projections,
tense shows up on the verb. When tense is not adjacent to a verbal element, then
do-support is required to house tense. Bobaljik follows Halle & Marantz’s (1993)
view that in a simple declarative sentence in English “inflectional affixes and verbal
heads may merge under (some form of) adjacency (Bobaljik 1994:5).” When an
inflectional affix and a verb are adjacent, then the affix can attach onto the verb,
even if they are in different projections. For example, Bobaljik proposes that in
the configuration in (21), the heads X and Y are adjacent. The trace intervening
between these heads does not block adjacency.
(21) . . . X [Y P trace [Y ′ Y . . . (Bobaljik 1994:2)
According to my understanding of Bobaljik’s analysis, a matrix non-subject whconstruction such as (22a) has the structure in (22b).
(22)
(a) When did Sam eat the horseradish? (Bobaljik 1994:7)
150
(b) [CP When do+INFLP AST 1 [IP Sam t1 [V P eat the horseradish]]]?
An inflectional head INFL raises to C. The subject Sam intervenes between INFL
and the verb eat, thereby preventing INFL from affixing onto eat. Thus, a form of
do must be inserted in C to house INFL.
When a subject wh-phrase moves, the results differ notably in that the adjacency
relation between INFL and the verb is not disrupted. This can be seen in (23a-b).
(23)
(a) Who ate my horseradish? (Bobaljik 1994:7)
(b) [CP Who1 INFLP AST 2 [IP t1 t2 [V P eat my horseradish]]]?
(23b) shows the underlying structure of (23a). The INFL head raises to C and the
subject who raises to [Spec, CP]. Since there is no intervening subject, adjacency
is established between INFL in C and the verb in the VP. Therefore, INFL is able
to show up on the verb. The trace of the subject does not block adjacency, and so
there is no need for insertion of do.
5.3.3 Analysis of English facts
With a slight reformulation to fit with my proposals, Bobaljik’s analysis accounts
for the wh-construction facts as follows. When a non-subject wh-phrase raises to
[Spec, TypP], adjacency between the Tense element which has moved to Typ and
v is blocked by the intervening subject. Thus, if there is no auxiliary to house the
tense element, a form of do is inserted (do-support). Below is a diagram of (15a).
The wh-phrase raises to [Spec, TypP] and the tense element raises to Typ. Since
Typ is not adjacent to v , due to the intervening subject she, a form of do is inserted
to house the tense element.7
7
Note that in a matrix yes/no construction, there always will be a subject intervening between
the Qu-morpheme and T. Thus, do-support is always required when there is no auxiliary that
carries tense.
151
(24)
TypP
DP1
What
Typ’
Typ
∅FQu +did2
TP
DP
she3
T’
T
t2
vP
t3
v’
v
eat
DP
t1
When a subject wh-phrase moves, there is no element intervening between the
tense element in Typ and v (the trace of the subject does not count as an intervener).
Therefore, tense is able to show up on the matrix verb in cases in which there is
no auxiliary. A diagram of (15b) is shown below. Adjacency is indicated by the
double-headed arrow between Typ and v .
152
(25)
TypP
DP1
Typ’
Who
Typ
∅FQu +∅P AST 2
TP
t1
T’
T
t2
vP
t1
v’
v
ate
DP
dinner
I next turn to embedded wh-constructions. Remember that in these constructions there is no movement of a tense element, as can be seen in (11a-b), repeated
below. Example (11b) is ill-formed as a non-emphatic wh-construction.
(11)
(a) I wonder [what she ate].
(b) *I wonder [what did she eat].
I propose that an embedded wh-construction in English contains a different type of
Qu-morpheme from a matrix clause. The Qu-morpheme is [-AFFIX]. Therefore, it
does not force a tense head to move to Typ. A simplified diagram of the embedded
clause of (11a) is shown below in (26).
153
(26)
TypP
DP
what1
Typ’
Typ
∅FQu
TP
she ate t1
An embedded clause and a matrix clause thus differ with respect to the type
of Qu-morpheme that occurs; a matrix clause contains a [+AFFIX] Qu-morpheme
and an embedded clause contains a [-AFFIX] Qu-morpheme. This difference between Qu-morphemes relies on the features that these Qu-morphemes contain. The
matrix Qu-morpheme has a [+AFFIX] feature and a feature that I loosely refer to
as [+MATRIX] that requires it to only occur in a matrix clause, as shown in (27a).
The embedded Qu-morpheme has a [-AFFIX] feature and a [-MATRIX] feature that
requires it to occur in an embedded clause, as shown in (27b). It should also be
noted that both the matrix and embedded clause Qu-morphemes that occur in whconstructions are [-OVERT], unlike the [+OVERT] Qu-morpheme that occurs in
embedded yes/no constructions (see chapter 4.2.1-4.2.2).
(27)
Clause
Features
(a)
Matrix
[+MATRIX], [+AFFIX], [-OVERT]
(b)
Embedded
[-MATRIX], [-AFFIX], [-OVERT]
The particular features contained within the matrix and embedded English
Qu-morphemes determine their syntactic behavior. A language could have Qumorhemes with different sets of features. As discussed in chapter 4.2.3, there are
dialects of English that differ from standard English with respect to allowing a
[+AFFIX] Qu-morpheme to occur in an embedded clause in a yes/no construction.
In the next section, I discuss how in wh-constructions, dialects such as Ozark
154
English and Belfast English allow a [+AFFIX] Qu-morpheme to appear in an
embedded clause, and dialects such as South African Indian English allow a
[-AFFIX] Qu-morpheme to occur in a matrix clause.
5.3.4 Ozark English
In Ozark English, wh-constructions can be formed differently than in standard
American English. In this section, I argue that these differences result from Ozark
English containing a different set of Qu-morphemes than found in Standard American English.
Ozark English, unlike in ‘standard English,’ allows both [+AFFIX] and [-AFFIX]
Qu-morphemes in an embedded wh-construction. Example (28a) shows a complex
sentence with an embedded wh-construction that is formed with movement of an
auxiliary element, as would occur in a matrix wh-construction. I assume that the
structure is as shown in (28b) with a [+AFFIX] Qu-morpheme that attracts the
tense element to Typ.
(28)
(a) I want to find out who is she cheating with. (Barrett 2008)
(b) I want to find out [T ypP who2 [T yp ∅Qu -is1 [T P she [T t1 cheating with
t2 ]]]].
Ozark English also allows an embedded wh-construction to be formed as in standard
English with no movement of a tense element, as shown in (29a). In this case, the
embedded clause contains a [-AFFIX] Qu-morpheme that does not attract the tense
element, as shown in (29b).
(29)
(a) I want to find out who she is cheating with. (Barrett 2008)
(b) I want to find out [T ypP who1 [T yp ∅Qu she is cheating with t1 ]].
Ozark English then appears to contain two types of Qu-morphemes that can
occur in an embedded clause, as shown in (30a-b). These Qu-morphemes differ in
terms of their [AFFIX] feature.
155
(30)
(a) Qu[+AF F IX]
(b) Qu[−AF F IX]
A matrix non-subject wh-construction in Ozark English also requires subjectauxiliary inversion (Barrett 2008). Thus, as in standard American English, a matrix
clause contains a [+AFFIX] Qu-morpheme. Unlike in standard English, in Ozark
English, the [+AFFIX] Qu-morpheme can also occur in embedded clauses. Example
(31a) shows the [+AFFIX] Qu-morpheme. It has no features determining what type
of clause it can occur in. It is free to occur in either a matrix or embedded clause.
(31b), on the other hand, shows the [-AFFIX] Qu-morpheme, which has a feature
that requires it to occur in an embedded clause.
(31)
(a) Qu[+AF F IX]
(b) Qu[−AF F IX,−M AT RIX]
Not surprisingly, these Qu-morphemes also differ in their semantics.
According to Rusty Barrett (p.c), when there is inversion (under my analysis, ‘inversion’ is caused by a [+AFFIX] Qu-morpheme), then the embedded whconstruction refers to something that need not be true - there is no specified truth
value. When there is no inversion, the content of the embedded clause is believed
to be true. For example, Barrett (2008) writes that the embedded wh-construction
in (28a) above, which contains inversion, implies “that she might not be cheating.”
On the other hand, the embedded wh-construction in (29a) above, which lacks inversion, implies that “it is certain she is cheating.” Barrett notes that this semantic
distinction also accounts for the data in (32a-b). (32a) is fine because, in all likelihood, it is true that ‘he’ has a name. (32b) is ill-formed because the embedded
clause, which is dependent on the assumption that ‘he’ has a name, is incompatible
with a Qu-morpheme that does not imply truth.
(32)
(a) I wonder what his name is.
156
(b) *I wonder what is his name. (Barrett 2008)
These facts suggest that the [+AFFIX] embedded clause Qu-morpheme in Ozark
English does not carry with it a truth value, whereas a [-AFFIX] Qu-morpheme
indicates that a wh-construction refers to something that a speaker believes to be
true.
Several examples demonstrate some differences between wh-constructions with
and without inversion. According to Barrett, (33a), which I claim has the structure in (33b) with a [+AFFIX] Qu-morpheme, indicates that the speaker “can’t
understand a word she is saying.”
(33)
(a) I don’t understand what is she saying. (can’t understand a word she is
saying) (Barrett 2008)
(b) I don’t understand [T ypP what2 [T yp ∅Qu -is1 [T P she [T t1 saying t2 ]]]].
On the other hand, (34a), which I assume has the structure in (34b) with a [-AFFIX]
Qu-morpheme, indicates that the speaker thinks “her meaning/intention is unclear.”
(34)
(a) I don’t understand what she is saying. (her meaning/intention is
unclear) (Barrett 2008)
(b) I don’t understand [T ypP what1 [T yp ∅Qu she is saying t1 ]].
(35a), which I assume has the structure in (35b) with a [+AFFIX] Qu-morpheme is
an insult.
(35)
(a) I wonder where does she buy her clothes. (insult) (Barrett 2008)
(b) I wonder [T ypP where2 [T yp ∅Qu -does1 [T P she [T t1 buy her clothes t2 ]]]].
(36a), with the structure in (36b) with a [-AFFIX] Qu-morpheme, however, is a
compliment.
(36)
(a) I wonder where she buys her clothes. (compliment) (Barrett 2008)
157
(b) I wonder [T ypP where1 [T yp ∅Qu she buys her clothes t1 ]].
The Ozark Qu-morphemes are summarized in (37a-c).
(37)
Features
Meaning
(a) [+AFFIX]
No truth value
(c)
True
[-AFFIX], [-MATRIX]
(37a) is a Qu-morpheme that occurs in both embedded and matrix clauses,
as it lacks any specification for the type of clause that it can occur in, and it does
not carry a truth value. (37b) occurs in embedded clauses only. It is not an affix,
and thus does not trigger inversion, and it implies truth of the embedded clause.
In this section, I have only discussed Ozark English, but note that there are
other English dialects which also differ from ‘standard English’ with respect to
wh-construction formation. Belfast English and Hiberno English allow auxiliary
movement in an embedded clause, as shown in (38) and (39).
(38) Belfast English:
I wonder [what street does he live in]. (Henry 1992:288)
(39) Hiberno English:
Did he tell you [how did he do it]? (McCloskey 1991:294)
These dialects thus appear to contain a [+AFFIX] Qu-morpheme that can occur
in an embedded clause. Also, South African Indian English does not require tense
movement in a matrix clause, as shown in (40).
(40) What I must do? If my father say I must go an’ plough today, what I
can do?(Mesthrie 1992:47)
This dialect then appears to contain a matrix clause [-AFFIX] Qu-morpheme.
158
5.3.5 Summary
On the one hand, the facts regarding the Qu-morpheme in English wh-constructions
are straightforward. The Qu-morpheme is Merged directly in Typ. On the other
hand, there are certain complexities involving the [±AFFIX] value of the Qumorpheme, and also concerning where a tense element occurs. In ‘standard English,’
a matrix wh-construction contains a [+AFFIX] Qu-morpheme. The [+AFFIX] feature forces the tense head to move to Typ. Tense must either surface on an auxiliary
or on a main verb; in the latter case, Tense must be adjacent to the verb. When a
non-subject wh-phrase raises to [Spec, TypP], the tense head, which has moved to
Typ is not adjacent to the verb in v because of the intervening subject. Thus, if
the tense head does not already occur on an auxiliary, then a form of do must be
inserted. When a subject wh-phrase raises to [Spec, TypP], then the tense head is
adjacent to the verb. Therefore, if there is no auxiliary, tense is able to occur on
the verb, and no form of do is required. An embedded wh-construction in ‘standard
English’ contains a [-AFFIX] Qu-morpheme, and so there is no movement of a tense
element. In Ozark English, unlike in ‘standard English,’ an embedded clause allows both [+AFFIX] and [-AFFIX] Qu-morphemes to occur. These Qu-morphemes
also differ in their semantics. Other English dialects such as Belfast, Hiberno, and
South African Indian English similarly differ from ‘standard English’ with respect
to the types of Qu-morphemes that they contain. Note that various dialects of
English demonstrate variation with respect to the requirements they place on the
Qu-morphemes in matrix and embedded clauses. A matrix Qu-morpheme must be
[+AFFIX] in standard English, but not in South African Indian English. A Qumorpheme must be [-AFFIX] in embedded clauses in Standard English, but not in
Ozark English or Belfast English.
159
5.4 Qu-morphemes in various positions
In certain wh-constructions in Sinhala and Okinawan, a Qu-morpheme can either
appear adjacent to a wh-phrase in a TP-internal position, as in (41a), or separate
from it, as in (41b).8
(41)
(a) [T ypP . . . wh-phrase Qu . . . ]
(b) [T ypP . . . wh-phrase . . . Qu ]
As I argued was the case in certain yes/no constructions discussed in chapter 4, in
this section I argue that when a Qu-morpheme contains a Focus-feature, it must
occur adjacent to a wh-phrase, as in (41a). Otherwise it is Merged directly in a
clause peripheral position corresponding to Typ, as shown in (41b).
In Sinhala, a Qu-morpheme can either appear adjacent to (in a TP-internal
position) or separated from (in a clause-peripheral position) certain types of whadjuncts.9 These wh-adjuncts are referred to by Hagstrom (2004:230) as “scalar
wh-words.” They include kiidenek ‘how many (animate),’ kiiyak ‘how many (inanimate),’ and [email protected]@ ‘how much.’ Example (42a) shows the Qu-morpheme in clausefinal position separated from the in-situ wh-adjunct [email protected]@ ‘how much.’ In this
case, the default ‘-A’ ending appears on the verb. In (42b), the Qu-morpheme
appears adjacent to the wh-phrase and the special ‘-E’ ending shows up on the verb.
(42) Sinhala:
(a) Salli
[email protected]@ dunna [email protected]?
money how much give.A Qu
‘How much money did (you) give?’
8
In some languages, the position of the Qu-morpheme directly after the wh-phrase in (41a) and
at the end of the clause in (41b) may be the opposite, with the Qu-morpheme directly preceding
the wh-phrase or appearing at the beginning of a clause.
9
I discuss wh-arguments in Sinhala in the following sections.
160
(b) Salli
[email protected]@ [email protected] dunne?
money how much Qu give.E
‘How much money was it that (you) gave?’ (Sumangala 1992:248)
As can be seen in the glosses, the interpretations of these examples differ. A standard
wh-question interpretation arises when the Qu-morpheme appears in clause-final
position. When the Qu-morpheme appears TP-internally, the wh-phrase functions
as an identificational focus, as indicated by the English gloss containing a cleft (see
chapter 4.4 for discussion of focus).
In Okinawan, the facts are similar. The Qu-morpheme ga may either appear
adjacent to or separated from a matrix wh-phrase. Note that unlike in Sinhala in
which variability in the position of the Qu-morpheme in matrix clauses only occurs
with some wh-adjuncts, in Okinawan, this variability is allowed with wh-arguments.
Examples are shown in (43a-b). When the Qu-morpheme is adjacent to the whphrase in (43b), the matrix verb occurs with the special verbal agreement ending
‘-RA.’10
(43) Okinawan:
(a) Taa-ga
ich-u-ga?
who-nom go-pres-Qu
‘Who is going?’
(b) Taa-ga-ga
ich-u-ra?
who-nom-QU go-pres-RA
‘I wonder who on earth is going?’ (Miyara 2001:27)
The gloss provided by Miyara for (43b) indicates, via the phrase who on earth, that
when the Qu-morpheme is adjacent to the wh-phrase, the wh-phrase is focused.
10
As discussed in Chapter 4.4.2, I assume that ga is always a Qu-morpheme, contrary to Miyara
(2001) who claims that when ga occurs TP-internally it is an emphatic marker. I also assume that
the ‘-RA’ ending is verbal agreement, unlike Miyara who claims that it is a Qu-morpheme.
161
This appears to be the same type of phenomenon found in Sinhala. The wh-phrase
with an adjacent Qu-morpheme is an identificational focus.
In Sinhala, when a wh-phrase occurs within an embedded clause that is not an
island and the wh-phrase has embedded scope, there can be two possible positions
for the Qu-morpheme. The Qu-morpheme can occur adjacent to the embedded whphrase, as shown in (44a), or it can occur at the edge of the embedded clause as in
(44b).
(44)
(a) [T ypP . . . [T ypP . . . wh-phrase Qu . . . ] . . . ]
(b) [T ypP . . . [T ypP . . . wh-phrase . . . Qu ] . . . ]
However, only certain matrix verbs allow the Qu-morpheme to appear in these
two positions within an embedded wh-complement clause. Some of these verbs
are “[email protected] ‘know,’ hoya [email protected]@[email protected] ‘examine,’ parikSaa [email protected]@[email protected] ‘look into, inspect,’ and [email protected] ‘understand’ (Kishimoto 2005:8).”11 In (45a) below, the Qumorpheme [email protected] follows the wh-phrase kau ‘who’ in the embedded clause, whereas in
(45b), it occurs at the embedded clause-periphery, following the verb aawa, but
preceding the complementizer [email protected] ‘that.’12 In this way, the Qu-morpheme may
either be adjacent to or separated from the wh-phrase.13
11
I discuss these verbs in more detail in section 5.6, where I also discuss verbs such as ‘[email protected]
‘ask,’ [email protected] [email protected]@[email protected] ‘question,’ and [email protected] ‘consider’ (Kishimoto 2005:8)” that require the
Qu-morpheme to appear adjacent to a wh-phrase in a complement clause.
12
These examples contain the complementizer [email protected] ‘that’ in addition to a Qu-morpheme in the
embedded clause. This fact can be accounted for if [email protected] is in ForceP and the Qu-feature of the
Qu-morpheme is in TypP. See chapter 2.5 for discussion of clause structure.
13
Note that the form of the subject meaning ‘who’ differs. It is kau in (45a) and kauru in (45b).
According to Kishimoto (2005:6), “[t]he wh-word meaning ‘who’ is kauru, but when [email protected] is adjacent
to the wh-word, -ru is dropped, as in kau [email protected]”
162
(45) Sinhala:
(a) Ranjit [kau [email protected] aawe
[email protected]] [email protected]
Ranjit who Q came.E that
know.A
‘Ranjit knows who came.’ (Kishimoto 2005:7)
(b) Ranjit [kauru aawa
[email protected] [email protected]] [email protected]
Ranjit who
came.A Qu that
know.A
‘Ranjit knows who came.’ (Kishimoto 2005:6)
As expected, there is a meaning difference depending on the position of the Qumorpheme. When the Qu-morpheme is adjacent to the wh-phrase, the wh-phrase
is an identificational focus, and when the Qu-morpheme is separated from the whphrase, the wh-phrase is not an identificational focus. In other words (45a) means
something along the lines of ‘Ranjit knows who it is that came.’ Although these
differences in interpretation are not indicated in the glosses by Kishimoto, they
are supported by Gair & Sumangala (1991), who write that the wh-phrase in an
embedded clause such as (46) below is focused when it is adjacent to a Qu-morpheme
(Gair & Sumangala 1991:98). This example is comparable to (45a) above.
(46) Sinhala:
[Siri mokak [email protected] keruwe
[email protected]] amma [email protected]
Siri what
Qu do-past.E comp mother think-past.A
‘Mother thought (about) what Siri did. (Gair & Sumangala 1991:98)
In (47) below, which is comparable to (45b), Gair & Sumangala (1991:96) write that
the wh-phrase is not focused.
(47) Sinhala:
[[email protected] kauru hæduwa [email protected]] danne nææ.
that who
did.A
Qu know not
‘I don’t know who could have done that.’ (Gair & Sumangala 1991:96)
163
Gair & Sumangala (1991:96) attribute the lack of focus in this type of construction
to what they refer to as “general doubt.”
If this analysis is on the right track, the Sinhala data indicate that when a Qumorpheme is adjacent to a wh-phrase, that phrase is an identificational focus. I
propose that in this case, the Qu-morpheme contains a Focus-feature. This is the
same type of Qu-morpheme that I argued occurs TP-internally in yes/no constructions in these languages, as discussed in chapter 4.4. The Focus-feature forces the
Qu-morpheme to focus a TP-internal element. The wh-phrase must have scope in a
particular clause, and as such, if anything is focused, it must be the wh-phrase, and
so the Qu-morpheme adjoins to it. The structure of a wh-phrase with an adjoined
Qu-morpheme is given in (48).
(48)
whP
whP
Qu[FF oc ,FQu ]
wh-phrase
The Qu- and Focus-features raise as a feature bundle from the Qu-morpheme, and
the Qu-morpheme remains in-situ. These features raise to Foc, where the Focusfeature remains and focuses the wh-phrase, and the Qu-feature raises to Typ where
it types the clause as an interrogative. Movement of these features is motivated by
EPP features. A diagram is shown in (49). The Focus-feature in Foc shows up as
the verbal agreement ending ‘-E’ in Sinhala and ‘-RA’ in Okinawan.
164
(49)
TypP
Typ’
FocP
TypEP P
FQu2
Foc’
TP
FocEP P
-E/-RA[FF oc1 ]
T’
vP
T
. . . wh-phrase+(Qu+t[FF oc1 ,FQu2 ] ) . . .
When the Qu-morpheme occurs at the edge of a clause in these languages, the
wh-phrase does not receive an identificational focus interpretation. Rather, there is
a standard wh-construction interpretation in which the wh-phrase is an information
focus. In this case, the Qu-morpheme lacks a Focus-feature. It is Merged directly
in Typ where it types a clause as an interrogative. A diagram is shown in (50).
Because there is no Focus-feature in Foc, the verb does not show up with the special
agreement ending.
165
(50)
TypP
Typ’
TP
Typ
Qu[FQu ]
T’
vP
T
. . . wh-phrase . . .
Also, in these constructions, a wh-feature (which is a distinct element from a Qufeature) associated with a wh-phrase establishes a relationship with Typ, thereby
giving the wh-phrase scope. I examine the behavior of wh-features in the next
chapter.
I next turn to embedded clauses which contain a wh-phrase that has matrix
scope. In both Sinhala and Okinawan, when a wh-phrase in an embedded clause
has matrix scope, a Qu-morpheme either occurs adjacent to or separated from the
wh-phrase. When separated from the wh-phrase, the Qu-morpheme can either occur
at the edge of the embedded clause, or at the edge of the matrix clause. The three
possible positions of the Qu-morpheme are shown in (51a-c).
(51)
(a) [T ypP . . . [T ypP . . . wh-phrase Qu . . . ] . . . ]
(b) [T ypP . . . [T ypP . . . wh-phrase . . . ] Qu . . . ]
(c) [T ypP . . . [T ypP . . . wh-phrase . . . ] . . . Qu ]
Examples (52a-b) from Sinhala show that the Qu-morpheme can occur adjacent
to the wh-phrase or at the edge of the embedded clause. In the latter case, note
that the Qu-morpheme follows the embedded clause complementizer, suggesting
that the Qu-morpheme is adjoined to the entire embedded clause. I assume that
166
the complementizer [email protected] ‘that’ is in ForceP (see the discussion of clause structure
in chapter 2.5).
(52) Sinhala:
(a) Chitra [kau [email protected] aawa
[email protected]] kiiwe?
Chitra who Qu came.A that
said.E
‘Who did Chitra say came?’
[email protected]] [email protected] kiiwe?
(b) Chitra [kauru aawa
came.A that
Qu said.E
Chitra who
‘Who did Chitra say came?’ (Kishimoto 2005:21)
Example (53b) shows that the Qu-morpheme can also appear at the edge of the
matrix clause. Compare this with (53a) in which the Qu-morpheme is adjacent to
the wh-phrase in the embedded clause.
(53) Sinhala:
(a) Ranjit [Chitra kiidenek
[email protected] dækka [email protected]] danne?
Ranjit Chitra how.many Qu saw.A that
know.E
‘How many (people) does Ranjit know that Chitra saw?’ (Kishimoto
2005:7)
(b) Ranjit [kiidenek [email protected] [email protected]] [email protected] [email protected]?
Ranjit how.many come.A that
know.A Qu
‘How many (people) does Ranjit know will come?’ (Kishimoto 2005:6)
As discussed in chapter 4.4, when a Qu-morpheme in an embedded clause has matrix
scope, the special verbal agreement only shows up in the matrix clause. Thus, in
examples (52a) and (53a), only the matrix verb appears with the special verbal
ending ‘E.’ In (53b), in which the Qu-morpheme appears in clause-final position of
the matrix clause, the matrix verb appears with the default ‘A’ ending.
In these examples, the Qu-morpheme focuses different constituents. In (52a),
the Qu-morpheme focuses the wh-phrase and in (52b), it focuses the embedded
167
clause. These differences are not indicated in Kishimoto’s glosses, but they are
indicated in his descriptions of the meanings of these examples. Kishimoto (2005:21)
explains that an answer to (52a) can “provide just the value of the wh-phrase as
a minimal answer.” However, when answering (52b), “it is necessary to repeat the
embedded clause with the complementizer, alongside the value of the wh-phrase.” In
(52a) the answer must refer to the wh-phrase, but not necessarily to the embedded
clause, because the wh-phrase is an identificational focus, and the embedded clause
is not. In (52b), the answer must refer to the entire embedded clause because the
embedded clause is focused. Example (53b) with the Qu-morpheme at the edge
of the matrix clause presumably does not contain any TP-internal identificational
focus. This example functions like example (42a) above, which also has a clause
final Qu-morpheme, and lacks an identificational focus.
Similar examples in Okinawan are shown in (54). In (54a), the Qu-morpheme
is adjacent to the embedded clause subject taa-ga ‘who-nom’ and the matrix verb
shows the special ‘-RA’ ending. As in Sinhala, when a Qu-morpheme in an embedded clause has matrix scope, the special verbal agreement only shows up in the
matrix clause (see chapter 4.4). In (54b), the Qu-morpheme is at the edge of the
embedded clause, and the matrix verb again shows the ‘-RA’ ending. In (54c), the
Qu-morpheme is at the edge of the matrix clause and the ‘-RA’ verbal ending is
absent.
(54) Okinawan:
(a) [CP Taa-ga-ga
suba
kada-n-di]
John-oo
umutoo-ra?
who-nom-Qu noodles ate-ind-comp John-top think-RA
‘Who the hell does John think ate the noodles?’ (Miyara 2001:52)
(b) [CP Taa-ga
suba
kada-n-di]-ga
John-oo
umutoo-ra?
who-nom noodles ate-ind-comp-Qu John-top think-RA
‘Who the hell does John think ate the noodles?’ (Miyara 2001:38)
168
(c) [CP Taa-ga
suba
kada-n-di]
John-oo
umutoo-ga?
who-nom noodles ete-ind-comp John-top think-Qu
‘Who does John think ate the noodles?’ (Miyara 2001:52)
My analysis given in chapter 4 that a Qu-morpheme with a Focus-feature appears
in a TP-internal position and focuses an adjacent phrase predicts the differences in
the glosses of (54a-b) when compared with (54c). In (54a-b), the glosses containing
who the hell indicate identificational focus. Although the glosses for (54a-b) are
the same, I assume that the differing positions of the Qu-morphemes indicates that
different constituents are focused. In (54c), there is no focused phrase within the
clause, and thus there is no focused element in the gloss.
I account as follows for these Sinhala and Okinawan examples with an embedded wh-phrase that obtains matrix scope. A TP-internal Qu-morpheme contains a
Focus-feature that forces it to be Merged in a position adjoined to either the whphrase or the embedded clause containing the wh-phrase. In each case, the Qu- and
Focus-features of the Qu-morpheme move on to the matrix Foc and Typ projections. Movement is driven by EPP features. The Focus-feature in the matrix Foc is
pronounced as the verbal affix -E in Sinhala and -RA in Okinawan. The verb raises
at least to T, although possibly even to Foc. In the simplified diagram in (55), the
Qu-morpheme is adjoined to the wh-phrase within the embedded TypP and in (56),
the Qu-morpheme is adjoined to the embedded TypP.
169
(55)
TypP
Typ’
FocP
TypEP P
FQu2
Foc’
TP
FocEP P
-E/-RA[FF oc1]
T’
vP
T
. . . [T ypP . . . wh-phrase+(Qu+t[FF oc1 ,FQu2 ] ) . . . ] . . .
(56)
TypP
Typ’
FocP
TypEP P
FQu2
Foc’
TP
FocEP P
-E/-RA[FF oc1]
T’
vP
T
. . . [T ypP . . . wh-phrase . . . ]+(Qu+t[FF oc1 ,FQu2] ) . . .
The Focus-feature raises to the matrix Foc and focuses the wh-phrase, as in (55),
or it focuses the embedded clause containing the wh-phrase, as in (56), and the Qufeature raises to Typ where it types the matrix clause as an interrogative. When the
170
Qu-morpheme occurs in a clause-peripheral position, it lacks a Focus-feature and is
simply Merged in the matrix Typ. A diagram is shown below.
(57)
TypP
Typ’
TP
Typ
Qu[FQu ]
T’
vP
T
. . . wh-phrase . . .
If this analysis is on the right track, in a wh-construction in languages such as
Sinhala and Okinawan, a Qu-morpheme with a Focus-feature adjoins to a wh-phrase.
Movement of the Focus-feaure to Foc gives the wh-phrase scope and also focuses it.
The Qu-morpheme’s Qu-feature also moves to Typ. When a Qu-morpheme appears
in a clause-peripheral position, it lacks a Focus-feature and is Merged directly in
Typ.
5.5 Qu-morphemes and island effects
I next turn to wh-constructions in Sinhala and Okinawan that require the Qumorpheme to appear at the edge of an embedded clause. When a Qu-morpheme occurs inside of certain types of embedded clauses, such as a complex-DP, whether/if clause, or adjunct clause, then there is an island effect (see chapter 2.6). If the
Qu-morpheme occurs at the edge of this type of clause, then there is no island
effect. The configuration in (58a) is ill-formed but that in (58b) is acceptable.
(58)
(a) *[T ypP . . . [complex−DP/whether/if −clause/adjunct−clause . . . wh-phrase Qu . . . ]
...]
171
(b) [T ypP . . . [complex−DP/whether/if −clause/adjunct−clause . . . wh-phrase . . . ] Qu
...]
In this section, I show that these island effects result from Qu-feature movement.
In (59a), the Qu-morpheme is adjacent to kaa-ú@ ‘who-dat’ inside of a complexDP, and ill-formedness results. The complex-DP then functions as an island. In
(59b), the Qu-morpheme is adjacent to the complex-DP, and the result is wellformed.
(59) Sinhala:
(a) *Oyaa [DP [T ypP Chitra kaa-ú@ [email protected] [email protected]] [email protected]] kieuwe?
you
Chitra who-dat Qu gave
book read.E
‘You read the book that Chitra gave to whom?
(b) Oyaa [DP [T ypP Chitra kaa-ú@ [email protected]] [email protected]] [email protected] kieuwe?
you
Chitra who-dat gave
book Qu read.E
‘You read the book that Chitra gave to who? (Kishimoto 2005:29)
The same is true for adjunct clauses. In the ill-formed (60a) the Qu-morpheme
is inside of the adjunct clause (the embedded TypP), and in the well-formed (60b),
the Qu-morpheme is at the edge of the clause.
(60) Sinhala:
(a) *[T ypP Chitra [email protected] [email protected] [email protected] koú@] Ranjit [email protected]
Chitra what
Qu ate time Ranjit surprise
unee?
became.E
‘What1 was Ranjit surprised when Chitra ate t1 ? (Kishimoto 2005:29)
(a) [T ypP Chitra [email protected] [email protected] koú@] [email protected] Ranjit [email protected] unee?
Chitra what
ate time Qu Ranjit surprise became.E
‘What1 was Ranjit surprised when Chitra ate t1 ? (Kishimoto 2005:30)
172
Similarly, when a wh-phrase with an adjacent Qu-morpheme occurs within an
embedded whether/if -clause the result is ill-formed, as in (61a). Otherwise, when
the Qu-morpheme is at the edge of the whether/if -clause, the result is well-formed,
as shown in (61b).
(61) Sinhala:
(a) ?*Ranjit [T ypP Chitra [email protected] [email protected] kieuwa [email protected]æ[email protected] [email protected]]
Ranjit
Chitra what
Qu read.A whether that
danne?
know.E
‘What1 does Ranjit know whether Chitra read t1 ?
(b) Ranjit [T ypP Chitra [email protected] kieuwa [email protected]æ[email protected] [email protected]] [email protected]
Ranjit
Chitra what
read.A whether that
Qu
danne?
know.E
‘What1 does Ranjit know whether Chitra read t1 ? (Kishimoto
2005:29-30)
The facts are identical in Okinawan. Example (62a) shows a wh-phrase and adjacent Qu-morpheme in a complex-DP. This is ill-formed, but when the Qu-morpheme
is at the edge of the complex-DP, well-formedness results, as in (62b).
(62) Okinawan:
(a) *[DP [T ypP taa-ga-ga
ka-chee-ru]
shimuchi]
who-nom-Qu write-have-nom book(-acc)-top
John-oo
yuda-ra?
John-TOP read-RA
‘[Who the hell]1 did John read the book that t1 had written?’
(b) [DP [T ypP taa-ga
ka-chee-ru]
shimuchi]-ga
who-nom write-have-nom book(-acc)-top-ga
John-oo
yuda-ra?
John-TOP read-RA
173
‘[Who the hell]1 did John read the book that t1 had written?’ (Miyara
2001:52)
When the Qu-morpheme is within an adjunct clause as in (63a), ill-formedness
results. When it is at the edge of the adjunct clause as in (63b), the result is
well-formed.
(63) Okinawan:
(a) *[T ypP Nuu-ga
chichi-gachiinaa] John-oo
benkyoosa-ra?
what(-acc)-Qu listening-while
John-top studied-RA
‘[What the hell]1 did John study while listening to t1 ?’
(b) [T ypP Nuu
chichi-gachiinaa]-ga John-oo
benkyoosa-ra?
what(-acc) listening-while-Qu
John-top studied-RA
‘[What the hell]1 did John study while listening to t1 ?’ (Miyara 2001:38)
These constructions suggest that it is the position of the Qu-morpheme that determines whether or not there is an island effect. Specifically, when a Qu-morpheme
appears in certain types of clauses such as a complex-DP, whether/if -clause, or adjunct clause, there is an island effect. When the Qu-morpheme appears at the edge
of these types of clauses, there is no island effect. Hagstrom (1998) argues that when
the Qu-morpheme is contained within an island it undergoes covert movement that
causes an island effect. I adopt this analysis, in a slightly modified version. I claim
that the Qu-morpheme does not undergo LF movement. Instead, it remains in-situ,
but its’ Qu- and Focus-features move at Spell-Out. The Qu-morpheme does not
move because in these languages, Qu- and Focus-features are able to separate from
it. These features, however, are subject to blocking effects that result in the island
effects discussed here.
First of all, I propose that the complex-DP and adjunct-island effects have the
same cause, which is the inability to extract a feature out of an adjunct because
the adjunct has been renumerated (see discussion of renumeration in chapter 2.6.2).
174
In (59a) and (62a), the Qu-morpheme is contained within an adjunct modifier of a
nominal within a Complex-DP. In (60a) and (63a), the Qu-morpheme is within an
adjunct clause. According to Johnson’s (2002) notion of renumeration, an adjunct
is assembled and reinserted into a numeration. It essentially functions as a fixed
lexical item, and nothing can be extracted from it. Therefore, Qu- and Focusfeatures cannot escape from an adjunct and raise to Typ and Foc. For example, in
(59a), repeated below, the adjunct clause is assembled via Merge.
(59) Sinhala:
(a) *Oyaa [DP [T ypP Chitra kaa-ú@ [email protected] [email protected]] [email protected]] kieuwe?
you
Chitra who-dat Qu gave
book read.E
‘You read the book that Chitra gave to whom? (Kishimoto 2005:29)
(64) shows the numeration of (59a), after the adjunct has been assembled and renumerated.
(64) N = {Oyaa, < Chitra kaa-ú@ [email protected] [email protected]>, [email protected], kieuwe }
The adjunct clause functions as a lexical item that is selected from the numeration
and Merged with the nominal [email protected] ‘book’ within the complex-DP. Since the adjunct
clause is a fixed lexical item, Qu- and Focus-features cannot move out of it.
In a construction such as (60a), repeated below, the Qu-morpheme is also contained within an adjunct clause that is renumerated.
(60) Sinhala:
(a) *[T ypP Chitra [email protected] [email protected] [email protected] koú@] Ranjit [email protected]
Chitra what
Qu ate time Ranjit surprise
unee?
became.E
‘What1 was Ranjit surprised when Chitra ate t1 ? (Kishimoto 2005:29)
175
The adjunct clause is formed via Merge and then put back into the numeration as
a single lexical item. (65) shows the numeration of (60a), once the adjunct clause
is renumerated.
(65) N = {<Chitra [email protected] [email protected] [email protected] koú@>, Ranjit, [email protected], unee}
This numeration results in an ill-formed construction because the Qu- and Focusfeatures are unable to move out of the adjunct-clause.
When the Qu-morpheme appears at the edge of a complex-DP or adjunct-clause,
it is an element that is base-generated outside of the clause. For example, the
Numeration of (59b), repeated below, is as shown in (66).
(59) Sinhala:
(b) Oyaa [DP [T ypP Chitra kaa-ú@ [email protected]] [email protected]] [email protected] kieuwe?
you
Chitra who-dat gave
book Qu read.E
‘You read the book that Chitra gave to who? (Kishimoto 2005:29)
(66) N = {Oyaa, < Chitra kaa-ú@ [email protected]>, [email protected] , [email protected], kieuwe }
The Qu-morpheme is not part of the complex-DP. The Qu-morpheme Merges with
the Complex-DP, and its Qu- and Focus-features are able to move freely to Typ
and Foc. The same is true when the Qu-morpheme appears outside of the adjunctclause in (60b), repeated below. It is outside of the adjunct clause, as shown in the
Numeration in (67). The Qu-morpheme is able to Merge with the adjunct-clause
and its Qu- and Focus-features can move freely to Typ and Foc.
(60) Sinhala:
(b) [T ypP Chitra [email protected] [email protected] koú@] [email protected] Ranjit [email protected] unee?
Chitra what
ate time Qu Ranjit surprise became.E
‘What was Ranjit surprised when Chitra ate? (Kishimoto 2005:30)
(67) N = {<Chitra [email protected] [email protected] koú@>, [email protected], Ranjit, [email protected], unee}
176
In a construction such as (61a), repeated below, the Qu-morpheme is contained
within a whether/if -clause. In this case, unlike with an adjunct clause, ill-formedness
results from a violation of the Minimal Link (MLC) condition (see chapter 2.6 and
2.7 for discussion of the MLC).
(61) Sinhala:
(a) ?*Ranjit [T ypP Chitra [email protected] [email protected] kieuwa [email protected]æ[email protected] [email protected]]
Ranjit
Chitra what
Qu read.A whether that
danne?
know.E
‘What1 does Ranjit know whether Chitra read t1 ?
Movement of the Qu- and Focus-features to Typ is blocked by [email protected]æ[email protected] ‘whether’
in the embedded clause. This blocking effect can be accounted for as resulting from
[email protected]æ[email protected] ‘whether’ being an element of the same type (having a quantificational
element) as a Qu- and Focus-feature. I propose that movement of these features is
driven by an EPP feature in the embedded Typ, but [email protected]æ[email protected] ‘whether’ blocks the
features from landing in Typ, thus preventing them from moving on to the matrix
clause. A simplified diagram is shown below.
(68)
ForceP
Force’
TypP
Typ’
TP
Chitra [email protected] [email protected]+t[FF oc,FQu ] kieuwa
Chitra what Qu read-A
Force
[email protected]
that
TypEP P
[email protected]æ[email protected]
whether
177
In summary, a Qu-morpheme cannot appear adjacent to a wh-phrase in certain
types of embedded clauses such as complex-DPs, adjunct clauses, and whether/if clauses because movement of Qu- and Focus-features from the Qu-morpheme cannot
occur, either due to renumeration or an MLC effect. If my proposal that a Focusfeature forces TP-internal Merge of a Qu-morpheme is correct, then when a whphrase is within an island, a Qu-morpheme cannot focus the wh-phrase. The clause
containing the wh-phrase, however, can be focused by an adjacent Qu-morpheme.
Thus, when a Qu-morpheme appears at the edge of this type of clause, the clause
functions as an identificational focus.14
5.6 Remaining problems regarding TP-internal Qu-morphemes in Sinhala
In this section, I examine some perplexing facts regarding Qu-morphemes that must
appear in a TP-internal position in Sinhala. First of all, I discuss the fact that
certain verbs impose restrictions on the position of the Qu-morpheme. Specifically,
certain verbs require a Qu-morpheme to appear adjacent to a wh-phrase, when the
wh-phrase occurs in a clausal complement to the verb. Secondly, in a matrix clause,
a Qu-morpheme generally has to appear adjacent to a wh-phrase.
5.6.1 Embedded wh-constructions
As discussed in section 5.4, in Sinhala, a Qu-morpheme may either appear adjacent to or separate from a wh-phrase when the wh-phrase is in an embedded
clausal complement of certain verbs, such as “[email protected] ‘know,’ hoya [email protected]@[email protected] ‘examine,’ parikSaa [email protected]@[email protected] ‘look into, inspect,’ and [email protected] ‘understand’ (Kishimoto 2005:8).” However, not all verbs that can take an interrogative complement
clause allow this freedom in the position of the Qu-morpheme.
14
In a well-formed construction of this sort in which a Qu-feature appears at the edge of a
potential island, the embedded wh-phrase obtains scope without incurring an island effect. I
address this issue in chapter 6.
178
When verbs such as “‘[email protected] ‘ask,’ [email protected] [email protected]@[email protected] ‘question,’ and [email protected]
‘consider’ (Kishimoto 2005:8),” have a complement clause that is a wh-construction,
the Qu-morpheme must appear adjacent to the wh-phrase in the configuration in
(69a). (69b) is ill-formed.
(69)
(a) [T ypP . . . [T ypP . . . wh-phrase Qu . . . ] . . . ]
(b) *[T ypP . . . [T ypP . . . wh-phrase . . . Qu ] . . . ]
For example, (70a-b) show that this requirement holds when the matrix verb is
æhuwa ‘asked.’ When the Qu-morpheme is adjacent to the wh-phrase in (70a), the
result is well-formed, but when it is separated from the wh-phrase in (70b), the
result is ill-formed.
(70) Sinhala:
(a) Ranjit [kauru [email protected] aawe
[email protected]] æhuwa.
Ranjit who
Qu came.E that
asked.A
‘Ranjit asked who came.’
(b) *Ranjit [kauru aawa
[email protected] [email protected]] æhuwa.
Ranjit who
came.A Qu that
asked.A
‘Ranjit asked who came.’ (Kishimoto 2005:8)
Whether or not a Qu-morpheme in a complement clause can separate from an
associated wh-phrase then is dependent on the matrix verb. The following examples
make this clear. Compare (70a-b) above with (45a-b), repeated below, in which the
matrix verb is [email protected] ‘know.’ Whereas the Qu-morpheme can be separated from
the wh-phrase kauru ‘who’ in (45b), the same is not true in the virtually identical
(70b) above. The only difference between these constructions is the matrix verb.
(45) Sinhala:
(a) Ranjit [kau [email protected] aawe
[email protected]] [email protected]
Ranjit who Qu came.E that
know.A
‘Ranjit knows who came.’ (Kishimoto 2005:7)
179
(b) Ranjit [kauru aawa
[email protected] [email protected]] [email protected]
Ranjit who
came.A Qu that
know.A
‘Ranjit knows who came.’ (Kishimoto 2005:6)
Assuming that when the Qu-morpheme is adjoined to the wh-phrase, as in (70a),
the wh-phrase is an identificational focus, as I have claimed is the case in the previous
section, then a wh-phrase in the complement clause of certain verbs must be an
identificational focus.
In order to determine why a wh-phrase in the complement of certain verbs must
be an identificational focus, it is necessary to examine the distinction between verbs
that require a Qu-morpheme to appear adjacent to a wh-phrase in a complement
clause and those that do not. (71) shows verbs that fall into these two categories.
This list is by no means exhaustive. Qu-separate refers to verbs that allow the Qumorpheme to separate from a wh-phrase in a complement clause and Qu-adjacent
refers to verbs that require the Qu-morpheme to be adjacent to a wh-phrase.
(71)
Qu-separate
Qu-adjacent
“[email protected] ‘know’
‘[email protected] ‘ask’
hoya [email protected]@[email protected] ‘examine’
[email protected] [email protected]@[email protected] ‘question’
parikSaa [email protected]@[email protected] ‘look into, inspect’
[email protected] ‘consider’
[email protected] ‘understand’
Unfortunately, it is not clear what the distinction between these two types of
verbs is. Kishimoto (2005:9) writes that “there is no correlation between the class
of predicates permitting the clause-final Q-placement in the complement clause and
the class of predicates selecting a non-interrogative complement.” For example, both
[email protected] ‘know’ and hoya [email protected]@[email protected] ‘examine’ allow the Qu-morpheme to appear
both adjacent to the wh-phrase and clause-finally in the complement clause, yet
[email protected] ‘know’ allows a declarative complement, but hoya [email protected]@[email protected] ‘examine’
180
does not, as shown below.
(72) Sihhala:
Ranjit [Chitra aawa [email protected]] [email protected]/*hoya [email protected]@[email protected]
Ranjit Chitra came that
know-A/examine-A
‘Ranjit knows/examines that Chitra came.’ (Kishimoto 2005:9)
The distinction between the two types of verbs likely has to do with their semantics. Gair & Sumangala (1991:96) write that when a Qu-morpheme can occur
separated from a wh-phrase in an embedded clause, the embedded clause refers to
“general doubt.” If this is the case, then the class of verbs that does not allow an
embedded clause to express general doubt requires the Qu-morpheme to be adjoined
to a wh-phrase, and the class of verbs that allows a complement clause to express
general doubt allows the Qu-morpheme to be separated from a wh-phrase. The verbs
in the first column of (73) allow a complement to express general doubt, whereas
the verbs in the second column do not.
(73)
Qu-separate: permit general doubt
Qu-adjacent
“[email protected] ‘know’
‘[email protected] ‘ask’
hoya [email protected]@[email protected] ‘examine’
[email protected] [email protected]@[email protected] ‘question’
parikSaa [email protected]@[email protected] ‘look into, inspect’
[email protected] ‘consider’
[email protected] ‘understand’
However, if general doubt is involved, why certain verbs allow an interrogative complement to express general doubt and other verbs do not becomes an issue.
Further examination of the semantic distinction between these two types of verbs
is required.
181
5.6.2 Matrix wh-constructions
I next turn to another instance in Sinhala in which a Qu-morpheme must appear
adjacent to a wh-phrase. In a matrix clause, a Qu-morpheme generally must be
adjacent to its associated wh-phrase,15 as shown in (74a-b).
(74)
(a) [T ypP . . . wh-phrase [email protected] . . . ]
(b) *[T ypP . . . wh-phrase . . . [email protected] ]
For example, the Qu-morpheme must attach onto the matrix wh-phrase in (75a)
below. The verb also shows up with the special ‘-E’ ending. When the Qu-morpheme
is separated from the wh-phrase in (75b), ill-formedness results.
(75) Sinhala:
(a) Oyaa mokak [email protected] dække?
you what
Q see-past.E
‘What did you see?’
(b) *Oyaa mokak dække
[email protected]?
you
what
see-past.E Qu (Sumangala 1992:212)
One possibility is that in a matrix clause, a wh-phrase of this sort simply must
be an identificational focus. If this is the case, then the Qu-morpheme contains a
Focus-feature which forces it to adjoin to the wh-phrase. There is a problem for
this analysis though. According to Sumangala, (75a) has two possible meanings,
both given in (76). Meaning (i) is a standard wh-construction in which the whphrase presumably only receives an information focus interpretation, and in (ii) the
wh-phrase functions as an identificational focus, as can be seen in the clefted gloss.
15
The exception are scalar wh-words such as kiidenek ‘how many (animate),’ kiiyak ‘how many
(inanimate),’ and [email protected]@ ‘how much.’ See example (42) above.
182
(76) Sinhala:
Oyaa [email protected] dække?
you what-Q
see-past.E
(i) ‘What did you see?’
(ii) ‘What is it that you saw?’ (Sumangala 1992:212)
The fact that an identificational focus-interpretation is permitted, as indicated by
meaning (ii) is accounted for straightforwardly. The Qu-morpheme contains a Focusfeature and adjoins to the wh-phrase. The Focus- and Qu-features then undergo
movement to Foc and Typ. However, if Sumangala is correct about meaning (i)
being permitted, then focus is not always the reason why the Qu-morpheme must
appear adjacent to the wh-phrase.
Why exactly a Qu-morpheme must adjoin to a non-scalar wh-phrase in a matrix
clause thus remains to be explained. Gair & Sumangala (1991) suggest that an
explanation for this distinction between scalar and non-scalar wh-phrases has to
do with the semantics of the wh-phrase, although they also express their lack of
understanding of this distinction. Gair & Sumangala (1991:97) write the following.
At present, we have no explanation for why quantifier [scalar] WHs
should be exceptions to the general rule [that requires a Qu-morpheme
to be adjacent to a wh-phrase], though we assume that it is a function
of their semantic character.
Note that the requirement that a Qu-morpheme occur adjacent to a non-scalar whphrase clearly is not universal. In Okinawan, a language that like Sinhala allows
some variability in the positions of Qu-morphemes in wh-constructions, the Qumorpheme can be separated from a non-scalar matrix wh-phrase, as can be seen in
(43a-b) above.
183
5.6.3 Summary
In summary, the Sinhala Qu-morpheme must appear adjacent to a wh-phrase in the
two following situations given in (77a-b).
(77)
(a) an embedded clausal complement of certain matrix verbs such as
‘[email protected] ‘ask,’ [email protected] [email protected]@[email protected] ‘question,’ and [email protected] ‘consider’
(b) a matrix clause with a non-scalar wh-phrase
These facts leave much unexplained. I have argued that in situation (77a), the
Qu-morpheme must occur adjacent to the wh-phrase because the wh-phrase must
be an identificational focus. Why this is the case is an issue that I leave unresolved, but the explanation appears to have to do with the semantics of a matrix
verb. If Sumangala’s claim that a wh-phrase in situation (77b) need not have an
identificational focus interpretation is correct, then there may be some constraint
that requires the Qu-morpheme to be adjacent to a non-scalar wh-phrase in a matrix
clause, regardless of whether or not the Qu-morpheme has a Focus-feature. Furthermore, if Sumangala is correct, then whereas if a Qu-morpheme has a Focus-feature
it must be adjacent to a wh-phrase, as discussed in section 5.4, the converse is not
necessarily true. If a Qu-morpheme is adjacent to a wh-phrase there may be instances in which the Qu-morpheme lacks a Focus-feature. Further examination of
this issue is required.
5.7 Comparison with previous analyses
Previous analyses of TP-internal Qu-morphemes have relied on notions of operator
and LF movement to account for the facts. I briefly compare some of these analyses
with my own.
Kishimoto (2005) claims that in a wh-construction, a Qu-morpheme is always
Merged in a TP-internal position adjoined to a wh-constituent. Kishimoto writes
184
that “the Q-particle [email protected] serves primarily to delimit a wh-constituent in a wh-question
(16).” As I understand this, the term ‘delimit’ refers essentially to focusing an element, as Kishimoto (2005:16-17) writes that [email protected] ”serves to pick out a focused constituent in a wh-question.” Evidence for this proposal is that when a Qu-morpheme
occurs in clause-final position as in (78a), a minimal answer of (78b) is fine.
(78)
(a) Q: Chitra kiidenek
dækka [email protected]?
Chitra how.many saw.A Qu
‘How many (people) did Chitra see?’
(b) A: tundenek.
two
‘Two (people).’ (Kishimoto 2005:19)
Kishimoto argues that if the Qu-morpheme were initially Merged in a clauseperipheral position it would delimit the entire wh-construction, and thus a minimal
answer that just refers to the wh-phrase should not be allowed. However, under my
analysis, a clause-final Qu-morpheme is Merged directly in Typ and it simply lacks
a Focus-feature. Therefore, in a construction such as (78a), the Qu-morpheme does
not focus the entire clause, and a minimal answer that only refers to the wh-phrase
should be acceptable.
Kishimoto argues that a Qu-morpheme is an operator that raises to the scope
position of the wh-phrase. Notably, this movement can occur either at PF, in which
case the Qu-morpheme appears in a clause-peripheral position, or at LF, in which
case the Qu-morpheme remains in a TP-internal position adjoined to a delimited
wh-constituent. Kishimoto proposes that a verb in the clause where the wh-phrase
has scope has a [+Q] feature that is either strong or weak. When the [+Q] feature is
weak, the verb shows up with the -E ending, and the Qu-morpheme remains in-situ
adjoined to a wh-constituent. A weak [+Q] feature can be checked at LF. Therefore,
the Qu-morpheme moves to [Spec, CP] at LF and checks the [+Q] feature. When
the [+Q] feature is strong, it forces overt movement of the Qu-morpheme to the
185
clause-periphery. Movement of the Qu-morpheme checks the strong [+Q] feature
on the verb. If I understand correctly, the verb raises to C where its [+Q] feature
is eliminated via “a Spec-head configuration (23)” between the verb in C and the
Qu-morpheme in [Spec, CP].
My analysis differs from Kishimoto’s in several respects with regard to the nature
of a Qu-morpheme and with regard to what triggers Qu-morpheme movement. First
of all, I assume that a Qu-morpheme is a head and not a specifier. Specifiers in
Sinhala appear to precede their heads. Thus, if the Qu-morpheme were a specifier,
one would expect that when in the clause periphery, it would come at the beginning
of a clause. The notion that the Qu-morpheme is a head accounts straightforwardly
for its clause-final position when it is not in a TP-internal position. Also, Kishimoto
accounts for whether a Qu-morpheme appears in a TP-internal or a clause-peripheral
position, and for the appearance of the verbal ending ‘-E’ in terms of a strong or
weak [+Q] feature on a verb. As I see it, a Qu-feature is responsible for typing a
clause and is associated with a clause peripheral Typ projection, not with a verb.
My analysis accounts for the ‘E’ ending on a verb as simply being a moved Focusfeature, and whether or not a Qu-morpheme appears in a TP-internal or clauseperipheral position is dependent on whether or not a Qu-morpheme has a Focusfeature. An advantage of my analysis is that it directly connects the position of the
Qu-morpheme with the presence of a focus-interpretation when the Qu-morpheme
is in a TP-internal position, and with the lack of a focus interpretation when the
Qu-morpheme is in a clause-peripheral position.
I next turn to Okinawan. Sugahara (1996) argues that what I have been referring
to as a Qu-morpheme, ga, is actually a [+WH] element that is the head of a quantificational phrase QP. This particle ga takes as its complement a wh-phrase, which
Sugahara argues is an indeterminate pronominal that is bound by a null operator
that has a [+WH Scope] feature, and that is located in the specifier of the QP. The
structure of a wh-phrase with an adjacent Qu-morpheme is given as follows, from
186
Sugahara (1996:239).
(79)
QP
Op[+W HScope]
Q’
XP
Q
ga[+W H]
indeterminate pronominal
Furthermore, Sugahara argues that the verbal agreement ending ‘-RA’ has a [+WH
Scope] feature and is the head of a Modal Phrase (MP) that is located in a position
above TP, which accounts for why it follows tense. The null [+WH Scope] operator
in the [Spec, QP] position moves to the specifier of the modal phrase to check the
[+WH Scope] feature of ‘-RA.’
Sugahara’s analysis has several aspects that I find problematic. First of all, if
the Qu-morpheme is the head of a TP-internal QP, then it is not clear where it is
when it appears in a clause-peripheral position. Also, this analysis does not make
a connection between the fact that the TP-internal ga is homophonous with the
ga that occurs in a yes/no construction. As I see it, ga occurs in yes/no and whconstructions because it is a Qu-morpheme. It is not a wh-element. Lastly, I account
for the special verbal ending ‘-RA’ as being the pronunciation of a Focus-feature,
as ‘-RA’ occurs when there is a TP-internal Qu-morpheme adjoined to a focused
wh-phrase. There is no need to postulate movement of a null wh-operator to explain
its existence.
Miyara (2001) takes a different view of the nature of the Okinawan Qu-morpheme
ga. Miyara assumes that the clause-final ga is a Qu-morpheme, but the TP-internal
ga is an emphatic particle. Thus, there are two distinct ga elements. Furthermore,
Miyara assumes that the special verbal agreement ending ‘-RA’ that occurs when
a Qu-morpheme appears in a TP-internal position is a Qu-morpheme. This can
be seen in the glosses of the following constructions. In (80a), the clause final ga
187
is glossed as a Qu-morpheme, but in (80b), the TP-internal ga is glossed as an
emphatic particle and the ‘-RA’ ending is glossed as a Qu-morpheme.
(80)
(a) Taa-ga
ich-u-ga?
who-nom go-pres-Qu
‘Who is going?’
(b) Taa-ga-ga
ich-u-ra?
who-nom-emph go-pres-Qu
‘Who is going?’ (Miyara 2001:27)
The glossing of the TP-internal ga as an emphatic particle is in agreement with
my notion that the TP-internal Qu-morpheme has a Focus-feature, but it misses
the connection between the TP-internal and the clause-peripheral occurences of ga.
Both versions of ga serve the function of typing a clause as an interrogative, and
so I think that the Qu-morpheme label is appropriate. Furthermore, I do not think
there is any clear evidence that ‘-RA’ is a Qu-morpheme, since it only occurs when
there is a TP-internal ga, whereas ga occurs in all wh-constructions.
Hagstrom (1998) argues that in a wh-construction in wh-in-situ languages such
as Sinhala, Okinawan, Premodern Japanese, and Modern Japanese, a Qu-morpheme
is base generated in a position adjacent to a wh-phrase. From this position the Qumorpheme moves to C. This movement can be overt, in which case the Qu-morpheme
must appear in the clause periphery, or this movement can be at LF, in which case
the Qu-morpheme appears in a TP-internal position. There are two aspects of this
analysis that I think can be revised. First of all, although Hagstrom acknowledges
that a TP-internal Qu-morpheme focuses a constituent in a yes/no construction, if I
understand correctly, his analysis does not account for the focus facts. My analysis
accounts for these facts with the notion that a TP-internal Qu-morpheme contains
a Focus-feature and focuses a particular wh-phrase. Secondly, if a Qu-morpheme
is always base generated as part of a wh-phrase in a wh-construction, then the
issue arises of what happens in a yes/no construction. In yes/no constructions in
188
which there is no focused phrase, the Qu-morpheme appears in the clause-periphery.
Thus, in these cases, it seems most likely that the Qu-morpheme is base generated
in Typ, as there is no TP-internal phrase for it to be base generated next to. Under
my analysis, the wh-construction facts are identical; if there is no identificationally
focused wh-phrase, then the Qu-morpheme is base generated in Typ. Thus, it is not
clear that a Qu-morpheme must always be base generated in a position adjacent to
a wh-phrase.
Another aspect that I differ from Hagstrom on is with respect to the special verbal endings in Sinhala. Hagstrom writes that when the Sinhala Qu-morpheme remains adjoined to a wh-phrase, the ‘-E’ suffix indicates that it will move at LF. When
the Qu-morpheme moves overtly, then the ‘-E’ suffix does not occur. Hagstrom
(1998:33) writes the following.
. . . the ‘-E’ suffix is a morphological reflection of an “unsatisfied property” of the surface representation.
‘-E’ reflects a feature which is
“checked” or “satisfied” by the movement of the focus or interrogative
particle to the clause periphery. If this movement has not taken place
overtly, ‘-E’ appears indicating that this movement is “yet to ocur,” i.e.
occurs covertly.
In other words, ‘-E’ indicates that the Qu-morpheme will move at LF. If the Qumorpheme moves overtly, then the ‘-E’ suffix does not occur on the verb. Therefore,
“[t]he unsatisifed property marked by ‘-E’ is only resolved by a suffixation of the
focus/question particle to the verb (33).” As I see it, there is no LF movement
involved. The special verbal agreement ending is simply a Focus-feature that has
moved overtly from a TP-internal Qu-morpheme to Foc, where it is Spelled-Out as
‘-E.’
In summary, although the analyses discussed in this section make interesting
claims that are worthy of closer examination, they have problems. They rely on
notions that a Qu-morpheme is a specifier, null operator movement, strong and
189
weak Qu-features, and/or LF movement. My analysis does not require any of these
elements. Furthermore, these analyses do not, for the most part, account for the focusing effects involving Qu-morphemes, whereas my analysis does; a Qu-morpheme
with a Focus-feature focuses an associated phrase. Lastly, these analyses account for
Qu-morpheme positions in wh-constructions, but they do not extend to yes/no constructions. My analysis has the advantage of accounting for wh-construction facts
in the same manner as in yes/no constructions. The position of a Qu-morpheme
in a yes/no or wh-construction is dependent on whether or not it contains a Focusfeature.
5.8 Conclusion
In this chapter, I examined cross-linguistic variation in the positions of Qumorphemes in several different types of wh-constructions. In an optimal derivation
of a wh-construction, a Qu-morpheme should be Merged directly in Typ, where
its Qu-feature can type a clause without any movement. However, this ‘optimal
derivation’ clearly does not always occur, as there are constructions in which a Qumorpheme does not appear in Typ. I have accounted for this variation as primarily
resulting from whether or not a Qu-morpheme contains a Focus-feature.
I have examined the two types of languages listed in (81). A Type 1 language
requires its Qu-morpheme to be Merged directly in Typ in the clause-periphery.
A Type 2 language allows a Qu-morpheme to either be Merged in Typ or in a
TP-internal position.
(81)
Language Type
Typ
1
X
2
X
TP-internal
X
190
In the Type 2 languages that I discussed (Sinhala and Okinawan), the position of the Qu-morpheme depends on whether or not it has a Focus-feature. If
it has a Focus-feature, it must be Merged in a TP-internal position adjoined to
a wh-phrase or a constituent containing a wh-phrase. If it lacks a Focus-feature,
then it is Merged in the clause-periphery in Typ. As discussed in section 5.6,
Sumangala claims that a standard non-identificational focus interpretation of a
wh-phrase is possible even when a Qu-morpheme is adjoined to a matrix non-scalar
wh-phrase. If true, this is an exception to this last claim that a Qu-morpheme
without a Focus-feature is Merged in the clause-periphery. However, this issue
requires further examination.
In Sinhala and Okinawan, although a Qu-morpheme can appear in a TP-internal
position, it cannot occur inside of an ‘island’ because movement of its Qu- and
Focus-features cannot proceed out of the relevant clause, either because the clause
has been renumerated or because an intervening element causes an MLC violation.
However, in these languages a Qu-morpheme can be Merged in a position adjacent
to a potential island, and in this case it serves the function of focusing the island,
as well as typing the clause.
This analysis has several positive aspects. It accounts for the special focusinterpretations associated with TP-internal Qu-morphemes that occurs in whconstructions. This analysis is also compatible with the analysis of yes/no constructions in the previous chapter, and thus it accounts for similarities between
yes/no and wh-construction formation.
In this chapter, I accounted for the syntactic distribution of Qu-morphemes in
certain wh-constructions. Crucially, a wh-construction contains a wh-phrase. In the
next chapter, I account for the syntactic distribution of wh-phrases by examining
the role of another interrogative feature; a wh-feature.
191
CHAPTER 6
Wh-features in single wh-constructions
6.1 Introduction
In this chapter I focus on how a wh-feature establishes a relationship with Typ in
single wh-constructions. I argue that it is this relationship that accounts for much
of the crosslinguistic variation in wh-constructions.
A wh-construction obviously requires a wh-element. When a Qu-morpheme occurs and there is no wh-phrase, as in (1a) from Japanese, a yes/no construction
results. Example (1b) is a wh-construction because the Qu-morpheme ka co-occurs
with a wh-phrase.
(1)
Japanese:
(a) Anata-wa kono-hon-ga
suki desu ka?
you-top this-book-nom like be
Qu
‘Do you like this book?’
(b) Anata-wa nani-ga
suki desu ka?
You-nom what-nom like be
Qu
‘What do you like?’
I essentially adopt the view argued for by Katz & Postal (1964), Aoun & Li (1993),
and Denham (2000), among others, that a wh-construction contains a unique whfeature that is separate from a Qu-feature. In order for a wh-construction to be
formed, Typ must contain a Qu-feature to type the clause as an interrogative, and
in addition, a wh-feature must value a probe in Typ. I will argue that in a whconstruction, a single Typ head contains both a Qu-feature and also a probe that is
valued by a wh-feature. This probe is responsible for giving a wh-phrase scope.
192
Evidence that both a Qu-feature and a wh-feature co-occur in TypP can be seen
in languages that have both an overt Qu-morpheme and a wh-phrase that occur in
a single TypP projection. In Sharanahua,1 the Qu-morpheme mun occurs in clause
final position and a wh-phrase must occur at the beginning of a clause (Scott &
Frantz 1974), as shown in (2).
(2)
Sharanahua:
Ahuua
min rutu-a-mun?
What-thing you kill-compl-Qu
‘What did you kill?’ (Scott & Frantz 1974:84)
Similarly, Tlingit2 requires a Qu-morpheme and a wh-phrase to appear in a clauseperipheral position (Cable 2007). In (3), the wh-phrase daa ‘what’ occurs in clauseinitial position, followed by the Qu-morpheme sá ‘Qu.’
(3)
Tlingit:
Daa sá kéet
ax/’a?
what Qu killerwhale he.eats.it
‘What do killerwhales eat?’ (Cable 2007:64)
In these wh-constructions in (2) and (3), both the wh-phrase and the Qu-morpheme
are likely in TypP. The wh-phrase moves from within the TP to [Spec, TypP], and
the Qu-morpheme is in Typ. The Qu-morpheme has a Qu-feature (FQu ) and the
wh-phrase has a wh-feature (Fwh ) that values a probe in Typ, as indicated by the
Fwh
subscript on the probe Prb (see section 6.4 for further details on how the Probe
is valued). Thus, the Typ head in these constructions contains a Qu-feature and
also a probe that is valued by the wh-feature. The TypP of (2) is shown in (4), and
the TypP of (3) is shown in (5).
1
2
This a Panoan language spoken in Peru (Scott & Frantz 1974:75).
This is a “Na-Dene language of Alaska, British Columbia and the Yukon (Cable 2007:21).”
193
(4)
TypP
DP1
ahuua[Fwh ]
what-thing
Typ’
TP
Typ
Prb[Fwh ] , mun[FQu ]
. . . t1 . . .
(5)
TypP
DP1
Typ’
daa[Fwh ]
Typ
what Prb[Fwh ] ,sá[FQu ]
TP
. . . t1 . . .
I discuss the details of these languages further in section 6.4. The key point here
is that wh-constructions in these languages provide evidence that both a Qu- and a
wh-element co-occur in the same TypP projection.
In the following sections, I argue that there are three ways, listed in (6a-c), that
a wh-feature can obtain scope in Typ.
(6)
(a) Agree
(b) Wh-feature movement
(c) Wh-phrasal movement
Movement distinguishes these three relationships. A wh-feature and a probe in
Typ can form an Agree relation (6a), which crucially does not result in island or
intervention effects. A wh-feature can move to Typ (6b), or a wh-phrase can move
to [Spec, TypP] (6c). These latter movement options are subject to island and
intervention effects.
I propose that which of the three basic methods in (6a-c) is utilized in a whconstruction is dependent on the following factors: 1) the presence or absence of
194
an EPP feature that attracts a wh-element in Typ, and 2) the type of EPP feature
that occurs (when present). A Typ head may or may not contain an EPP feature
that attracts a wh-element, depending on the language. When there is no EPP
feature, or no EPP feature that attracts a wh-element, a wh-feature undergoes an
Agree relation with a probe in Typ. This relationship crucially does not involve
any movement. When there is an EPP feature in Typ, some form of movement of a
wh-element to TypP must occur. The EPP feature determines how the probe in the
scopal TypP head is valued; either by XP movement of a wh-phrase to [Spec, TypP]
or by X0 movement of a wh-feature to Typ. This notion that an EPP feature can
attract an XP or an X0 element was proposed by Alexiadou & Anagnostopoulou
(1998). This view is summarized by Miyagawa (2001) as follows.
The EPP may be satisfied either by moving an appropriate XP to the
specifier of the head with the EPP-feature or by raising an appropriate
head to the head with the EPP-feature (Miyagawa 2001:295).
I refer to an EPP feature that forces XP movement as an EPPXP feature3 and I
0
refer to an EPP feature that forces X0 movement as an EPPX feature. These three
different types of wh-construction, utilizing Agree, wh-feature movement, or whphrasal movement are summarized in (7a-c). In 7a) there either is no EPP feature,
or if an EPP feature is present, it does not attracts a wh-element.
(7)
Relationship
EPP
(a)
Agree
∅
(b)
Wh-feature movement
EPPX
(c)
Wh-phrasal movement
EPPXP
3
0
This is the similar to the traditional EPP feature of Chomsky (1981), although in accord with
Chomsky (1999, 2000), this EPP feature is associated with a projection other than T.
195
Example (8) shows an Agree relation between a probe Prb in Typ and the whfeature of an in-situ wh-phrase located within the TP. The probe is valued by the
wh-feature of the wh-phrase, and there is no movement of a wh-element. The Agree
relation is represented by the matching Fwh1 subscripts on the probe in Typ and on
the wh-phrase within the TP.
(8)
TypP
Typ’
TP
Typ
Prb[Fwh1 ]
. . . wh-phrase[Fwh1 ] . . .
0
Example (9) shows wh-feature movement. An EPPX feature located in Typ
attracts the wh-feature of an in-situ wh-phrase, and the wh-feature values a probe,
if present,4 as shown in (9).
(9)
TypP
Typ’
TP
TypEP P X 0
Prb+Fwh1
. . . wh-phrase+tFwh1 . . .
The wh-feature is an element that I assume is contained within the head of a whphrase. So if a wh-phrase is a DP, then the wh-feature is located within the D head,
and when the wh-feature moves, it moves from this position, as shown below.
4
A probe need not be present, such as when a wh-feature moves through an embedded clause
on its way to a higher clause. In this case, I assume that the embedded clause has an EPP feature,
but it does not have a probe. The probe is located in the scopal TypP head.
196
(10)
DP
D’
D
Fwh1
Example (11) shows wh-phrasal movement. In this case, an EPPXP feature
forces a wh-phrase to move to [Spec, TypP]. Then, the wh-feature associated with
the wh-phrase values a probe, if present, via a Spec-head relation.
(11)
TypP
DP1
Typ’
wh-phrase[Fwh ] TP
TypEP P XP
Prb[Fwh ]
. . . t1 . . .
A crucial claim in the following sections is that Agree is not subject to intervention and island effects, whereas movement, whether it be feature movement or
phrasal movement, is subject to these effects.5 Here, I take the position that an EPP
feature is sensitive to feature type, whereas Agree is only sensitive to feature content.6 Specifically, an EPP feature in Typ attracts the closest scope bearing element;
generally quantificational and/or focus elements. Thus, if a particular scope bearing element that is not a wh-feature intervenes between the EPP and a wh-element,
attraction of the wh-element is blocked. This exact nature of these blocking effects
is not entirely clear to me, but it is most likely that Ill-formedness results for one
of the following reasons: 1) a feature associated with the intervening scope bearing
element raises to Typ, but, not being a wh-feature, it is not compatible with the
5
As discussed in chapter 2.2.4, Chomsky (1999, 2000), among others, takes the position that
Agree is subject to blocking effects.
6
I thank Heidi Harley (p.c.) for suggesting this to me.
197
probe, 2) the wh-feature is unable to value the probe in Typ since it is not attracted
by the EPP, and/or 3) the non-wh-quantificational feature is attracted by the EPP
feature, but it is unable to move, since it is in a position where it already values
an uninterpretable feature. Agree, on the other hand, results when a probe simply
searches for a feature of the appropriate content, which in the relevant cases discussed in this chapter, is a wh-feature. Any other features of the same type, i.e.,
have a scope bearing quality, but which lack a wh-feature, are invisible to an Agree
relation, in the relevant cases discussed here. Agree may even hold when a whphrase is contained within a potential island. Thus, even if a wh-phrase is contained
within a renumerated clause, there can be an Agree relation between the wh-feature
of the wh-phrase and a probe in Typ because Agree does not involve any movement
out of the clause. I elaborate on these distinctions between Agree and movement in
the following sections.
In the following sections, I examine how Agree, wh-feature movement and whphrasal movement are utilized to establish a relationship between Typ and a whfeature. In section 6.2, I argue that an Agree relation occurs between Typ and
a wh-feature in certain wh-in-situ languages. Section 6.3 discusses languages in
which a wh-phrase remains in-situ but its wh-feature moves. Section 6.4 focuses on
languages in which there is wh-phrasal movement. Section 6.5 discusses languages
that allow a wh-phrase either to undergo full wh-movement to a scopal TypP or
partial wh-movement to a non-scopal TypP. Section 6.6 discusses Malay, a language
which allows Agree, full wh-movement, and partial wh-movement. Section 6.7 is
about French, a language which utilizes both overt wh-phrasal movement and whfeature movement. Section 6.8 concludes the chapter.
6.2 Agree
In this section, I argue that in single wh-constructions in Sinhala, Okinawan, and
Mandarin, there is no movement of a wh-element because Typ does not contain an
198
EPP feature that attracts a wh-element. Neither the wh-feature nor its associated
wh-phrase undergo any movement, and so there are no island- or intervention-effects
caused by wh-feature movement. There is, however, an Agree relation that is formed
between the probe in Typ and a wh-feature. This Agree relation gives the wh-phrase
scope and, together with a Qu-feature, types a clause as a wh-construction.
The following examples demonstrate that Sinhala is a wh-in-situ language. Example (12a) is a typical SOV construction. Example (12b) is a corresponding whconstruction in which the object has been replaced with the wh-phrase [email protected]
‘what.’ Note that the object wh-phrase obtains scope but does not undergo any
movement; it occurs in the typical preverbal position.
(12) Sinhala:
(a) Chitra [email protected] gatta.
Chitra book bought.A
Chitra bought the book.
(b) Chitra [email protected] [email protected] gatte
Chitra what
Qu bought.E
What did Chitra buy? (Kishimoto 2005:3-4)
Examples (13a-b) demonstrate that Okinawan is also a wh-in-situ language. In
(13a), the object shishi ‘meat’ occurs in pre-verbal position where it recieves its
theta-role in this SOV language, and in the wh-construction (13b), the object whphrase nuu ‘what’ occurs in this same position.
(13) Okinawan:
(a) Taru-ya
shishi kamyi-N
Taru-top meat
eat-decl
‘Taru eats meat.’
199
(b) Taru-ya
nuu kamyi-ga
Taru-top what ate-Qu
‘What does Taru eat?’ (Sugahara 1996:236)
Examples (14a-b) demonstrate that Mandarin is a wh-in-situ language. In (14a)
the object follows the verb and when the object is a wh-phrase, as in (14b), it
occurs in the same post-verbal position in this SVO language, thereby resulting in
a wh-construction. The wh-phrase then does not need to move.
(14) Mandarin:
(a) Ta pian-le
Lisi.
he cheat-asp Lisi
‘He cheated Lisi.’ (Huang 1982:27)
(b) Ni kanjian-le shei?
you see-asp
who
‘Who did you see?’ (Huang 1982:253)
In these languages, a wh-phrase clearly does not need to move at Spell-Out.
Below I present evidence that not only does the wh-phrase not move, but the whfeature does not move. Rather, the wh-feature establishes an Agree relation with a
probe in Typ.
As discussed in chapter 5, a Qu-morpheme in a wh-construction in Sinhala and
Okinawan can appear in a TP-internal position adjoined to a wh-phrase, as long as
the wh-phrase is not contained within an island. When the Qu-morpheme is at the
edge of an island the result is well-formed. This can be seen in the Sinhala examples
(15a-b) and the Okinawan (16a-b) (these examples were originally presented as
(59a-b) and (62a-b) in chapter 5).
(15) Sinhala
(a) *Oyaa [DP [T ypP Chitra kaa-ú@ [email protected] [email protected]] [email protected]] kieuwe?
you
Chitra who-dat Qu gave
book read.E
Intended: ‘[To whom]1 did You read [the book that Chitra gave t1 ]?
200
(b) Oyaa [DP [T ypP Chitra kaa-ú@ [email protected]] [email protected]] [email protected] kieuwe?
you
Chitra who-dat gave
book Qu read.E
‘[To whom]1 did You read [the book that Chitra gave t1 ]? (Kishimoto
2005:29)
(16) Okinawan:
(a) *[DP [T ypP taa-ga-ga
ka-chee-ru]
shimuchi]
who-nom-Qu write-have-nom book(-acc)-top
John-oo
yuda-ra?
John-TOP read-RA
Intended: ‘[Who the hell]1 did John read [the book that t1 had
written]?’
(b) [DP [T ypP taa-ga
ka-chee-ru]
shimuchi]-ga
who-nom write-have-nom book(-acc)-top-ga
John-oo
yuda-ra?
John-TOP read-RA
‘[Who the hell]1 did John read [the book that t1 had written]?’ (Miyara
2001:52)
In examples (15-16a), the Qu-morpheme is contained within a complex-DP, resulting
in ill-formedness. In chapter 5, I argued that in Sinhala and Okinawan, there is
movement of Qu- and Focus-features from a Qu-morpheme and that this movement
is blocked when it proceeds from within an adjunct clause (due to renumeration),
or due to the MLC when it proceeds from a whether/if -clause. When the Qumorpheme is outside the complex-DP as in (15-16b), the result is well-formed. The
features associated with the Qu-morpheme are able to proceed uninhibited to their
final landing sites. A wh-feature, on the other hand, is a different element from
these Qu- and Focus-features.
In constructions such as (15-16b), the wh-phrases are able to obtain scope despite
the fact that they are contained within complex-DPs. If the wh-feature associated
with a wh-phrase undergoes movement, then this movement should result in islandeffects. Because of the lack of island effects, I propose that the wh-feature does
201
not move, but instead, forms an Agree relation with a probe in Typ. This lack of
movement is the result of there not being an EPP feature in Typ that attracts the
wh-feature; i.e., there is either no EPP feature, or if there is an EPP feature, it does
not attract a wh-element. Rather, a probe in Typ forms an Agree relation with the
wh-feature.7 This Agree relation is able to hold over potential island boundaries,
because there is no movement out of the ‘island.’ A simplified diagram of the wellformed constructions in (15-16b) is shown below in (17).
(17)
TypP
Typ’
FocP
...
Typ
Prb[Fwh1 ] ,FQu3
Foc
FF oc2
TypP
TypP
Qu+t[FF oc2] +t[FQu3]
. . . wh-phrase[Fwh1 ] . . .
The probe in Typ Agrees with the wh-feature, as signified by the Fwh1 subscripts
on the probe Prb in Typ and the in-situ wh-phrase. Following the discussion of
TP-internal Qu-morphemes in chapter 5, I assume that Qu- and Focus-features
associated with the TP-internal Qu-morpheme raise to Typ and Foc respectively.
The Focus-feature in Foc focuses the embedded clause and the Qu-morpheme types
the clause as an interrogative.
7
In these cases, there may be an EPP feature in Typ, but one that does not attract a wh-feature.
For example, in chapter 5, I argued that an EPP feature in Typ in Sinhala and Okinawan whconstructions can attract a TP-internal Qu-feature. This EPP feature crucially does not attract a
wh-feature.
202
I next turn to Mandarin Chinese. In chapter 5, I argued that the Qu-morpheme
in Mandarin, which either surfaces as ne or is null, is Merged directly in Typ. If this
is the case, then features associated with the Qu-morpheme cannot be responsible
for any intervention or island effects. This is because Mandarin’s Qu-morpheme
is always Merged in a position external to an island. Therefore, examination of
intervention and island effects should shed light on the behavior of the wh-feature.
If these effects are absent, as appears to be the case,8 then the argument that the
wh-feature does not move is supported.
In Mandarin, a wh-phrase may be c-commanded by a quantificational element
without any resulting ill-formedness. This shows that Mandarin wh-phrases are not
subject to intervention effects (see chapter 2.7 for discussion of intervention effects).
In the well-formed (18), the quantifer zhi ‘only’ or negation bu ‘not’ c-commands
the wh-argument shenme ‘what.’
(18) Mandarin:
Ta {zhi/bu} mai shenme?
he only/not sell what
‘What is the thing x such that he {only sells/does not sell} x?’ (Soh
2005:147)
Even though a quantificational element intervenes between Typ and the wh-phrase,
the result is well-formed. If the wh-feature were to move, then this movement should
be blocked by a quantificational feature associated with zhi ‘only’ or bu ‘not,’ as is
the case in languages such as Japanese (see section 6.3).9
Similarly, there are generally no island effects in this language. The following
examples in (19) demonstrate that the wh-phrases shei ‘who’ and shenme ‘what’
8
Note though that weishenme ‘why’ is an exception. It is subject to intervention- and island-
effects. I return to this issue in chapter 9.
9
Although I do not have conclusive data on intervention effects in Sinhala and Okinawan, my
analysis predicts that a wh-phrase should not be subject to intervention effects in these languages.
203
can occur within a whether/if -clause (19a), a complex-DP (19b), an adjunct clause
(19c), and a complex subject (19d).
(19) Mandarin:
(a) No whether/if island effect:
Ni xiang-xhidao [ta xi-bu-xihuan
shei]?
you wonder
he whether-or-not like who
‘Who1 do you wonder [whether he likes t1 ]?’ (Lasnik & Saito 1992:32)
(b) No complex-DP island effect:
Ni xihuan [DP shei xie
de
shu]?
you like
who write gen book
‘Who1 do you like [the book t1 wrote]?’
(c) No adjunct island effect:
Ta [T ypP yinwei ni shuo shenme hua] hen shengqi?
he
because you say what
word very angry
‘What1 was he angry [because you said t1 words]?’(Aoun & Li 1993:203)
(d) No subject island effect:
[DP shei xie
de
shu] zui
youqu?
who wrote gen book most interesting
‘Who1 are [books that t1 wrote] most interesting?’ (Lasnik & Saito
1992:122)
Examples (20a-c) below show that the wh-adjuncts sheme yuanyin ‘what reason,’
nali ‘where,’ and shemesihou ‘when’ also are not subject complex-DP island effects.
(20) Mandarin:
(a) [DP Ta we-le sheme yuanyin xie
de
shu] zui
he for
what
reason
write gen book most
youqu?
interesting
204
‘[For what reason]1 are [books that he wrote t1 ] most interesting?’
(Huang 1982:527)
(b) [DP Ta zai nali pai de
dianying] zui
hao?
he at where film gen movie
most good
‘Where1 are [movies that he filmed t1 ] the best?’
hao?
dianying] zui
(c) [DP Ta (zai) shemeshihou pai de
most good
film gen movie
he (at) when
‘When1 are [movies that he filmed t1 ] the best?’ (Huang 1982:529-530)
The lack of intervention and island effects suggests that in Mandarin, a whfeature associated with a wh-phrase does not move. A probe in Typ forms an Agree
relation with the wh-feature of a wh-phrase, as shown in (21).
(21)
TypP
Typ’
TP
Typ
Prb[Fwh1 ] , FQu
. . . wh-phrase[Fwh1 ] . . .
Since this is an Agree relation, it does not matter if the wh-phrase is c-commanded
by a scope bearing element or contained within a potential island.
In summary, Sinhala, Okinawan, and Mandarin are languages in which the whfeature of a wh-phrase does not move. Rather, it establishes scope by forming an
Agree relation with a probe in Typ. Evidence for Agree is that in these languages,
a wh-phrase can occur inside of a potential island. In Mandarin, there also are no
intervention effects.
6.3 Wh-feature movement
I next turn to single wh-constructions in which a wh-feature undergoes feature movement. In languages such as Japanese, Korean, and Persian, I propose that the rel-
205
0
evant Typ head has an EPPX feature that attracts an X0 category wh-feature so
that it can value a probe in Typ.
Japanese, Korean, and Persian all allow a wh-phrase to remain in-situ.10 Examples (22a-b) show that Japanese is a wh-in-situ language. In (22a), the object
hon-o ‘book-ACC’ occurs in the preverbal position in this SOV language, and in
(22b), the wh-phrasal object nani-o ‘what-ACC’ occurs in this same position. The
wh-phrase does not move, but a wh-question is formed.
(22) Japanese:
(a) Hideya-wa hon-o
katta.
Hideya-top book-acc bought.
‘Hideya bought a book.
(b) Hideya-wa nani-o
katta
no?
Hideya-top what-acc bought Qu.
‘What did Hideya buy?’
Similarly, a wh-phrase in Korean may remain in-situ. Example (23a) shows a declarative construction in this SOV language. Note that the object kae-lul ‘dog-acc’
directly precedes the verb. In the wh-construction (23b), the wh-phrasal object
muôs-ûl ‘what-acc’ remains in the preverbal position.
(23) Korean:
(a) Suna-ka
chaeg-ul
sa-ss-ta.
Suna-nom book-acc buy-past-decl
‘Suna bought a book.’ (Hyun Kyoung Jung, p.c.)
(b) Suna-ka muôs-ûl sa-ss-ni?
Sun-nom what-acc buy-past-Qu
‘What did Suna buy?’ (Beck & Kim 1997:339)
10
All of these languages allow scrambling, but in an unscrambled constructions, a wh-phrase
remains in-situ.
206
Persian, similarly is an SOV language in which a wh-phrase can remain in-situ.
In (24a), the object ketâb-ro ‘book-acc’ directly precedes the verb. In (24b) the
wh-phrasal object chi-ro ‘what-acc’ appears in this same preverbal position and a
wh-construction results.
(24) Persian:
(a) Kimea un ketâb-ro xund.
Kimea that book-acc read
‘Kimea read that book.’ (Karimi 2005:27)
(b) Kimea chi-ro
xund?
Kimea what-acc read
‘Kimea read that book.’ (Simin Karimi, p.c.)
In these languages, although the wh-phrase remains in-situ, there is evidence that
its wh-feature undergoes movement.
First of all, in all of these languages a Qu-morpheme appears to be Merged
directly in Typ. In chapter 5, I argued that because the Japanese Qu-morpheme
must always appear in clause-final position, it is Merged directly in Typ of the clause
that it types. Korean also requires a clause-final Qu-morpheme. Persian does not
generally have an overt Qu-morpheme in a matrix clause; although an interrogative
interpretation is indicated via rising intonation at the end of a clause (Simin Karimi,
p.c.).11 In these cases, I take the position that a null Qu-morpheme is present. I
know of no evidence that a Qu-morpheme can be Merged TP-internally in Korean
or Persian. If these languages have only a Qu-morpheme that is Merged directly in
Typ, then any intervention or island effects must result from something other than
Qu-feature movement.
The following data show that Japanese, Korean, and Persian have intervention
effects. In the Japanese (25a), the Negative Polarity Item (NPI) element sika ‘only’
11
In formal Persian, there is a Qu-morpheme āya that appears clause-initially in yes/no con-
structions (Simin Karimi, p.c.).
207
c-commands the wh-phrase nani-o ‘what-ACC,’ resulting in ill-formedness. In (25b)
the wh-phrase is scrambled to a position higher than the NPI and the result is wellformed.
(25) Japanese:
(a) ?*Taroo-sika nani-o
yoma-nai no?
Taroo-only
what-acc read-NEG Qu
‘What did only Taro read?’ (Tanaka 1997:159)
(b) Nani-o Taroo-sika t1 yoma-nai no?
what-acc Taroo-only
read-NEG Qu
‘What did only Taro read?’ (Tanaka 1997:162)
Intervention effects in Korean are shown in (26a-b).12 When the wh-phrase muôs-ûl
‘what-acc’ is c-commanded by amuto ‘anyone’ the result is ill-formedness, as shown
in (26a). When the wh-phrase is scrambled to clause initial position, as shown in
(26b), the result is well-formed.
(26) Korean:
(a) *Amuto muôs-ûl sa-chi
anh-ass-ni?
anyone
what-acc buy-CHI not do-past-Qu
(b) Muôs-ûl1 amuto t1 sa-chi
anh-ass-ni?
what-acc anyone
buy-CHI not do-past-Qu
‘What did no one buy?’ (Beck & Kim 1997:339)
Similarly, in the Persian (27a), the NPI hichkas ‘nobody’ c-commands the wh-phrase
chi-ro ‘what,’ resulting in ill-formedness.13 When the wh-phrase is scrambled to a
position above hichkas ‘nobody’ in (27b), the result is well-formed.
12
Following Beck & Kim (1997) I gloss chi as ‘CHI.’ Beck and Kim write that “[t]he status of
the verbal suffix chi is not clear (381)” and that it has been argued to be a nominalizer by some
and a complementizer by others.
13
This is well-formed as an echo-question (Simin Karimi, p.c.).
208
(27) Persian:
(a) *Hichkas chi-ro
na-xarid?
nobody
what-acc neg-bought
(b) Chi-ro1 hichkas t1 na-xarid?
what-acc nobody
neg-bought
‘What was it that no one bought?’ (Karimi & Taleghani 2007:180)
I turn to an explanation of the intervention effects and of the amelioration of these
effects resulting from scrambling.
The notion that there are intervention effects, at least in Japanese and Korean,
is not uncontroversial. Tomioka points out that there is a great deal of variability in
well-formedness judgments regarding intervention effects and intervention effects are
weaker in embedded clauses than in matrix clauses. However, Tomioka acknowledges
that scrambling of a wh-phrase over a potential intervener is better than leaving the
wh-phrase in a position c-commanded by an intervener, even in embedded clauses
where intervention effects are weaker. Tomioka (2007:1572) writes the following,
where ‘unscrambled’ refers to a construction in which an intervener c-commands a
wh-phrase.
Although there is no denying that the scrambled version is better than
the unscrambled counterpart, some feel that the unscrambeled examples
are merely marginal, while others find them bad enough to label them
as ‘ungrammatical.’
Even though there is variability in judgments and there are differences between
matrix and embedded clauses, the facts still point to the existence of intervention
effects. I leave aside the interesting complexities involving judgments of intervention
effects, although see Tomioka (2007) for a pragmatic account of the intervention
effects in Japanese anbid Korean.
209
I propose that the intervention effects in (25-27a) are the result of the blocking
of wh-feature movement. The motivation for wh-feature movement is an EPPX
0
0
feature in Typ that attracts a wh-feature. However, the EPPX feature is sensitive
to ‘type’ - it searches for a wh-feature but if there is a close element of the same
type, then this attraction is blocked. When there is a non-wh-scope bearing element
such as an NPI closer to Typ than the wh-feature, then attraction of the wh-feature
is blocked.14 As a result, the probe in Typ does not get valued by the wh-feature.
Example (28) shows the simplified structure of (25-27a). The Japanese sika ‘only,’
the Korean amuto ‘anyone,’ and the Persian hichkas ‘nobody’ are in the specifier or
what I loosely refer to as QuantP, to indicate the quantificational nature of these
0
elements. The EPPX is unable to attract the wh-feature due to the intervening
quantificational feature.
14
Intervention effects appear to be caused by a class of scope bearing elements that have a
quantificational and/or focus property. See chapter 2.7 for discussion of the class of elements that
cause intervention effects. For the sake of simplicity I refer to these intervening elements as scope
bearing and I place them in a projection that I refer to as a QuantP, where Quant refers to the
quality of quantification. But note that quantificational alone is not necessarily an indicator of the
relevant type of feature that causes intervention effects.
210
(28)
TypP
Typ’
QuantP
TypEP P X 0
Prb, Qu[FQu ]
Quant’
TP
Quant
FQuant1
Taroo-sika/amuto/hichkas+tFQuant1
Taro-only/anyone/nobody
T’
*
vP
. . . wh-phrase
T
[Fwh ]
...
In (25-27b), the wh-phrase appears in a scrambled position higher than the
0
quantificational element. In this case, the EPPX feature in Typ is able to attract
the wh-feature since there is no intervening element. A diagram is shown in (29). I
have indicated the wh-phrase as being in the specifier of a FocP projection, which I
assume houses a scrambled phrase.15
15
I leave aside discussion of the exact nature of the projection housing a scrambled phrase. For
discussion of scrambling see Saito (1985), Mahajan (1990), Saito & Fukui (1998), Bos̆ković &
Takahashi (1998), Miyagawa (1997a), Miyagawa (2001), and Karimi (2005), among others.
211
(29)
TypP
Typ’
FocP
TypEP P X 0
Prb+Fwh2 ,Qu[FQu ]
Foc’
DP1
wh-phrase+t[Fwh2 ] QuantP
Foc
Quant’
TP
Quant
FQuant
. . . t1 . . .
I next turn to island effects in these languages. In Japanese there appear to be
island effects when a wh-phrase occurs inside of an embedded whether/if construction (see chapter 2.6 for discussion of whether/if islands). In the Japanese example
(30), the wh-phrasal DP nani-o ‘what-acc’ can have scope in the embedded clause,
as indicated by the well-formed meaning (i), but according to Nishigauchi (1990), it
cannot have scope in the matrix clause, as shown by the ill-formedness of meaning
(ii).
(30) Japanese:
Satookun-wa [Suzukikun-ga nani-o
tabeta ka] oboeteimasu ka?
Sato-top
Suzuki-nom what-acc ate
Qu remember
Qu
(i) ‘Does Sato remember [what Suzuki ate]?’
(ii) *‘What1 does Sato remember [whether Suzuki ate t1 ]?’ (Nishigauchi
1990:30-31)
Example (31) is similar except that the embedded clause contains the Qu-morpheme
kadooka, which notably only occurs in embedded yes/no constructions; it is not
212
compatible with a wh-phrase (see chapter 3.3). Since the wh-phrase in the embedded clause cannot have embedded scope and is unable to have matrix scope,
ill-formedness results.
(31) Japanese:
*Satookun-wa [Suzukikun-ga nani-o
tabeta kadooka] oboeteimasu
Sato-top
Suzuki-nom what-acc ate
whether
remember
ka?
Qu
“What1 does Sato remember [whether Suzuki ate t1 ]?’ (Nishigauchi 1990:31)
The assumption that there are island effects of the sort shown in (30) and (31) is
in accord with Nishigauchi (1990), Watanabe (1992a, 1992b), and Richards (1998,
2001), among others. However, I note that the existence of this type of island effect
(generally referred to as a wh-island effect in the literature) is controversial. Watanabe (1992a:262), who claims that island effects exist in Japanese, acknowledges that
the “degree of unacceptability varies among different speakers (262).” Nishigauchi
(1990), who also claims that island effects exist, states that island effects can be
circumvented by heavy stress on a wh-phrase. Nishigauchi concludes that “a WHexpression which receives extra focus . . . is capable of taking wide scope, in violation
of the WH-island effect (35).” He attributes this fact to something other than an
island effect. Kitagawa (2005, 2006), Kitagawa & Fodor (2006), and Kitagawa et al.
(2004) present arguments that when spoken with the ‘correct prosody,’ there are
no wh-island effects in Japanese. With the acknowledgment that the issue of island
effects in Japanese is not clear, I proceed with the assumption that they do exist
for some speakers.
Korean also shows whether/if -island effects. According to Yoon (2006), the
Korean example (32) is fine with meaning (i) as a yes/no construction, in which
case mwues-ul ‘what-acc’ has embedded scope and functions as an indefinite. But
the wh-construction interpretation in (ii) is ill-formed, indicating an island effect.
213
(32) Korean:
Ne-nun [Yenghi-ka mwues-ul sa-ess-nunci]
al-ko sip-ni?
you-top Yenghi-nom what-acc buy-past-whether want to know-Qu
(i) ‘Do you want to know [whether Yenghi bought something]?’
(ii) *‘What1 do you want to know [whether Yenghi bought t1 ]?’ (Yoon
2006:387)
Note that in Japanese and Korean, wh-phrases are not generally subject to
Complex DP-island effects. I return to this issue in chapter 7.
The Persian examples (33a-c) show that a wh-phrase cannot be contained within
an adjunct clause (33a), a complex-DP (33b), or a whether/if -clause (33c).
(33) Persian:
(a) Adjunct island:
*Parviz raghsid
[chonke ki
unjâ bud]?
Parviz dance-3sg because who there was
Intended: ‘Who1 did Parviz dance [because t1 was there]?’ (Karimi &
Taleghani 2007:180)
(b) Complex-DP island:
*Kimea [DP pesar-i-ro
[ke to diruz
kojaa did-i]] be
Kimea
boy-rel-raa that you yesterday where saw-2sg to
man moarrefi
kard?
me introduced did
Intended: ‘Where1 did Kimea introduce to me [the boy you saw t1
yesterday]?’ (Simin Karimi, p.c.)
(c) Whether/if -island:
Kimea az Parviz porsid [ke Haesan chi xarid]?
Kimea of Parviz asked that Hassan what bought
Intended *‘What1 did Parviz ask Kimea [whether Hassan bought t1 ]?’
(Simin Karimi, p.c.)
214
Examples (33a-b) are fine as echo questions, but not as wh-constructions, and (33c)
is fine as a yes/no construction, but ill-formed as a wh-construction (Simin Karimi,
p.c.).
The existence of island effects in Japanese, Korean, and Persian suggests that
0
the wh-feature undergoes movement. I propose that the EPPX feature in Typ
attracts the wh-feature of the wh-phrase, but this attraction is blocked by whether/if,
due to an MLC violation. For example, in (34), ka(dooka)/nunci/ke ‘whether’ in
0
Japanese, Korean, and Persian, respectively prevents the EPPX in the matrix Typ
from attracting the wh-feature of the wh-phrase contained within the embedded
clause.
(34)
TypP
Typ’
...
TypEP P X 0
Prb
TypP
Typ’
TP
Typ
ka(dooka)/nunci/ke
. . . wh-phrase[Fwh ] . . .
The adjunct-island and complex-DP island effects in Persian result from renumeration of the adjunct-clause containing the wh-phrase. Once the adjunct clause is
assembled and placed back into the numeration, a wh-feature cannot move out of
it. The numerations of (33a-b) are shown below in (35a-b) respectively. The adjunct clauses are linearized constituents that have been renumerated, and as such,
a wh-feature cannot move out of them.
(35)
(a) N = {Parviz, raghsid, <chonke ki unjâ bud>}
215
(b) N = {Kimea, pesar-i-ro, <ke to diruz kojaa did-i>, be, man, moarrefi,
kard}
Note that this analysis shares similarities with Maki (1995) and Maki & Ochi
(1998) who argue that in Japanese wh-constructions, there is movement of a whfeature to a [+Q] Comp position. Maki (1995) argues that wh-feature movement
occurs at LF; however, Maki & Ochi (1998) argue, as I do, that wh-feature movement
occurs before LF, and that this wh-feature movement is subject to certain island
effects. The key point that we differ on is what drives wh-feature movement. I claim
0
that an EPPX feature forces wh-feature movement, and they claim that movement
is driven by the need to “delete the strong feature of the COMP (Maki & Ochi
1998:496).” An advantage of my analysis is that it does not rely on the notion that
COMP contains a strong feature.
In summary, in languages such as Japanese, Korean, and Persian, a Typ head
0
contains an EPPX feature that attracts a wh-feature to Typ, where it values a
probe. Intervention effects arise when there is an intervening element of the same
0
type, due to the inability of the EPPX feature to attract the wh-feature. Furthermore, a wh-feature cannot move out of a renumerated clause.
6.4 Wh-phrasal movement
Some languages require a wh-phrase in a simple wh-construction to move to a clause
peripheral position, a position I assume to be [Spec, TypP]. In this case, Typ contains
an EPPXP feature that attracts a wh-phrase to the [Spec, TypP] position. If this
is a scopal TypP position, then there is a probe that the wh-feature values via a
Spec-head relation.
English is a wh-movement language; a wh-phrase in a single wh-construction
cannot remain in-situ. Example (36a), although fine as an echo question, is illformed as a standard wh-construction. (36b) in which the wh-phrase moves to the
216
front of the construction is the required form.
(36)
(a) *You bought what?
(b) What1 did you buy t1 ?
Movement of a wh-phrase is accounted for as follows. The English Typ head
contains an EPPXP feature that attracts a wh-phrase to [Spec, TypP]. The whfeature associated with the wh-phrase then values the probe in Typ via a Spec-head
relation.
(37)
TypP
DP1
wh-phrase[F wh1] TypEP P XP
Prb[F wh1]
Typ’
TP
. . . t1
Unlike in languages such as Japanese, Korean, and Persian, as shown in (2527a), in English, a scope bearing element can c-command the base position of a
wh-phrase, as shown in (38a-b).
(38)
(a) What1 did only Bob read t1 ?
(b) What1 did nobody buy t1 ?
The well-formedness here results results from phrasal movement. I assume that
a quantificational feature associated with only or nobody occurs in Quant. This
quantificational feature can block X0 movement of a wh-feature, as in Japanese
and Korean, but it will not block XP movement of a wh-phrase. A diagram of
(38a-b) is shown below. The quantificational element only/nobody occurs in [Spec,
QuantP].16 I propose that the quantifier feature associated with only/nobody checks
16
It is likely that only is base-generated in this position, but nobody, being a subject, moves to
this position.
217
an uninterpretable feature in Quant. Thus, the quantifier feature is associated with
the head of QuantP and cannot block movement of an XP element.
(39)
TypP
DP1
Typ’
what[F wh1] TypEP P XP
Prb[Fwh1 ] ,∅[FQu ] +did
QuantP
DP
Quant’
only/nobody1 Quant
QuantF1
TP
. . . t1
I argued in the previous section that wh-feature movement is subject to island effects. Wh-phrasal movement is also subject to island effects. As is well
known, in English movement of a wh-phrase out of certain types of clauses results
in island-effects.17 The following examples (originally presented as (29) in chapter 2.6) demonstrate that in English, a wh-phrase cannot move out of a wh-island,
whether/if island, complex-DP-island, adjunct-island, or subject-island.
(40)
(a) Wh-island:
*What1 do you wonder [who bought t1 ]? (Huang 1982:456)
(b) Whether/If island
??What1 do you wonder [whether/if John saw t1 ]? (adapted from
Lasnik & Saito 1992:11)
17
I leave aside D-linked wh-phrases. See Pesetsky (1987, 2000), among others.
218
(c) Complex NP/DP-island:
??What1 did you read [a report [that John bought t1 ]]? (Lasnik &
Saito 1992:12)
(d) Adjunct island:
*Which book1 did John go to class [after he read t1 ]? (Lasnik & Saito
1992:12)
(e) Subject island:
?*Who1 did [a story about t1 ] amuse you? (Lasnik & Saito 1992:42)
The existence of island effects is a straightforward result of wh-movement.
The causes of these island violations, though, is not so straightforward (see chapter 2.6). The wh-island effect in (40a) can be attributed to an MLC violation. Those
in (40c-e) can be attributed to renumeration of a non-complement. In (40c) the adjunct clause modifier of the nominal report is sent to Spell-Out and renumerated.
In (40d), the adjunct clause after he read which book is renumerated and in (40e),
the subject clause is renumerated. As discussed in chapter 2.6.2, the whether/if island effect in (40b) is not easy to account for. If whether and if are heads, then
there is no MLC violation, and if whether/if John saw what is a complement, then
this clause should not be renumerated. On the other hand, if whether/if John saw
what is an adjunct clause, then the island effect is accounted for, since the adjunct
clause would have to be renumerated. This latter possibility accounts nicely for the
island effect here. However, wonder appears to select for an interrogative clause,
suggesting that the whether/if clause is a complement. Thus, I leave this issue for
further investigation.
Sharanahua is another language with overt wh-movement.
Scott & Frantz
(1974:84) write that in this language, “the interrogative word of content questions
obligatorily occurs in sentence-initial position.” In (41a), the object chasho ‘deer’
appears in clause-second position in this SOV language. However, in (41b), the ob-
219
ject wh-phrase ahuua ‘what-thing’ occurs in clause-initial position. It has undergone
movement.
(41) Sharanahua:
(a) Min chasho rutu-a-quin.
you deer
kill-compl-decl
‘You killed a deer.’
(b) Ahuua
min rutu-a-mun?
What-thing you kill-compl-Qu
‘What did you kill?’ (Scott & Frantz 1974:84)
A Sharanahua wh-construction also contains an overt clause-final Qu-morpheme
mun, as can be seen in (41b). Evidence that mun is a Qu-morpheme can be seen in
(42). The statement in (42a) becomes a yes/no construction when the clause-final
declarative ending quin is replaced with mun ‘Qu,’ as shown in (42b).
(42) Sharanahua:
(a) Min chasho rutu-a-quin.
you deer
kill-compl-decl
‘You killed a deer.’
(b) Min chasho rutu-a-mun?
you deer
kill-compl-Qu
‘Did you kill a deer.’ (Frantz 1973:532)
The morpheme mun is responsible for the interrogative interpretation of this construction because it is a Qu-morpheme.
Sharanahua then is a language in which one can see both an overt Qu-morpheme
and overt wh-movement. A construction such as (41b) is accounted for as shown in
(43).
220
(43)
TypP
DP1
ahuua[FW h1 ]
‘what-thing’
Typ’
TP
TypEP P XP
Prb[FW h1 ] ,mun[FQu ]
min t1 rutu-a
‘you t1 kill-compl’
The Qu-morpheme mun is in Typ. The EPPXP feature in Typ forces the wh-phrase
ahuua ‘what-thing’ to move to [Spec, TypP]. The wh-feature associated with the
wh-phrase values the probe in Typ via a Spec-head relation.
Tlingit is also a language with overt wh-movement, and like Sharanahua it has
an overt Qu-morpheme. For example, in (44), the Qu-morpheme sá follows the
clause-initial wh-phrase daa.
(44) Tlingit:
Daa sá kéet
ax/’a?
what Qu killerwhale he.eats.it
‘What do killerwhales eat?’ (Cable 2007:64)
The wh-phrase must appear in a pre-predicate position. Cable (2007:62) writes that
“[i]n a Tlingit wh-question . . . the phrase understood to be the wh-operator must
appear left of the main predicate of the clause.” This fact is accounted for if the
Typ head has an EPPXP feature that forces movement of the wh-phrase to [Spec,
TypP]. Example (44) is accounted for as shown in (45).
221
(45)
TypP
DP1
Typ’
daa[Fwh1 ] TypEP P XP
‘what’ Prb[Fwh1 ] ,sá[FQu ]
TP
. . . t1 . . .
In this section, I have proposed that in wh-constructions in languages such as
English, Sharanahua, and Tlingit, the Typ head contains an EPPXP feature that
forces a wh-phrase to move to [Spec, TypP]. This wh-phrasal movement can result
in island effects, as is the case in English.18
6.5 Partial wh-movement
Some languages can form wh-constructions either with overt movement of a whphrase to its scope position or using partial wh-movement (McDaniel 1989) in which
a wh-phrase moves to the edge of an embedded clause, but has scope in a matrix
clause. A single language then can form a wh-construction with either wh-phrasal
movement or partial wh-movement. In this section, I attempt to explain why this
is the case.
I propose that languages that allow both full wh-movement and partial whmovement have these two options because a Typ head can contain either an EPPXP
0
feature or an EPPX feature. In a full wh-movement construction, an EPPXP feature
drives movement of a wh-phrase to a [Spec, TypP] position. If there is movement
through multiple clauses then there can be multiple EPPXP features. In a partial
wh-movement construction, first an EPPXP feature forces movement of a wh-phrase
0
to [Spec, TypP] of an embedded clause (or clauses), and then an EPPX feature
18
Note that I do not have data on island effects in Sharanahua. Also, island effects in Tlingit
can be circumvented for what I believe are reasons independent of wh-phrasal movement. I return
to this issue in chapter 7.
222
forces movement of the wh-feature to the matrix Typ (and sometimes also to an
0
embedded Typ). Interestingly, in these languages the EPPX feature cannot be
used freely. Rather, it can only occur if it c-commands an EPPXP feature contained
in TypP of the next lower embedded clause. I turn to an explanation of the details
of partial wh-movement in German and Albanian.
6.5.1 German
Examples (46a-b) below demonstrate that a wh-construction in German can be
formed via full wh-phrasal movement. The wh-phrase mit wem ‘with whom’ moves
overtly from the embedded clause to the initial position of the matrix clause.
(46) German:
(a) [P P Mit wem]1 glaubt Hans [T ypP t1 daSS Jakob jetz t1
with whom thinks Hans
that Jakob now
spricht]]?
talks
‘With whom does Hans think that Jakob is now talking?’
(b) [P P Mit wem]1 glaubst du
with whom believe you
t1 daSS Jakob t1 gesprochen
that Jakob
talked
‘With whom do you believe that
(Cheng 2000:78)
[T ypP t1 daSS Hans meint [T ypP
that Hans thinks
hat]]?
has
Hans thinks that Jakob talked?’
The full wh-movement facts can be accounted for straightforwardly if each relevant
Typ head contains an EPPXP feature that forces movement of the wh-phrase to the
embedded [Spec, TypP] and then to the matrix [Spec, TypP]. A simplified diagram
of the full wh-movement (46a-b) is shown below in (47). An EPPXP feature located
in both the embedded and matrix Typ heads forces overt phrasal movement of the
wh-phrase. When the wh-phrase arrives in the matrix TypP, its wh-feature values
the probe in Typ via a Spec-head relation.
223
(47)
TypP
PP1
Typ’
TP
...
mit wem[Fwh1 ] TypEP P XP
with whom Prb[Fwh1 ]
TypP
t1
Typ’
TypEP P XP
TP
. . . t1 . . .
In addition to full wh-phrasal movement, a wh-phrase can undergo partial whmovement. In (48), the wh-phrase mit wem ‘with whom’ raises to the specifier of
the embedded TypP, which Cheng (2000:78) notes “does not normally host a [+wh]
element since verbs such as glauben (‘to believe/think’) do not take an embedded
question.” The wh-phrase, however, has matrix scope which is marked by was in
the initial position of the matrix clause. In this case, was essentially functions as a
scope marker.
(48) German:
Was1 glaubt Hans [[Mit wem]1 daSS Jakob jetz t1 spricht]]?
WH
thinks Hans with whom that Jakob now
talks
‘With whom does Hans think that Jakob is now talking?’ (Cheng 2000:78)
The facts become more complex when a wh-phrase is multiply-embedded. In (49a),
the wh-phrase moves to [Spec, TypP] of the higher embedded clause. In (49b), the
wh-phrase moves to [Spec, TypP] of the lower embedded clause. In both cases,
224
matrix scope is marked by was. Example (49a) contains only a single was, whereas
(49b) contains was both in the higher embedded clause and in the matrix clause.
(49) German:
(a) Was1 glaubst du [T ypP [Mit wem]1 Hans meint [T ypP t1
WH
believe you
with whom Hans thinks
daSS Jakob t1 gesprochen hat]]?
that Jakob
talked
has
‘With whom do you believe that Hans thinks that Jakob talked?’
(b) Was1 glaubst du [T ypP was1 Hans meint [T ypP [Mit wem]1
WH
believe you
WH Hans thinks
with whom
Jakob t1 gesprochen hat]]?
Jakob
talked
has
‘With whom do you believe that Hans thinks that Jakob talked?’
(Cheng 2000:79)
According to Cheng (2000), if the wh-phrase remains in the lower clause in a
construction such as (49b), in some dialects of German, there must be the wh-scope
marker was in each higher clause. Example (50), in which the wh-phrase remains
in the lower clause and the higher embedded clause lacks a wh-scope marker, is
ill-formed.
(50) German:
*Was1 glaubst du [T ypP daSS Hans meint [P P Mit wem]1
WH
believe you
that Hans thinks
with whom
Jakob t1 gesprochen hat]]?
Jakob
talked
has
‘With whom do you believe that Hans thinks that Jakob talked?’ (Cheng
2000:79)
As can be seen above, a partial wh-movement construction in German utilizes
the wh-element was. Cheng (2000:80) argues that when used as a wh-scope marker,
225
was “is the wh-feature of the wh-phrase.”19 If was is a wh-feature then, since it
is a functional element, it should be an X0 category element that occurs in Typ.
There is evidence suggesting that this is the case. (51a) shows that a wh-phrase, in
this case mit wem ‘with whom,’ may co-occur with a complementizer, daB ‘that.’
However, in (51b), was cannot co-occur with daB ‘that.’
(51) German:
(a) Ich weiB nicht [[P P mit wem]1 (daB) du meinst t1 daSS
I
know not
with whom that
you think
that
Jakob t1 gesprochen hat].
Jakob
talked
has
‘I don’t know with whom you think that Jakob has talked.’
(b) Ich weiB nicht [was1 (*daB) du meinst [P P mit wem]1
I
know not WH
that
you think
with whom
(daB) Jakob t1 gesprochen hat].
that Jakob
talked has
‘I don’t know with whom you think that Jakob has talked.’ (Cheng
2000:96)
The ill-formedness of (51b) is accounted for straightforwardly if was is a head that is
in Typ. It cannot co-coccur with daB since daB is also in Typ.20 Further evidence
that was is a moved wh-feature can be seen by examining intervention and island
effects.
Example (52a) shows that movement of a wh-phrase is not blocked by the intervening negation nicht, whereas (52b) shows that partial wh-movement is blocked
by negation. Example (52b) is fine if negation is absent.
19
20
Was also has a use as a wh-phrase corresponding to what in English.
There is another possibility, which is that daB is actually the head of a FinP (see chapters
2.4-2.5) and was is in Typ. If this is the case, then the reason why daB and was cannot co-occur
is not due to their being in the same projection. I leave this issue for further analysis.
226
(52)
(a) [Mit wem]1 glaubst du
(nicht), dass Hans t1
with
whom believe you
not that Hans
gesporchen hat?
spoken has
‘With whom do(n’t) you believe that Hans has spoken?’
(b) Was1 glaubst du (*nicht), [Mit wem]1 Hans t1
WH
believe you not
with
whom Hans
gesporchen hat?
spoken has
‘What do(n’t) you believe with whom Hans has spoken?’ (Rizzi
1994:368)
The existence of intervention effects is expected if partial wh-movement involves
feature movement. In example (52b), the wh-feature is unable to move to the
matrix Typ when negation nicht intervenes between the wh-phrase and Typ.
Another piece of evidence supporting the idea that the wh-feature moves in partial wh-movement constructions is the extreme sensitivity of partial wh-movement
to islands. In the words of Cheng (2000:86), “partial wh-movement is more islandsensitive than overt extraction involving arguments.” Cheng (2000:87) gives the following examples in (53a-b), originally from Müller & Sternefeld (1996:21) and Müller
(1997:18-20), which show a wh-phrase that originates in a complex-DP. When there
is movement of the wh-feature out of the complex-DP in the partial wh-movement
example in (53a), the result is worse than when there is wh-phrasal movement out
of the complex-DP, as in (53b) .
(53) German (Complex-DP island):
(a) *Was hast du [DP ein Gerücht t2 ] gehört [typ2 wen1 Ede
WH
have you
a
rumor
heard
whom Ede
t1 mag]?
likes
227
(b) ??Wen1 hast du [DP ein Gerücht t2 ] gehört [typ2 t1 daSS
whom
have you
a
rumor
heard
that
Ede t1 mag]?
Ede
likes
Cheng (2000) argues that these facts are the result of feature movement being more
sensitive to island effects than phrasal movement. Why exactly feature movement
is more island-sensitive than phrasal movement is an issue that I return to in the
next section.
In this manner, partial wh-movement constructions appear to involve wh-phrasal
movement to an embedded TypP followed by wh-feature movement to a higher
TypP. I propose that the motivation for each type of movement is determined by
the EPP feature contained within Typ. For example, in (48), repeated below, the
embedded Typ contains an EPPXP feature and the matrix Typ contains an EPPX
feature.
(48) German:
Was1 glaubt Hans [[Mit wem]1 daSS Jakob jetz t1 spricht]]?
WH
thinks Hans with whom that Jakob now
talks
‘With whom does Hans think that Jakob is now talking?’ (Cheng 2000:78)
A simplified diagram is shown below in (54).
0
228
(54)
TypP
Typ’
TP
...
TypEP P X 0
Prb+was[Fwh ]
TypP
DP1
Typ’
mit wem+t[Fwh ] TypEP P XP
with whom
daB
that
TP
Jakob jetz t1 spricht
Jakob now talks
The embedded Typ contains an EPPXP feature that forces the wh-phrase to move
to the embedded [Spec, TypP]. The matrix clause, on the other hand, contains an
0
EPPX feature that forces the wh-feature of the wh-phrase to raise to the matrix
Typ head where it values a probe and is pronounced as was.21
In a partial wh-movement construction in some German dialects, there can be
variation in the position of a wh-phrase and in the number of occurrences of the whscope marker was. In (49a), repeated below, the wh-phrase mit wem ‘with whom’
moves to [Spec, TypP] of the higher embedded clause and its scope is marked by
was. In (49b), the wh-phrase only moves to the [Spec, TypP] of the lower embedded
clause, and there are two occurrences of was, one in the higher embedded clause and
one in the matrix clause.
21
0
Note that in order for a Typ head to have an EPPX feature, it must c-command a Typ head
in the next lower embedded clause that contains an EPPXP feature. I return to this issue in
section 6.5.3.
229
(49) German:
(a) Was1 glaubst du [T ypP
WH
believe you
daSS Jakob t1
that Jakob
‘With whom do you believe
[Mit wem]1 Hans meint [T ypP t1
with whom Hans thinks
gesprochen hat]]?
talked
has
that Hans thinks that Jakob talked?’
(b) Was1 glaubst du [T ypP was1 Hans meint [T ypP [Mit wem]1
WH
believe you
WH Hans thinks
with whom
Jakob t1 gesprochen hat]]?
Jakob
talked
has
‘With whom do you believe that Hans thinks that Jakob talked?’
(Cheng 2000:79)
Other than the positions of the wh-phrases and the number of occurrences of was,
these examples are identical. I propose that the difference between (49a) and (49b)
results from the types of EPP features selected by the relevant Typ heads. In
example (49a) both embedded Typ heads contain an EPPXP feature, and the matrix
0
clause Typ contains an EPPX feature, as shown in (55). The EPPXP features force
the wh-phrase to move overtly through the lower embedded [Spec, TypP] to the
0
higher embedded [Spec, TypP]. Then the EPPX in the matrix clause attracts the
wh-feature to the matrix Typ.
230
(55)
TypP
Typ’
TP
...
TypEP P X 0
Prb+was[Fwh ]
TypP
DP1
Typ’
mit wem+t[Fwh ] TypEP P XP
with whom
TP
...
TypP
t1
Typ’
TypEP P XP
daB
that
TP
. . . t1 . . .
In example (49b) on the other hand, both the matrix and the higher embedded
0
clauses contain an EPPX feature. Therefore, the wh-phrase moves to the lower
embedded [Spec, TypP] to statisfy the EPPXP feature. Then, the wh-feature alone
moves to the higher embedded Typ and again to the matrix Typ, in each case to
0
satisfy an EPPX feature. This wh-feature is pronounced as was. A diagram is
shown below.
231
(56)
TypP
Typ’
TypEP P X 0
Prb+was[Fwh1 ]
TP
...
TypP
Typ’
TP
...
TypEP P X 0
was[Fwh1 ]
TypP
DP1
Typ’
mit wem+t[Fwh1 ] TypEP P XP
with whom
TP
. . . t1 . . .
In this manner, the differences between these two types of constructions hinge on
the type of EPP feature that Typ contains.22
6.5.2 Albanian
Albanian is also a language with partial wh-movement, although unlike in German,
there does not appear to be an overt scope marker in a partial wh-movement con22
In these partial wh-movement constructions in which a wh-phrase is multiply embedded, the
0
facts regarding the positioning of the EPPX and EPPXP features are the same as in a construction
in which a wh-phrase originates in a singly embedded clause (see footnote 21). The TypP housing
the moved wh-phrase, has an EPPXP feature, and this Typ must be c-commanded by a Typ head
0
with an EPPX feature in the next higher clause. I return to this issue in section 6.5.3.
232
struction.
First of all, a wh-phrase may undergo full wh-movement, as in (57).
(57) Albanian:
Kë1
mendon se
Maria ka takuar t1 ?
Who-acc think-2S that mary has met
‘Who do you think that Mary met?’ (Turano 1998:162)
In a construction such as (57), an EPPXP feature in Typ attracts the wh-phrase to
[Spec, TypP].
Example (58) is a partial wh-movement construction. It corresponds to (57)
except that instead of moving to the front of the matrix clause, the wh-phrase kë
‘who’ only moves to the front of the embedded clause. Also, the Qu-morpheme a
appears in the initial position of the matrix clause. This morpheme is absent in
(57). The morpheme se is presumably in ForceP.
(58) Albanian:
A1 mendon [se kë1
ka takuar Maria t1 ]?
Qu think-2S that who-acc has met
Mary
‘Who do you think that Mary met?’ (Turano 1998:163)
Note that a partial wh-movement construction requires the presence of the particle a. Example (59) shows that a ‘Qu’ results in a yes/no construction, which
shows that it is a Qu-morpheme.
(59) Albanian:
A ke
punuar dje?
Qu have-2d worked yesterday
‘Did you work yesterday?’ (Turano 1998:164)
233
Although a ‘Qu’ occurs in yes/no and partial wh-movement constructions, it
does not occur in a full wh-movement construction, such as (57) above. Example
(60) below shows that a moved wh-phrase and a ‘Qu’ cannot co-occur.
(60) *Kë
a mendon se
ka takuar Maria?
who-acc Qu think-2s that has met
Maria
Intended: ‘Who do you think that Mary met?’ (Turano 1998:170)
This fact can be accounted for straightforwardly as resulting from the Doubly Filled
Comp Filter (Keyser 1975, Chomsky & Lasnik 1977), which is a constraint, found
in some languages such as English, that in its original formulation prevents there
from being a wh-word that is directly followed by a complementizer. Within a more
modern MP framework, it is a constraint that prevents there from being both a
specifier and a head in a CP (a TypP if my clause structure is utilized).23 Specifically,
if the Doubly Filled Comp Filter holds in Albanian, then both an overt wh-phrase
and an overt Qu-morpheme cannot co-occur in a single TypP projection. In a
partial wh-movement construction, there is no element in [Spec, TypP], and thus
the Qu-morpheme a is pronounced.
Turano (1998) writes that a construction such as (60) is ill-formed because,
following Koopman (1996), “for any given projection overt elements may appear
either in the head or in the Spec, but not in both positions (171).” This view
that both the head and specifier of a projection may not be filled is essentially
identical to the Doubly Filled Comp Filter. Turano writes that in this case, both
the Qu-morpheme and the wh-phrase check a Focus-feature in the head of a FocP
projection. Since a Focus-feature cannot be checked twice, the Qu-morpheme and
wh-phrase cannot co-occur. My analysis differs in that I assume that when a whphrase moves, there is still a Qu-morpheme present, but it is not pronounced. Both
23
I thank Heidi Harley (p.c.) for suggesting to me that the Doubly Filled Comp Filter may be
the key to understanding this phenomenon.
234
the Qu-morpheme and wh-phrase occur in TypP and there is no checking of a Focusfeature.
The fact that a ‘Qu’ occurs in a partial wh-movement construction, but not in
a full wh-movement construction leads Turano (1998:164) to argue that it “can be
analyzed as a [+WH] C0 element.” As I see the facts, a is not a wh-element. It is
merely a Qu-morpheme with a Qu-feature and when it co-occurs with a wh-feature,
then a wh-construction results. First of all, as shown in (59) above, a is used to
type a clause as a yes/no construction; in which case a clearly is not a [+WH]
element. Further evidence that a is not a wh-elment can be seen by examining
partial wh-movement constructions with multiply embedded wh-phrases. In (61a),
the wh-phrase kush ‘who’ is in [Spec, TypP] of the doubly-embedded clause, and
the Qu-morpheme a appears in the matrix clause. The ill-formed (61b) on the other
hand is identical to (61a) except that a also occurs in the higher embedded clause.
(61) Albanian:
(a) A1 mendon [T ypP se
Maria thotë [T ypP se
kush1 ka
Qu think-2S
that Mary says
that who
has
lexuar librin]]?
read
book-the
‘Who do you think that Mary says read the book?’
(b) *A1 mendon [T ypP a1 Maria thotë [T ypP se
kush1 ka
Qu think-2S
Qu Mary says
that who
has
lexuar librin]]?
book-the
ead
‘Who do you think that Mary says read the book?’
The ill-formedness of (61b) is accounted for straightforwardly if a is a Qu-morpheme.
It occurs in the matrix clause because this clause is an interrogative. It cannot occur
in the higher embedded clause because this clause is not an interrogative. If a were
a wh-element akin to the wh-scope marker was in German then it should be able to
235
occur in a non-interrogative embedded clause in a partial wh-movement construction
such as (61b), contrary to fact.
A partial wh-movement construction such as (58), repeated below, can then be
accounted for as follows.
(58) Albanian:
A1 mendon [se kë1
ka takuar Maria t1 ]?
Qu think-2S that who-acc has met
Mary
‘Who do you think that Mary met?’ (Turano 1998:163)
The embedded Typ head contains an EPPXP feature that attracts the wh-phrase
0
kë ‘who’ to its specifier position. The matrix Typ contains an EPPX feature that
attracts the wh-feature of the wh-phrase to the matrix Typ. Since the specifier of
Typ is empty, the Qu-morpheme a, which is a separate element from the wh-feature,
is pronounced without violating the Doubly Filled Comp Filter. A diagram is shown
below. Note that se ‘that’ is in the ForceP head in the embedded clause.
236
(62)
TypP
Typ’
TP
...
TypEP P X 0
Prb+Fwh1 ,a[FQu ]
ForceP
Force’
se
that
TypP
DP1
kë+t[Fwh1 ]
who-acc Typ
EP P XP
Typ’
TP
ka takuar Maria t1
has met Mary
Example (61a), repeated below, contains a multiply embedded wh-phrase that
has matrix scope.
(61) Albanian:
(a) A1 mendon [T ypP se
Maria thotë [T ypP se
kush1 ka
Qu think-2S
that Mary says
that who
has
lexuar librin]]?
read
book-the
‘Who do you think that Mary says read the book?’
In this case, there is an EPPXP feature in the lowest embedded Typ that forces the
wh-phrase kush to move to its specifier position. The higher embedded Typ and the
0
matrix Typ each contain an EPPX feature that forces movement of the wh-feature.
237
The Qu-morpheme a types the matrix clause as an interrogative and thus it appears
in the matrix Typ. A simplified diagram is shown below.
(63)
TypP
Typ’
TP
...
TypEP P X 0
Prb+Fwh ,aFQu
TypP
Typ’
TP
...
TypEP P X 0
t[Fwh ]
TypP
DP1
kush+t[Fwh ]
who
Typ’
TypEP P XP
TP
. . . t1 . . .
In summary, partial wh-movement constructions in Albanian, like in German,
are formed when there is a Typ head that contains an EPPXP feature and a Typ
0
head that contains an EPPX feature.24 Unlike in German, though, a moved whfeature is not pronounced. Also, unlike in German, Albanian contains an overt
Qu-morpheme that is pronounced in a partial wh-movement construction.
24
0
As in German, the Typ head with the EPPX feature must c-command a Typ head with an
EPPXP feature. I discuss this issue in the next section.
238
6.5.3 Remaining issues concerning partial wh-movement
Both German and Albanian are languages that require movement of a wh-phrase. I
propose that there is the following constraint in these languages.
(64) The closest c-commanding Typ head of a clause with a wh-phrase must
contain an EPPXP feature.
This constraint requires movement of a wh-phrase to the specifier of the closest ccommanding TypP. If this constraint is satisfied, then there can also be an EPPX
0
0
feature. This constraint explains why an EPPX feature cannot occur freely in any
Typ head.
0
First of all, (64) accounts for the fact that a Typ head with an EPPX feature
cannot precede a Typ head with an EPPXP feature. The structure in (65), in which
0
an embedded Typ contains an EPPX feature that attracts a wh-feature and a higher
Typ contains an EPPXP that attracts a wh-phrase is not allowed.
(65)
TypP
Typ’
TypEP P XP
TP
...
TypP
Typ’
TypEP P X 0
TP
. . . wh-phrase . . .
The EPPXP feature in the higher Typ attracts a wh-phrase from the embedded
clause. In order to arrive in the specifier of the matrix TypP, the wh-phrase would
239
have to move through the embedded [Spec, TypP], but this is not possible because
0
the embedded Typ contains an EPPX feature that only attracts the wh-feature,
and not the wh-phrase. According to Phase Theory, only an element at the edge
of a phase is accessible to higher phrases. Thus, if the embedded wh-phrase does
not move to the embedded [Spec, TypP], it is not able to move out of the lower
phase (the embedded TypP) and into a higher phase. Therefore, there is a ban on
a construction with the configuration in (65).
The constraint in (64) also straightforwardly accounts for the fact that a whphrase cannot remain in-situ in these languages. For example, (66a-b) from German
are ill-formed because the wh-phrase does not move.
(66) German:
(a) *Was1 ist er wem1 begegnet?
WH
is he who
met
(Sabel 2000:419)
(b) *Was1 glaubst du [CP was1 Hans meint [CP Jakob [mit
WH
believe you
WH Hans thinks
Jakob with
wem]1 gesprochen hat]]?
has
whom talked
‘With whom do you believe that Hans thinks that Jakob has talked?’
(Cheng 2000:80)
Similarly, the Albanian (67) is ill-formed because the wh-phrase remains in-situ.25
(67) Albanian:
*A pe
kush rrugës?
Qu saw-2s who street
‘Who did you see in the street?’ (Turano 1998:150)
25
Note that (67) is well-formed as a yes/no construction meaning Did you see anyone in the
street?’, in which case the wh-element functions as an indefinite (Turano 1998:150).
240
These are ill-formed as wh-constructions because Typ lacks an EPPXP feature,
thereby violating (64).
Cheng (2000) has a different analysis of why wh-phrasal movement must precede
wh-feature movement; however, it has some problems. Cheng (2000:81) writes the
following.
Following Chomsky (1995; class lecture fall 1995), I assume that overt
wh-movement involves a two-step movement: feature movement and category movement. Feature movement falls under Attract F (i.e., as a
result of feature attraction/checking) while category movement (the socalled generalized pied piping) is for PF convergence. Further, Chomsky
proposes that after a category undergoes movement, an automatic repair
strategy takes place to ensure that the feature(s) will not be scattered
(Cheng 2000:81).
Cheng, following Chomsky, claims that a wh-feature raises to C and then the whphrase raises to the specifier of CP so that a repair strategy can occur. The repair
strategy is described as “a mechanism which puts the feature bundle back into the
category (Cheng 2000:81).” The wh-feature separates from the wh-phrase, undergoes
movement to C, and then the wh-phrase moves to [Spec, CP] so that the feature
can move back into the wh-phrase. According to Cheng, a wh-phrase must move
before partial wh-movement can occur because the repair strategy must occur. If
the wh-phrase remains in-situ and its feature raises to C, then there is no point in
the derivation in which the repair strategy can occur. Cheng (2000:82) writes that
“category movement is a necessary step preceding the repair strategy.”
A problem with Cheng’s analysis is that, in my opinion, it predicts that whphrasal movement should behave like wh-feature movement with respect to island
effects. As shown in (53a-b), repeated below, partial wh-movement out of an island
is worse than full wh-movement out of an island.
241
(53) German (Complex-DP island):
(a) *Was hast du [DP ein Gerücht t2 ] gehört [typ2 wen1 Ede
WH
have you
a
rumor
heard
whom Ede
t1 mag]?
likes
(b) ??Wen1 hast du [DP ein Gerücht t2 ] gehört [typ2 t1 daSS
whom
have you
a
rumor
heard
that
Ede t2 mag]?
Ede
likes
(Cheng 2000:87)
In constructions such as (53a), partial wh-movement involves wh-feature movement
alone out of an island. When there is wh-phrasal movement as in (53b), according
to Cheng, the wh-feature moves to C and the wh-phrase moves to [Spec, CP] for the
repair strategy to occur. Since, under Cheng’s analysis, the wh-feature undergoes
head-movement regardless of whether or not there is partial or full wh-movement,
full wh-phrasal movement out of an island should be just as bad as partial whmovement. It maybe should even be worse, since two elements (a wh-phrase and
a wh-feature) move in the case of full wh-movement, but only a wh-feature moves
in the case of partial wh-movement. Yet wh-phrasal movement is not just as bad
as partial wh-movement. Partial wh-movement is more island sensitive than full
wh-movement. Cheng accounts for this fact by suggesting that category movement
somehow lessens an island violation. Cheng (2000:87) writes the following.
In cases where full movement takes place, both feature movement and
category movement are involved. However, in such cases, the violations
incurred by feature movement are “cancelled” by category movement.
Thus, according to Cheng, category movement which follows feature movement
lessens an island effect incurred by feature movement.26 However, it is not clear
26
Cheng states that the island effects are ‘canceled’ but this is not quite correct, since island
effects still occur, but are not as severe.
242
why category movement would lessen the severity of an island effect. Since there
is movement of two elements out of an island, an island effect should worsen. Instead, as I see it, the difference in the severity of island effects resulting from full
wh-movement versus partial wh-movement results from the difference between movement of a feature, which is an X0 category element, and XP movement of a phrase.
This view of the facts raises the question of why head-movement is more sensitive
to island effects than phrasal movement is. However, I do not have a solution at
this time. The island effects in the complex-DP constructions in (53a-b) can be
attributed to renumeration of the adjunct clause modifier of the nominal. Why
exactly movement of a wh-feature is worse than movement of a wh-phrase out of a
complex-DP is not clear to me, and is an issue that requires further examination.
In conclusion, partial wh-movement languages such as German and Albanian
require a wh-phrase to undergo movement before its wh-feature can separate from
0
it and raise to adjoin to a wh-expletive. There can only be an EPPX feature in a
Typ head if there is already an EPPXP feature in a lower embedded Typ. A partial
wh-movement construction must have the configuration shown in (68a), although
there can be any number of Typ heads with an EPPXP feature as long as they are
0
c-commanded by a Typ with an EPPX feature, and there can be any number of
0
Typ heads with an EPPX feature as long as they c-command a Typ head with an
EPPXP feature. The configuration in (68b) which results in wh-feature movement
preceding wh-phrasal movement, and the configuration in (68c) which results in a
wh-in-situ construction are not allowed.
(68)
(a) [T ypP [T yp
EP P X
0
. . . [T ypP [T ypEP P XP . . . ]]]]
(b) *[T ypP [T ypEP P XP . . . [T ypP [T yp
(c) *[T ypP [T yp
EP P X
0
EP P X
0
. . . ]]]]
. . . ]]
The facts of (68) are forced by (64) which requires a wh-construction to have an
EPPXP feature in the closest Typ to c-command a relevant wh-phrase and by these
243
0
languages containing an EPPX feature in their lexicons.
6.6 Wh-phrasal movement, Agree, and partial wh-movement in a single language:
Malay
Malay27 is interesting in that it allows a wh-construction to be formed in a variety
of ways. Malay, like German and Albanian, allows a wh-phrase either to move to its
scope position or to undergo partial-wh-movement, and in addition, it allows a whphrase to remain in-situ. I propose that these facts are the result of Malay allowing
three different types of Typ heads, one without an EPP feature, one with an EPPXP
0
feature, and one with an EPPX feature, although as in the partial wh-movement
0
languages discussed above, the EPPX feature can only occur if there is already an
EPPXP feature in a lower Typ head.
I begin by presenting the relevant data from Malay. First, a wh-phrase can
undergo movement to the position where it takes scope, as in (69).
(69) Malay:
Siapa1 (yang) [Bill harap [yang t1 akan
membeli baju
Who
(that) Bill hope that
will buy clothes for
untuknya]]?
him
‘Who does Bill hope will buy clothes for him?’ (Cole & Hermon 1998:224)
The second option is for a wh-phrase to remain in-situ, as shown in (70).28
27
This is a language spoken in Brunei, Indonesia, Malaysia, and Singapore. The data that I
present are from Cole & Hermon (1998) who state that their data come from “the Malay spoken
by educated speakers in Singapore.” See also Saddy (1990) for a discussion of Indonesian, which
is closely related to Malay, and which shows the same variety of wh-constructions.
28
Note that unlike a wh-phrase such as apa ‘what,’ kenapa ‘why’ cannot remain in-situ. See
chapter 9 for further discussion.
244
(70) Malay:
Ali memberitahu kamu tadi
[Fatimah baca apa]?
Ali informed
you
just now Fatimah read what
‘What did Ali tell you Fatimah was reading?’ (Cole & Hermon 1998:224)
The third option is for a wh-phrase to undergo partial movement, as shown in (71).
(71) Malay:
Jon fikir [kenapa1 (yang) Mary rasa [Ali dipecat
t1 ]]?
John think why
(that) Mary feel Ali was fired
‘Why does John think (that) Mary felt Ali was fired?’ (Cole & Hermon
1998:225)
The wh-phrase moves to the initial position of an embedded clause but has scope in
a higher clause. Note that there is no overt wh-expletive.
When a wh-phrase moves to its scope position as in (69), according to my analysis, there is an EPPXP feature in Typ that forces overt movement of the wh-phrase,
and the wh-feature values the probe in Typ via a Spec-head relation. A diagram is
shown in (72).
(72)
TypP
DP1
wh-phrase[Fwh ] TypEP P XP
Prb[Fwh ]
Typ’
TP
. . . t1
Evidence supporting the idea that there is overt wh-movement is the existence of
island effects. For example, in (73), the wh-phrase di mana ‘at where’ moves from
inside a complex-DP, resulting in ill-formedness. See Cole & Hermon (1998:227-228)
for further examples.
245
(73) Malay:
*[Di mana]1 kamu fikir Ali suka [DP perempuan yang tinggal
At
where
you
think Ali like
woman
who live
t1 ]?
Intended: ‘Where1 do you think Ali likes the woman who lives t1 ?’ (Cole &
Hermon 1998:227)
The ill-formedness of this construction is accounted for straightforwardly if the
EPPXP feature in Typ is unable to attract the wh-phrase out of a renumerated
adjunct clause within the complex-DP.
When a wh-phrase remains in-situ as in (70), there are two possibilities; one
is that the wh-feature alone moves, and the other is that the wh-feature forms an
Agree relation with Typ. Evidence suggests that the latter is the case; there is an
Agree relation. This is because an in-situ wh-phrase is not subject to island effects.
Example (74) shows that the wh-phrase di mana ‘at where’ can remain in-situ and
result in a perfectly well-formed wh-construction, which is identical to the ill-formed
(73), except that the wh-phrase remains in-situ.
(74) Malay:
Kamu fikir [Ali suka [DP perempuan yang tinggal di mana]]?
you
think Ali like
woman
that live
at where
‘Where1 do you think Ali likes the woman who lives t1 ?’ (Cole & Hermon
1998:228)
The well-formedness of this construction can be accounted for if there is no EPP
feature in Typ. The wh-feature undergoes an Agree relation with a probe in Typ,
and Agree can occur between the matrix Typ and a wh-feature contained within a
renumerated wh-phrase. Since there is no movement of the wh-feature, there is no
island effect. A diagram is shown in (75).
246
(75)
TypP
Typ’
Typ
Prb[Fwh1 ]
TP
. . . wh-phrase[Fwh1 ] . . .
In a partial wh-movement construction, there are island effects, just as in a
construction in which the wh-phrase moves to its scope position. Example (76a)
shows that when the wh-phrase remains in-situ within the complex-DP the result is
well-formed. When the wh-phrase undergoes partial movement in (76b), the result
is ill-formed.
(76) Malay:
(a) Kamu sayang [DP perempuan yang Ali fikir
[yang telah
you
love
woman
that Ali thinks that already
berjumpa siapa]]?
meet
who
‘Who1 do you love [the woman who Ali thinks met t1 ]?’
[(dengan)
(b) *Kamu sayang [DP perempuan yang Ali fikir
you
love
woman
that Ali thinks (with)
siapa1 yang telah
jumpa t1 ]]?
who
that already meet
Intended: ‘Who1 do you love [the woman who Ali thinks met t1 ]?’ (Cole
& Hermon 1998:235)
The ill-formedness of (76b) results from wh-feature movement. The wh-phrase moves
to an embedded [Spec, TypP]. Then movement of the wh-feature is blocked by the
complex-DP.
A partial wh-movement construction can then be accounted for as shown in (77).
There is an EPPXP feature in the embedded Typ that forces wh-phrasal movement
247
0
to the embedded [Spec, TypP]. The matrix Typ contains an EPPX feature that
attracts the wh-feature from the wh-phrase within the embedded clause.
(77)
TypP
Typ’
TP
...
TypEP P X 0
Prb+ Fwh
TypP
DP1
wh-phrase+t[Fwh ] TypEP P XP
yang
that
Typ’
TP
. . . t1 . . .
Note that this partial wh-movement works just as in German and Albanian; an
0
EPPX feature can only occur if there is an an EPPXP feature in a lower embedded
clause. Remember that I claimed that German and Albanian have the constraint in
(64), repeated below.
(64) The closest c-commanding Typ head of a clause with a wh-phrase must
contain an EPPXP feature.
This same constraint appears to hold in Malay, although with a twist. It is not
the case that a wh-construction always must have an EPPXP feature in Typ, as
indicated by the possibility of Agree in this language. Rather, when there is whphrasal movement of any sort, there must be EPPXP feature in Typ. Thus I revise
(64) as follows.
(78) In a construction with wh-phrasal movement, the lowest c-commanding Typ
head of a clause with a wh-phrase must contain an EPPXP feature.
248
The Malay wh-construction facts then raise the question of why so many different
options for forming a wh-construction are allowed in a single language. The answer
lies in the possible underlying numerations of a wh-construction. There are three
basic numerations that are allowed, as shown in (79a-c), which differ in terms of
whether or not there is an EPP feature in Typ and the type of EPP feature that
occurs in Typ.
(79)
Wh-construction type
EPP
(a)
Wh-movement
EPPXP
(b)
Wh-in-situ
∅
(c)
Partial wh-movement
[T ypP [T yp
EP P X
0
. . . [T ypP [T ypEP P XP . . . ]]]]
In a wh-movement construction (79a), Typ contains an EPPXP feature that
forces the wh-phrase to move to [Spec, TypP]. In a wh-in-situ construction (79b),
Typ lacks an EPP feature, and thus Agree alone occurs. In a partial wh-movement
construction, there is an EPPXP feature in an embedded Typ that forces movement
0
of the wh-phrase to [Spec, TypP] and there is an EPPX feature in a higher
Typ head that forces wh-feature movement. These options exist in various other
languages. Thus, Malay is a language which can form a wh-construction in the same
way as in English with wh-phrasal movement, as in Mandarin with no movement
of a wh-element, or as in German and Albanian, with partial wh-movement.
Before I end this section, I briefly compare my analysis with that of Cole &
Hermon (1998), who take a different approach to the Malay wh-construction facts.
They assume that a wh-phrase consists of an operator and a variable, and that
whether or not the variable and operator form a single word depends on the language.
In English, they propose that an operator and variable form a single lexical item,
thus accounting for the requirement for full wh-movement in single wh-constructions,
and in Chinese, an operator and variable are separate, thus accounting for why a
249
wh-phrase can remain in-situ and why it is not subject to island effects. In Malay,
on the other hand, they propose that the operator and variable can either be a
single lexical item, or they can be separate, and that these facts account for the
wide variety of Malay wh-constructions.
Cole & Hermon (1998) take the position that in a wh-construction, a C head contains what they refer to as a “STRONG Q (wh-feature) that needs to be checked by
a wh-feature that moves into the checking domain of C (Cole & Hermon 1998:223).”
In other words, there is a Q feature in C that needs to be checked by a wh-feature.
Cole & Hermon (1998) argue that there are two possibilities for the structure
of a wh-phrase. One type of wh-phrase consists of an operator and a variable that
form a single lexical item, represented as “[OP+VAR].” This single wh-phrase is
used in both full and partial wh-movement constructions. The other type consists
of a separated operator and variable, represented as “[OP.. ..VAR],” and this results
in a wh-in-situ construction.
Evidence that a wh-word and an operator can be separate is that a wh-word in
Malay can behave as a variable. For example, apa ‘what’ can be used in the following
example, where Cole and Hermon argue that it is “bound by the existential quantifier
represented by -pun ‘also’ (240).”
(80) Malay:
Dia tidak membeli
apa-pun untuk saya.
he not MENG-buy what-also for
me
‘He did not buy anything for me.’ (Cole & Hermon 1998:239)
Based on this use of a wh-word as a variable, Cole & Hermon argue that a wh-word
can occur separately from an operator.
In a full wh-movement construction, there is a single lexical item [OP+VAR]
that undergoes movement to [Spec, CP] to check the STRONG Q feature in C by
discharging its wh-feature. The need to check the STRONG Q feature in C results
250
in overt wh-phrasal movement. Movement of a wh-phrase is subject to island effects,
as in (73) above.
In a partial wh-movement construction, again Cole and Hermon argue that there
is an [OP+VAR] wh-phrase, but in addition, unlike in a full wh-movement construction, there is a null wh-expletive that is Merged directly in the scopal [Spec, CP]
position, where it “satisfies the STRONG feature of Q (250).” Because the whexpletive is in the scopal [Spec, CP] position, the [OP+VAR] wh-phrase moves to a
non-scopal [Spec, CP] position. Then, at LF, the operator OP moves to the scopal
[Spec, CP] where it replaces the wh-expletive. From this position, the OP binds the
variable VAR. LF movement of this wh-operator is also subject to island effects, as
in (76b).
In a wh-in-situ construction, Cole and Hermon argue that there is an [OP..
..VAR] structure. The operator part of the wh-phrase is base generated directly
in [Spec, CP], where it checks the STRONG Q feature, and binds the wh-phrase.
Following Tsai (1994), Reinhart (1993, 1995), and Cole & Hermon (1995), they
attribute the lack of island effects in a wh-in-situ construction to unselective binding
of the wh-phrase by the operator. Unselective binding is a proposed form of binding
in which an operator binds a variable and this binding relation is not subject to
intervention effects.
There are several problems that I see with this analysis.
First of all, the motivation for wh-phrasal movement (in constructions with overt
wh-phrasal movement), and also wh-operator movement (in constructions with partial wh-movement) is the need to check a wh-feature on a STRONG Q feature in C.
This analysis thus lumps together Qu- and wh-features, which I believe are distinct
elements. The label Q indicates a feature that results in an interrogative; i.e., in a
yes/no or a wh-construction. However the STRONG Q of Cole and Hermon’s analysis requires checking by a wh-feature, and thus appears to be an element specific to a
wh-construction. This type of analysis appears to require that a yes/no construction
251
contain a different type of Q-feature than a wh-construction. My analysis unifies
the interrogative aspect of yes/no and wh-constructions by simply assuming that
the Qu-feature of a wh-construction is the same type of feature that appears in a
yes/no construction. The Qu-feature has nothing to do with motivating movement
of a wh-element. Rather, movement is motivated by an EPP feature.
Another problem with this analysis is that it relies on the notion that a wh-phrase
consists of an operator and a variable that can be separated, either at Spell-Out
(in a wh-in-situ construction), or at LF (in a partial wh-movement construction).
Specifically, because an operator is base generated in [Spec, CP], it can bind a
variable wh-word in a wh-in-situ construction, or it can bind a variable wh-word
in a non-scopal [Spec, CP] position. But Bruening (2007) demonstrates that there
are problems with this view. Bruening (2007) points out that there are wh-in-situ
languages such as Turkish (a fact noted by Cole and Hermon) that do not use
wh-words as variables, and there are languages that use wh-words as indefinites,
but that require overt wh-phrasal movement, such as Pasamaquoddy. Therefore,
whether or not a wh-word can be used as a variable in a language does not predict
whether or not a wh-word can remain in-situ. One implication of Bruening’s analysis
is that it may not be the case that a wh-phrase actually consists of an operator and
a variable. This notion is compatible with my analysis, in which I make no recourse
to the notion of null wh-operators.
Next, in a partial wh-movement construction, Cole and Hermon claim that a
null wh-expletive is replaced at LF by movement of a null wh-operator. The null
wh-expletive checks the STRONG Q feature in the scopal C position, and the null
wh-operator undergoes movement to replace the expletive, most likely in order to
give a wh-phrase scope. This analysis does not account for why a full [OP+VAR]
wh-phrase moves to a non-scopal [Spec, CP] position. If there is an expletive inserted
in the scopal C head, and the null wh-operator can move at LF, then the wh-phrase
should be able to remain in-situ at Spell-Out as its wh-operator could move at LF.
252
Under my account, a wh-phrase moves to a non-scopal [Spec, CP] position because
of a requirement that there be an EPPXP feature to drive wh-movement (see (64)),
0
in which case, there can be an EPPX feature in a higher Typ head. There is no
reliance on a null wh-expletive or movement of a null wh-operator.
Lastly, in a wh-in-situ construction, Cole and Hermon rely on the notion that
a wh-operator is base generated in the scopal [Spec, CP] position and unselectively
binds an associated in-situ wh-word. Under my analysis, there is an Agree relation
between a probe in Typ and the wh-feature of the wh-phrase. This Agree relation is
similar to the notion of unselective binding, although it is not an operator variable
relationship. Rather it is a probe and goal (wh-feature) relationship, and the whin-situ facts do not result from binding by a null wh-operator.
In conclusion, the basic differences between my analysis and that of Cole and
Hermon, are that, under my analysis, there is no wh-operator movement, nor is
there any LF movement. The wh-construction facts simply rely on the relationship
that a wh-feature forms with a probe and whether or not a Typ head contains an
EPP feature.29
6.7 French: wh-phrasal movement and wh-feature movement in a single language
In this section, I discuss French, a language in which a wh-phrase may either occur
in-situ or undergo movement, although the option for a wh-phrase to remain in-situ
is limited.
According to Bos̆ković, wh-in-situ in French is limited to one particular situation,
a matrix clause with a null complementizer, as in (81a), in which the wh-phrase qui
‘who’ remains in-situ. The wh-phrase may also move to clause-initial position as in
(81b).
29
Cole and Hermon also apply their analysis to other languages, with English being a language
in which a wh-phrase and operator are a single lexical item and Chinese a language in which a
wh-phrase always consists of a wh-word variable and a separate operator.
253
(81) French:
(a) Tu as
vu
qui?
you have seen whom
‘Who did you see?’
(b) Qui1 as-tu
vu
t1 ?
whom have-you seen
‘Who did you see?’ (Bos̆ković 1998)
In all other environments, a wh-phrase must move. Examples (82a-b) show a
matrix clause with an overt complementizer. The wh-phrase cannot remain in-situ,
as shown by the ill-formedness of (82a). (82b) in which the wh-phrase moves to
clause-initial position is fine.
(82) French:
(a) *Que tu as
vu
qui?
comp you have seen who
Intended: ‘Who have you seen?’
(b) Qui1 que
tu as
vu
t1 ?
who comp you have seen
‘Who have you seen?’ (Bos̆ković 2002:352)
Example (83a) is ill-formed because the wh-phrase remains in-situ in the embedded
clause. (83b) is fine because the wh-phrase has moved to the initial position of the
embedded clause.
(83) French:
(a) *Pierre a
demandé [tu as
embrassé qui].
Pierre has asked
you have kissed
who
Intended: ‘Pierre has asked who you have kissed.’
254
(b) Pierre a
demandé [qui1 tu as
embrassé t1 ]
Pierre has asked
who you have kissed
‘Pierre has asked who you have kissed.’ (Bos̆ković 2002:352)
Examples (84a-b) show that when an embedded wh-phrase has matrix scope it must
move. (84a) is ill-formed because the wh-phrase remains in-situ, whereas (84b) is
well-formed because the wh-phrase has moved to the beginning of the matrix clause.
(84) French:
(a) *Jean et
Marie croient [que Pierre a
embrassé qui]?
Jean and Marie believe that Pierre has kissed
who
Intended: ‘Who do Jean and Marie believe that Pierre has kissed?’
(b) Qui1 Jean et
Marie croient-ils [que Pierre a
embrassé t1 ]?
who Jean and Marie believe
that Pierre has kissed
‘Who do Jean and Marie believe that Pierre has kissed?’ (Bos̆ković
2002:352)
The constructions in which a wh-phrase moves overtly are accounted for straightforwardly according to my analysis. The TypP head contains an EPPXP feature
that attracts the wh-phrase to [Spec, TypP].
Accounting for what happens when a wh-phrase remains in-situ is more problematic. Bos̆ković (2000) argues that a complementizer with a strong feature can be
inserted into a derivation at LF, in which case the strong feature of the complementizer forces the wh-phrase to move at LF. Since movement is at LF, the wh-phrase
remains in-situ at Spell-Out. This LF insertion can only occur in a matrix clause
since Merge must expand a tree. However, a general assumption in current work in
the Minimalist Program (Chomsky 1995b) is that LF movement does not exist. If
this is the case, then another analysis is required.
When the wh-phrase remains in-situ, there are two possibilities. One is that the
0
wh-feature alone undergoes feature movement driven by an EPPX feature in Typ.
255
The other possibility is that there is an Agree relation between the probe in Typ
and the wh-feature, and there is no movement of the wh-feature. The existence
of intervention effects indicates that wh-feature movement is at work. When an
in-situ wh-phrase (in a matrix clause with a null complementizer) is c-commanded
by “quantifiers, negation, or modals (Cheng & Rooryck 2000:11),” the result is illformedness. In (85a), the wh-phrase is c-commanded by a quantifier, in (85b) by
negation, and in (85c) by a modal.
(85)
(a) *Tous les éstudiants ont rencontré qui?
all
the students have met
who
Intended: ‘Who did all the students meet?’ (Chang 1997:17, per Cheng
& Rooryck 2000:11)
(b) *Il n’
a
pas rencontré qui?
he neg has neg met
who
Intended: ‘Who didn’t he meet? (Chang 1997:19, per Cheng & Rooryck
2000:11)
(c) *Il peut rencontrer qui?
he can meet-sc inf who
Intended: ‘Who can he meet?’ (Cheng & Rooryck 2000:11)
The intervention effects in the ill-formed (85a-c) are accounted for straightforwardly
if the wh-feature moves, but this movement is blocked by a c-commanding scope
0
bearing element. There is an EPPX feature in Typ that is unable to attract the
wh-feature because of the intervener, thereby resulting in ill-formedness, as shown
below.
256
(86)
TypP
Typ’
TypEP P X 0
QuantP/NegP/ModP
Quant/Neg/Mod
*
TP
. . . wh-phrase[Fwh ] . . .
Note that intervention effects disappear when there is overt wh-movement. Example (87a) is ill-formed because negation c-commands an in-situ wh-phrase, but
when the wh-phrase precedes negation, as in (87b), the result is well-formed.
(87) French:
(a) ?*Jean ne mange pas quoi?
Jean
neg eats
neg what
Intended: ‘What doesn’t John eat?
(b) Que ne mange-t-il pas?
neg
what neg eats
‘What doesn’t he eat? (Bos̆ković 1998)
When the wh-phrase moves to clause initial position in (87b), it undergoes XP
movement, which is not subject to intervention effects.
These facts then indicate that when a wh-phrase remains in-situ, the wh-feature
0
separates from the associated wh-phrase and moves to satisfy an EPPX feature in
Typ. A simplified diagram of this type of construction is shown below.
257
(88)
TypP
Typ’
TP
TypEP P X 0
Prb+Fwh1
. . . wh-phrase+tFwh1 . . .
0
If this analysis is on the right track, then in French, an EPPX feature and
an EPPXP feature have a different distribution. An EPPXP feature can occur in
0
any Typ head, whether it be a matrix or embedded Typ. An EPPX feature can
never occur in an embedded clause, which is why an embedded wh-phrase can never
remain in-situ and have matrix scope, as shown in (83a-b) and (84a-b) above. An
0
EPPX feature is confined to only a matrix Typ head. In a simple matrix clause,
0
Typ can contain an EPPX feature. For example, (81a), repeated below as (89)
0
contains an EPPX feature in the matrix Typ.
(89) French:
[T yp
EP P X
0
Tu as
vu
qui]?
you have seen whom
‘Who did you see?’ (Bos̆ković 1998)
0
However, the EPPX feature is incompatible with an overt complementizer. For
example, (82a-b) above have the following structures in (90a-b). (90a) is ill-formed
0
because there is an overt complementzer and an EPPX together in Typ. (90a-b),
which has wh-phrasal movement, is fine because an overt complementzer and an
EPPXP can co-occur.
(90) French:
Que tu as
vu
qui]?
comp you have seen who
Intended: ‘Who have you seen?’
(a) *[T yp
EP P X
0
258
(b) [T ypEP P XP Qui1 que
tu as
vu
t1 ?
who comp you have seen
‘Who have you seen?’
French thus bans the following configuration in which que co-occurs with an
0
EPPX feature.
0
(91) *[T yp que, EPPX . . . ]
When que co-occurs with an EPPXP feature in Typ, 91 is violated. Why exactly
this restriction holds, though, is not clear to me.
Lastly, there is no partial wh-movement in French. For example, a wh-phrase
cannot move to the edge of an embedded clause, and have matrix scope. In (92),
qui ‘who’ has moved to the edge of the embedded clause, and it cannot have matrix scope. There cannot be a wh-scope marker; this is ill-formed if que ‘what’ is
present in the matrix clause to mark scope, and it is ill-formed, if que ‘what’ is absent. It is also ill-formed regardless of whether or not there is an embedded clause
complementizer que.
(92) *(Que) Jean at Marie croient [(que) qui1 Pierre a
embrasse
(what) Jean and Marie believe (that) who Pierrre has kissed
t1 ]?
‘Who do Jean and Marie believe that Pierre kissed?’ (Dalila Ayoun, p.c.)
The French facts thus raise the issue of why there is no partial wh-movement.
0
An EPPX feature should be able to attract the wh-feature of a wh-phrase that
has moved to the specifier of an embedded TypP, but this is not allowed. A whphrase cannot move to [Spec, TypP] of an embedded clause and have matrix scope.
The following configuration in which the embedded clause Typ contains an EPPXP
0
feature and the matrix clause Typ contains an EPPX feature is banned.
(93) *[T ypP [T yp
EP P X
0
. . . [T ypP [T ypEP P XP . . . ]]]].
259
At this point, I do not have an explanation for why this is the case. One possibility
0
though is that the EPPX feature in French is ‘weak’ in the sense that it can only
attract a wh-feature that is in the same clause.
0
Lastly, another perplexing fact is that the EPPX feature is confined to a matrix
clause. It cannot occur inside of an embedded clause. Why this would be the case
is not clear.
To summarize, in French, an EPPXP feature can occur in any Typ head, whereas
0
an an EPPX feature can only occur in a matrix Typ and it is incompatible with
0
an overt complementizer. Why exactly an EPPX has such a limited distribution in
French though requires further investigation.
6.8 Conclusion
In this chapter I have examined wh-constructions that are formed via Agree, whphrasal movement, wh-feature movement, and/or partial wh-movement. When Typ
lacks an EPP feature that attracts a wh-element (either there is no EPP feature, or
there is an EPP feature, but it does not attracts a wh-element), there is an Agree
relation between a wh-feature and a probe in Typ, as shown in (94), where ‘Relation’
refers to the type of relationship established between a wh-feature and Typ.
(94)
Languages
EPP
Relation
Sinhala, Okinawan, Mandarin
∅
Agree
0
When there is an EPPX feature in Typ, then there is wh-feature movement,
as shown in (95).
(95)
Languages
EPP
Japanese, Korean, Persian
EPPX
Relation
0
Wh-feature movement
260
When there is an EPPXP feature in Typ, there is full wh-phrasal movement
to [Spec, TypP], as shown in (96).
(96)
Languages
EPP
Relation
English, Sharanahua, Tlingit
EPPXP
Wh-phrasal movement
Languages that permit either wh-phrasal movement or partial wh-movement
0
allow both an EPPXP feature and an EPPX feature in Typ. (97a) with only an
EPPXP feature results in full wh-movement, and (97b) with both an an EPPXP
0
feature and an EPPX feature results in partial wh-movement. However, an EPPX
0
feature can only occur if an embedded clause has an EPPXP feature.
(97)
0
Languages
EPPXP
EPPX
(a)
German, Albanian
X
∅
Wh-phrasal movement
(b)
German, Albanian
X
X
Partial wh-movement
Relation
Then there are languages such as Malay, which allows all strategies except
for wh-feature movement, as shown in (98a-c). (98a) states that there is an Agree
relation, in which case a wh-phrase remains in-situ, when the Typ head lacks an
EPP feature. (98b) states that there is wh-phrasal movement when all relevant Typ
heads contain an EPPXP feature. (98c) states that there is partial wh-movement
0
when there is both an EPPXP feature and an EPPX feature. Again, the EPPX
feature can only occur if an embedded clause contains an EPPXP feature.
(98)
0
EPPXP
EPPX
(a)
∅
∅
Agree
(b)
X
∅
Wh-phrasal movement
(c)
X
X
Partial wh-movement
Relation
0
261
Lastly, there is French.
(99a) states that wh-phrasal movement occurs when
a Typ head contains an EPPXP feature. (99b) states that wh-feature movement
0
0
occurs when a Typ head contains an EPPX feature. However, an EPPX feature
can only occur in a matrix clause with a null complementizer.
(99)
0
Clause
EPPXP
EPPX
(a)
Any
X
∅
Wh-phrasal movement
(b)
Matrix, null complementizer
∅
X
Wh-feature movement
Relation
This analysis also suggests that there is a crucial distinction between movement
triggered by an EPP feature and Agree. An EPP feature attracts elements of a
particular type, thus resulting in intervention effects when there is an intervening
element of the appropriate type but the incorrect content. EPP-driven movement is
also subject to island effects when movement is out of a renumerated clause. Agree
is a relation that is sensitive to feature content, and is not subject to intervention
effects caused by elements of a particular type, if they do not have the relevant
content. Furthermore, since Agree does not involve movement, it can hold over
potential island boundaries.
In conclusion, I have accounted for some of the variation found in single whconstructions. The relationship between a probe in Typ and a wh-element is determined by whether or not there is an EPP feature, and the type of EPP feature. Also,
further research is required to account for the French facts; specifically, why it is that
0
0
only a matrix Typ can contain an EPPX feature and why the EPPX in a matrix
Typ is incompatible with an embedded clause that contains an EPPXP feature. In
the next chapter, I examine further influences of wh-features on wh-constructions.
262
CHAPTER 7
Wh-movement that forms a wh-phrase
7.1 Introduction
This chapter examines instances in which movement of a wh-element within a larger
phrase turns that larger phrase into a wh-constituent. For example, the complex-DP
in the Japanese (1) functions as a wh-phrase due to movement of the wh-feature
associated with dare ‘who.’ I return to this type of construction later in this chapter.
(1)
[DP dare
ga
kaita hon]
who-nom wrote book
‘book that who wrote’
I take the position that movement of a wh-element within a larger constituent is
motivated by an EPP feature. For example, in (2), a wh-feature associated with a
wh-phrase embedded inside of a larger XP raises to the head X, thereby turning the
0
XP into a wh-phrase. This wh-feature movement is motivated by an EPPX feature
in the head X.
(2)
XP[+wh]
X’
XEP P X 0
Fwh
YP
. . . wh-phrase+t[Fwh ] . . .
In (3), on the other hand, there is an EPPXP feature in the head X that attracts a
wh-phrase to its specifier position. The wh-feature of the wh-phrase turns the XP
into a wh-phrase.
263
(3)
XP[+wh]
DP1
wh-phrase[Fwh ] XEP P XP
X’
YP
. . . t1 . . .
In this chapter, I examine how this movement of a wh-element that turns a larger
phrase into a wh-phrase can circumvent potential island effects (see chapter 2.6.3
for discussion of island effects).
The organization of this chapter is as follows. In section 7.2, I examine whin-situ languages in which wh-feature movement can result in island circumvention
effects, and in section 7.3, I examine wh-movement languages in which wh-phrasal
movement can lead to island circumvention effects. Section 7.4 is the conclusion.
7.2 Wh-in-situ languages
There are wh-in-situ languages in which a wh-feature associated with a wh-phrase
raises to turn a potential island into a wh-phrase. This process allows a potential
island effect to be circumvented. In this section, I show how a reworked version of
the feature percolation analysis proposed by Nishigauchi (1986, 1990, 1999a, 1999b)
can account for island circumvention effects in the wh-in-situ languages of Japanese
and Korean.
As discussed in chapter 6, in Japanese, there are intervention effects and, at
least for some speakers, wh-island effects. However, complex-DP and adjunct-island
effects are lacking.1 When a wh-phrase is contained within a complex-DP, as in
(4a-b), the result is well-formed. There is no island effect. The wh-phrases darega ‘who-nom’ and donna riyuu-de ‘what reason-for’ can have scope outside of the
1
There are complex-DP and adjunct-island effects with naze ‘why.’ See chapter 9 for discussion.
264
complex-DP.
(4)
Japanese:
(a) Kimi-wa [DP [dare-ga kaita] hon]-o
yomimashita ka?
you-top
who-nom wrote book-acc read
Qu
‘Who1 did you read [books that t1 wrote]? (Nishigauchi 1990:40)
(b) Kare-ga [DP [donna riyuu-de kaita] hon]-ga
omosiroi
he-nom
what
reason-for wrote book-nom interesting
desu ka?
is
Qu
‘[For what reason]1 are [books that he wrote t1 ] interesting?’
(Nishigauchi 1990:92)
Similarly, wh-phrases are not subject to adjunct-island effects. Examples (5a-b)
show that doko-ni ‘where-dat’ and nani-o ‘what-acc’ can occur within an adjunct
clause, labeled as TypP, and have matrix scope.
(5)
Japanese:
(a) Taroo-ga [T ypP doko-ni
itta kara]
umaku itta no?
Taro-nom
where-dat went because well
went Qu
‘Where1 did things go well [because Taroo went t1 ?’ (Richards 2000:187)
(b) Taroo-ga [T ypP nani-o
katta
kara]
umaku itta no?
Taro-nom
what-acc bought because well
went Qu
‘What1 did things go well [because Taro bought t1 ]? (Junko Ginsburg,
p.c.)
If an adjunct clause is renumerated (see discussion of Johnson 2002 in chapter 2.6.2)
and if there is wh-feature movement in Japanese, then the fact that a wh-phrase
can have scope outside of a complex-DP or adjunct clause requires an explanation.
Nishigauchi (1986, 1990, 1999a, 1999b) argues that wh-arguments which occur
inside of DP-islands are able to avoid an island effect, not because a wh-phrase is
not subject to island effects, but rather, because a DP-island can be pied piped at
265
LF in Japanese. According to Nishigauchi (1999b), at LF the structure of (6) is as
shown in (7).
(6)
Kimi-wa [nani-o
katta
hito]-ni
atta no?
you-top what-acc bought person-dat met Qu
‘What1 did you see [the man who bought t1 ? (Nishigauchi 1999b:43)
(7)
LF:
[CP [DP nani1
[t1 -o katta] hito]2 kimi-wa t2 ni
atta no]?
what-acc -acc bought person you-top
dat met Qu
(Nishigauchi 1999b:44)
At LF the wh-phrase nani ‘what’ raises to [Spec, CP] of the relative clause modifier of the noun hito ‘person.’ Next there is feature percolation, whereby “the
wh-feature is percolated through the Spec positions (Nishigauchi 1999a:275)” from
[Spec, CP] of the relative clause up to [Spec, DP], thereby turning the complex-DP
into a wh-phrase. Then the complex-DP raises to [Spec, CP] of the matrix clause.
Since the entire DP becomes a wh-phrase, the wh-phrase does not move out of the
complex-DP, and so an island violation is avoided. Nishigauchi does not, to the
best of my knowledge, discuss adjunct clauses (outside of complex-DPs); however,
his analysis can be applied to these constructions in the same manner. An adjunct
clause becomes a wh-phrase, and moves to [Spec, CP] at LF.
I adopt the spirit of Nishigauchi’s analysis, which is the notion that a potential
island (adjunct clause) containing a wh-phrase becomes a wh-consituent, and as
such, there is no movement of a wh-element outside of the potential island. The
technical details of my analysis, however, are different from those of Nishigauchi, as
I assume that there is no wh-movement at LF. Rather, I take the following position.
In Japanese, a wh-feature associated with a wh-phrase contained within a complexDP or adjunct clause raises to turn the complex-DP or adjunct clause into a larger
266
wh-phrase. Then the wh-feature moves from this newly formed wh-phrase to the
scopal Typ head.
Example (8) gives the structure of a clause with a wh-phrase that is contained
within a complex-DP.
(8)
TypP
Typ’
TP
...
TypEP P X 0
Prb+Fwh , ka/no[FQu ]
DP[+wh]
D’
NP
DEP P X 0
Prb+tFwh
N’
TypP
N’
Typ’
N
TypEP P X 0
Prb+tFwh
TP
. . . wh-phrase+tFwh . . .
The wh-phrase originates inside of an adjunct clause, the embedded TypP. In this
0
case, the embedded Typ head contains an EPPX feature that attracts the whfeature to Typ. Furthermore, the Typ head contains a probe. Valuation of this
probe turns the adjunct clause into a wh-phrase. I assume that this non-interrogative
adjunct clause becomes a wh-phrase because there is no Qu-feature present. In other
words, valuation of a probe by a wh-feature in the absence of a Qu-feature forms
267
a wh-phrase. The adjunct clause is renumerated, but since it is now a wh-phrase,
its wh-feature is still accessible to further operations. The D head of the complex0
DP contains an EPPX feature and a probe. The wh-feature again raises to satisfy
0
the EPPX feature and value the probe, thereby turning the complex-DP into a
0
wh-phrase. Lastly, the matrix Typ contains an EPPX feature and a probe. The
wh-feature thus moves to the matrix Typ and values the probe. In this last case,
though, the wh-feature turns the matrix clause into a wh-construction because of
the presence of a Qu-feature.
The complex-DP island effect in (4a), repeated below, is accounted for as follows.
(4)
Japanese:
(a) Kimi-wa [DP [dare-ga kaita] hon]-o
yomimashita ka?
you-top
who-nom wrote book-acc read
Qu
‘Who1 did you read books that t1 wrote? (Nishigauchi 1990:40)
0
First, an EPPX feature in Typ of the embedded adjunct clause attracts the whfeature and the wh-feature values a probe, thereby turning the adjunct clause into
a wh-phrase, as shown in (9).
(9)
TypP[+wh]
Typ’
TP
TypEP P X 0
Prb+Fwh
dare+tFwh -ga kaita
who-nom wrote
This adjunct clause is renumerated, but since it is now a wh-phrase, its wh-feature
is accessible to further operations; i.e., the wh-feature can move from its position
in Typ of the adjunct clause because the adjunct clause is a wh-phrase. Then the
adjunct clause is Merged with the nominal hon ‘book’ and with a null D head to
268
0
form a complex-DP, as shown in (10). The D head contains an EPPX feature that
attracts the wh-feature from the Typ head of the adjunct clause. Furthermore, the
D head contains a probe that is valued by the wh-feature. Valuation of this probe
turns the complex-DP into a wh-phrase.
(10)
DP[+wh]
D’
NP
DEP P X 0
Prb+Fwh
N’
TypP+tFwh
N’
dare-ga kaita
who-nom wrote
N
hon
book
Then the complex-DP is Merged with the other elements of the numeration, and
the wh-feature in its D head is able to proceed freely to the matrix Typ head to
0
satisfy an EPPX feature and also to value a probe, thereby giving the complex-DP
matrix scope, as shown in (11).
269
(11)
TypP[+wh]
Typ’
TP
T’
DP1
kimi-wa
you-top
TypEP P X 0
Prb+Fwh ,noFQu
vP
t1
v’
DP+tFwh
dare-ga kaita hon-o
who-nom wrote book-acc
T
-mashita
past
v
yomiread
The same analysis accounts for examples such as (5a), repeated below, in which
a wh-phrase occurs within an adjunct-clause.
(5)
Japanese:
(a) Taroo-ga [T ypP doko-ni
itta kara]
umaku itta no?
Taro-nom
where-dat went because well
went Qu
‘Where1 did things go well [because Taroo went t1 ?’ (Richards 2000:187)
As shown in (12), the wh-feature associated with doko-ni ‘where-dat’ raises to the
Typ head of the adjunct clause where it values a probe, and thus turns the adjunct
clause into a wh-phrase. I assume that there is a ForceP that houses kara ‘because.’
Despite the ForceP, movement of the wh-feature can proceed from the adjunct clause
because the entire clause has become a wh-phrase.
270
(12)
ForceP
Force’
TypP[+wh]
Force
kara
because
Typ’
TP
TypEP P X 0
Prb+Fwh
doko+tFwh -ni itta
where-dat went
Then this adjunct clause wh-consituent Merges with the other elements of the clause,
as shown below. The wh-feature raises from the Typ head of the adjunct-clause to
0
the matrix Typ to satisfy an EPPX feature and value a probe, as shown in (13).2
2
I have left out the adjunct umaku ‘well’ for the sake of simplicity. Presumably, umaku ‘well’
is an adverbial that is Merged with the verb.
271
(13)
TypP[+wh]
Typ’
TP
TypEP P X 0
Prb+Fwh ,noFQu
T’
DP1
Taro-ga
you-nom
vP
t1
v’
T
-tta
past
ForceP
v’
Force’
TypP[+wh]
Typ’
TP
Force
kara
because
v
igo
TypEP P X 0
Prb+Fwh
doko+tFwh -ni itta
where-dat went
Movement of the wh-feature is not blocked by intervening v and T heads because
these heads contain elements that are different from a wh-feature in type, in that
they lack a scope bearing quality.
This analysis then relies on the notion that a complex-DP or adjunct clause can
become a wh-phrase in Japanese. Nishigauchi (1990, 1999a) presents evidence that
this is the case with respect to a complex-DP; a wh-construction with a complex-DP,
such as (14), can be answered with a response that refers to the whole complex-DP.
272
(14) Japanese:
Kimi-wa [DP dare-o egaita
hon]-o
yomimashita ka?
you-top
who-acc described book-acc read
Qu
‘You read a book such that it described who?’ (Nishigauchi 1999a:274)
Nishigauchi claims that the following are possible answers to (14).
(15) Japanese:
(a) Gates desu.
Gates be.
‘Its Gates.’
(b) Gates-o
egaita
hon desu.
Gates-acc described book be.
‘(It’s) the book that describes Gates.’ (Nishigauchi 1999a:276)
Nishigauchi (1999a) argues that because (15b) refers to the entire complex-DP, the
entire DP must be pied-piped to [Spec, CP] of the matrix clause. The shortened
answer in (15a) is allowed because “it is a further truncated form of [(15b)] by
means of a discourse deletion rule (276).” Under my analysis, there is no phrasal
movement. The wh-feature associated with the entire DP raises from the DP to the
matrix Typ, giving the complex-DP scope. Thus, the answer in (15b) that refers to
the entire complex-DP is permitted. The availability of (15a) can be accounted for
as a shortened form of (15b), as claimed by Nishigauchi.
The same facts appear to be true when a wh-phrase originates inside of an
adjunct clause, as in (5b), repeated below.
(5)
Japanese:
(b) Taroo-ga [T ypP doko-ni
itta kara]
umaku itta no?
Taro-nom
where-dat went because well
went Qu
‘Where1 did things go well [because Taroo went t1 ?’ (Richards 2000:187)
273
According to my consultant, a response as in (16b) that refers to the entire adjunctclause is not only allowed, but it is better than a response that refers only to the
wh-phrase, as in (16a).
(16)
(a) ?Kyoto-ni.
Kyoto-dat.
‘To Kyoto.’
(b) Kyoto-ni
itta kara.
Kyoto-dat went because.
‘Because he went to Kyoto.’ (Junko Ginsburg, p.c.)
The possibility of the response in (16b) is to be expected if the entire adjunct clause
functions as a wh-constituent. The marginality of (16a) could be an indication that,
at least for my consultant, deletion of the remainder of the adjunct clause is marked.
Korean also behaves in an identical manner to Japanese: complex-DP and
adjunct-island effects are absent.
(17a) shows a wh-phrase contained within a
complex-DP and (17b) shows a wh-phrase in an adjunct clause. According to Yoon
(2006), these are well-formed.
(17) Korean:
(a) [DP [Nwukuwu-ka ssu-n]
chayk]-i
epseci-ess-ni?
who-nom
write-adn book-nom disappear-past-Qu
‘The book that who wrote disappeared?’
(b) Ne-nun [T ypP Yenghi-ka mwues-ul machi-myen] ttena-l
you-top
Yenghi-nom what-acc finish-if
leave
ke-ni?
fut-Qu
‘You are going to leave if Yenghi finishes what?’ (Yoon 2006:387)
These constructions can be accounted for in the same manner as the Japanese data
discussed above. Wh-feature movement allows potential complex-DP and adjunctisland effects to be circumvented.
274
Richards (2000) gives evidence involving the scope of multiple wh-phrases that
supports Nishigauchi’s analysis. He points out that if two wh-phrases are contained
within a complex-DP or adjunct-clause they both must have the same scope, whereas
this is not the case if they are not in a complex-DP. In (18), there are two wh-phrases
contained in an embedded clause of the sort from which wh-movement can proceed
freely. For meaning (i), both dare-ga ‘who-nom’ and dare-o ‘who-acc’ have matrix
scope. For meaning (ii), dare-ga ‘who-nom’ has matrix scope and dare-o ‘who-acc’
has embedded scope.
(18) Japanese:
Keesatu-wa [dare-ga dare-o
korosita ka] sirabeteiru
no?
police-top who-nom who-sc acc killed
Qu are-investigating Qu
(i) ‘For which x and which y are the police investigating [whether x killed y]?
(ii) ‘For which x are the police investigating [for which y, x killed y]?
(Richards 2000:188-189)
In (18), the two embedded wh-phrases can either have scope together or separately.
However, when two wh-phrases occur within a complex-DP, as in (19), both whphrases must have the same scope. (19) can only have the meaning in (i) in which
both wh-phrases have the same scope, and not the meaning in (ii) in which each
wh-phrase has a different scope.
(19) Japanese:
Keesatu-wa [Nakamura-san-ga
[dare-ga dare-o
korosita
police-top Nakamura-hon-nom who-nom who-sc acc killed
tatemono-o] katta
ka] sirabeteiru
no?
building-acc bought Qu are-investigating Qu
(i) ‘For which x and which y are the police investigating [whether
Nakamura-san bought [a house where x killed y]]?
(ii) * ‘For which x are the police investigating [for which y Nakamura-san
bought [a house where x killed y]]? (Richards 2000:189-190)
275
Richards argues that the fact that both wh-phrases must have the same scope when
they are in a complex-DP is predicted if the complex-DP is pied-piped to [Spec, CP]
at LF. If both wh-phrases move together inside of the complex-DP then they should
have the same scope. If there were no pied-piping, then one wh-phrase should be
able to have scope inside of the complex-DP and another wh-phrase should be able
to have scope outside the complex-DP, contrary to fact.
Under my analysis, these facts are accounted for in terms of wh-feature movement, and not pied-piping at LF. I propose that the Typ head of the adjunct clause
0
modifier contained within the complex-DP contains an EPPX feature that attracts
the wh-feature of the higher wh-phrase. Movement of this wh-feature eliminates
0
the EPPX feature. Then, in the absence of an EPP feature, the probe in Typ
forms an Agree relation with the lower wh-feature (see chapter 8.3 for arguments
that a probe forms an Agree relation with a second wh-feature in Japanese). This
secondary Agree relation comes about because the probe searches for all available
wh-features. The TypP then becomes a wh-phrase that has two wh-features. A
diagram is shown below, in which it can be seen that the wh-feature associated
with dare-ga ‘who-nom’ raises to Typ and the wh-feature associated with dare-o
‘who-acc’ forms an Agree relation with the probe in Typ.
(20)
TypP[+wh]
Typ’
TP
TypEP P X 0
Prb[F wh2] +Fwh1
dare-ga+tFw h1 dare-o[F wh2] koroshita
who-nom who-acc killed
0
The D head of the complex DP has an EPPX feature that attracts the multiple
wh-feature bundle from the adjunct TypP, thereby turning the complex-DP into a
single wh-phrase, as shown below.
276
(21)
DP[+wh]
D’
NP
DEP P X 0
Prb+Fwh1,wh2
N’
TypP+tFwh1,wh2
N’
dare-ga dare-o koroshita
who-nom who-acc killed
N
tatemono
building
The wh-feature bundle associated with this newly formed wh-phrase then raises to
the matrix Typ, as shown in the following simplified diagram.
(22)
TypP[+wh]
Typ’
TP
...
TypEP P X 0
Prb+Fwh1,wh2,noFQu
DP+tFwh1,wh2
. . . [dare-ga dare-o koroshita tatemono]. . .
who-nom who-acc killed building
Since the complex-DP functions as a single wh-phrase which has matrix scope, it is
not possible for the lower wh-phrase to have scope in any position other than the
matrix clause.
If a wh-feature can turn a complex-DP or an adjunct clause into a wh-phrase
and thereby circumvent an island effect, then the issue arises of why it cannot do
the same with a whether/if clause. For example, in (23) (originally presented as
277
(31) in chapter 6), the wh-feature associated with nani-o ‘what-acc’ cannot turn
the embedded TypP into a wh-phrase from which wh-feature movement can occur.
(23) Japanese:
*Satookun-wa [Suzukikun-ga nani-o
tabeta ka] oboeteimasu ka?
Sato-top
Suzuki-nom what-acc ate
Qu remember
Qu
‘What1 does Sato remember [whether Suzuki ate t1 ]? (Nishigauchi
1990:30-31)
Richards (2000), who assumes Nishigauchi’s analysis that certain islands can
pied-pipe at LF, argues that if a wh-island were to pied-pipe at LF, then “the whphrase [would be] in two scope positions at once (199).” A wh-island that appears in
[Spec, CP] of a larger clause would contain a wh-phrase that is simultaneously in the
specifier of the island and the specifier of the larger clause at the same time. This
is ruled out because “a single wh-phrase cannot take scope in two different places,
presumably for semantic reasons (Richards 2000:199).” However, under Richard’s
analysis, even within a complex-DP or adjunct clause, a wh-phrase has scope in both
the complex-DP or adjunct clause and [Spec, CP] of a matrix clause. Therefore, I
think there is another explanation.
As I see the facts, a whether/if -island effect cannot be circumvented by whfeature movement for the following reasons given in (24a-b).
(24) A wh-phrase cannot be an interrogative clause - a phrase that is typed as an
interrogative clause by a Qu-feature cannot also be a wh-phrase.
Specifically, if an embedded clause is typed as a wh-construction by both a Qufeature and a wh-feature, then this embedded clause cannot also function as a whphrase that has scope in a higher TypP. For example, (25a) shows that a whphrase can move to [Spec, TypP] in English, but a whether clause cannot. The
ill-formedness of 25b) can be accounted for as resulting from a violation of (24).
278
The clause whether Jane ate what is an embedded wh-construction, containing Quand wh-features in its Typ head. This clause cannot also function as a wh-phrase
that has scope in the matrix TypP.
(25)
(a) What1 does Bill remember t1 ?
(b) [Whether Jane ate what]1 does Bill remember t1 ?
If a wh-feature were to raise to Typ of a whether/if -clause, it would turn the
clause into a larger wh-phrase. For example, wh-feature movement within the embedded clause of (23) would turn the embedded clause into a wh-construction with
a wh-feature in Typ, as shown below.
(26)
TypP[+wh]
Typ’
TP
TypEP P X 0
Prb+Fwh ,kaFQu
Suzukikun-ga nani+tFwh -o tabeta
Suzuki-nom what-acc ate
Then if the wh-feature were to undergo further movement to the matix Typ, as
shown below, it would serve the purpose of giving the whether clause matrix scope.
279
(27)
TypP[+wh]
Typ’
TP
Satookun-wa
Sato-top
TypEP P X 0
Prb+Fwh ,kaFQu
T’
vP
t1
v’
T
-imasu
sc pres
Typ’
v
oboeteremember
TypP[+wh]
TP
TypEP P X 0
Prb+tFwh ,kaFQu
Suzukikun-ga nani+tFwh -o tabeta
Suzuki-nom what-acc ate
But in this case, the whether/if clause, which is an interrogative wh-construction,
is also functioning as a wh-phrase that has matrix scope, in violation of (24) above.
Notably, there are languages in which, like Japanese and Korean, a wh-phrase
remains in-situ, and there is wh-feature movement, but this wh-feature movement
cannot circumvent an island effect.3 For example, wh-phrases can remain in-situ in
Persian, but there are adjunct- and DP-island effects, as shown in (28a-b) (originally
(33a-b) in chapter 6.3).
3
Wh-in-situ languages such as Mandarin, as discussed in chapter 6.2 do not have wh-feature
movement. Rather, a probe in a Typ head is able to form an Agree relation with a wh-feature.
Since there is no movement of a wh-element, there are no island effects.
280
(28) Persian:
(a) Adjunct island:
*Parviz raghsid
[chonke ki
unjâ bud]?
Parviz dance-3sg because who there was
Intended: ‘Who1 did Parviz dance [because t1 was there]?’ (Karimi &
Taleghani 2007:180)
(b) Complex-DP island:
*Kimea [DP pesar-i-ro
[ke to diruz
kojaa did-i]] be
Kimea
boy-rel-raa that you yesterday where saw-2sg to
man moarrefi
kard?
me introduced did
Intended: ‘Where1 did Kimea introduce to me [the boy you saw t1
yesterday]?’ (Simin Karimi, p.c.)
In Persian, then, it appears as though a wh-feature cannot undergo movement that
circumvents an island effect. This can be accounted for if a Typ head within an
adjunct clause in Persian simply cannot contain an EPP feature that can attract a
wh-feature. Therefore, there is a difference between a Typ head that occurs in an
adjunct clause in Persian as opposed to a Typ head that occurs in a non-adjunct
clause.
(29) shows the structure of the adjunct clause of (28a) above.
(29)
TypP
Typ’
Typ
TP
chonke
because ki unjâ bud
who there was
281
Crucially, there is no EPP feature in the Typ head to attract the wh-feature, and
therefore, the wh-feature does not turn this clause into a wh-phrase. Since it is not
a wh-phrase, its wh-feature cannot move out of the clause, which is an island (it has
been renumerated).
If this analysis is on the right track, then there is a parametric difference between
Typ heads of adjunct clauses in different langauges. In languages such as Japanese
and Korean, Typ in an adjunct clause can contain an EPP feature, whereas Typ of
an adjunct clause in languages such as Persian cannot.
Lastly, I focus on the following fact: subject-island effects are absent in languages
such as Japanese, Korean, and Persian.
(30) Japanese:
[[dare-ga kaita] hon]-ga
itiban omosiroi
no?
who-nom wrote book-nom most interesting Qu
‘Who1 are [the books that t1 wrote] most interesting? (Lasnik & Saito
1992:122)
(31) Korean:
Nu-ka
sseun chaek-i kajang chemiit-ni?
who-nom wrote book
most interesting-Qu
‘Who1 are [the books that t1 wrote] most interesting? (Hyun Kyoung Jung,
p.c)
(32) Persian:
[Ye ketaab az kodum nevisande] montasher shod
a
book of which writer
published became.
‘Which writer was [a book of t] published?’ (Simin Karimi, p.c.)
Notably, the wh-feature movement analysis can account for the lack of subject island
effects in Japanese and Korean. The wh-feature raises t o turn the complex-subject
282
into a wh-phrase and then the wh-feature raises to the matrix Typ. However, this
analysis cannot account for Persian, which does not appear to allow wh-feature
movement to circumvent an island effect.
One possibility is that in these languages that lack subject-island effects, a subject actually remains within the v P and that when in a v P internal position, a
subject is not an island for extraction. Karimi (2005) argues that a subject remains within the v P in Persian and that this v P internal position is not an island.
Miyagawa (2001, 2003) also argues that a subject can remain in the v P in Japanese.4
Why exactly a v P-internal subject would not be an island though is not so clear.
Following Koizumi (1995) and Lasnik (2003), Hornstein et al. (2007) write that in
4
According to Miyagawa (2001, 2003), either a subject or object must raise to [Spec, TP]. If
Miyagawa is correct, the following examples raise some questions.
(i) Boru-o [dare-ga tsukutta roboto]-ga nageta no?
ball-acc who-nom made robot-nom threw Qu
‘[The robot that who made] threw the ball?’
(ii) [Dare-ga tsukutta roboto]-ga boru-o nageta no?
who-nom made robot-nom ball-acc threw Qu
‘[The robot that who made] threw the ball?
In (i), the object boru-o ‘ball-acc’ can be in [Spec, TP], in which case the subject remains in
[Spec, v P]. Thus this example does not pose a problem for the view that a v P-internal subject
is not an island. However, in (ii), if Miyagawa is correct, the subject dare-ga tsukutta roboto-ga
‘the robot that who made’ must either be in [Spec, TP] or in a scrambled position above [Spec,
TP], in which case the object boru-o ‘ball-acc’ would be in [Spec, TP]. In either case, the subject
clause must move. If the subject is in [Spec, TP], then it should be an island, contrary to fact, if
only v P-internal subjects are not islands (and if wh-feature movement cannot circumvent an island
effect in Japanese). If the subject is in a scrambled position such as a FocP, then maybe since it
is in an A’-position, it can escape islandhood. On the other hand, as discussed above with respect
to adjunct clauses in Japanese, it could be that a subject does not function as an island because
wh-feature movement can turn a complex subject into a wh-phrase, in which case it would not
matter whether or not the subject has moved out of the v P.
283
Japanese and Korean, “it is reasonable to suppose that the association of subjects
internal to the VP is ultimately small-clausal (155).” Following Moro (2000), they
note that the elements of a small clause likely form a “symmetrical relation” at
some point in a derivation. I assume that the symmetrial relation would be along
the lines of that shown in (33).
(33)
vP
subject
v
If the subject is not an adjunct, in the sense of an adjunct defined by Johnson (2002)
(see (39) in chapter 2.6.2), then it need not be renumerated, and an element can be
extracted from it. However, this notion of a v P-internal subject being in a symmetric
relation with the verb raises some questions. For example, since both the subject
and v are in a symmetric c-command relation in (33), what prevents the label of
(33) from being that of the subject (e.g., DP) and not v P? Also, it is not clear where
an object would occur if the subject and verb are in a symmetrical relation. Lacking
any clear solution at this time, I leave this issue for further analysis.
In summary, a wh-phrase can occur within an adjunct clause in wh-in-situ lan0
guages such as Japanese and Korean because these languages allow an EPPX fea0
ture to occur within the Typ head of an adjunct clause. This EPPX feature forces
movement of the wh-feature to the edge of the adjunct clause and thus allows the
wh-feature to undergo further movement that is not blocked by renumeration of the
adjunct-clause. In this section, I have examined how movement of a wh-feature can
turn a larger phrase into a wh-phrase in languages such as Japanese and Korean,
but not in Persian. In the next section, I examine how movement of a wh-phrase
(rather than a wh-feature) within a larger phrase can turn the larger phrase into a
wh-phrase and circumvent potential island effects in some languages.
284
7.3 Overt wh-movement languages
In this section, I examine movement of a wh-phrase in certain island constructions.
I show that wh-phrasal movement can circumvent potential island effects in Basque
and Tlingit. This wh-phrasal movement is very similar to the wh-feature movement
of the previous section. However, there is one big difference; unlike with wh-featuer
movement, when there is wh-phrasal movement that circumvents an island effect,
there is clear surface movement of a wh-phrase. I also show how wh-phrasal movement cannot circumvent potential island effects in English.
7.3.1 Basque
Richards (2000:195) points out, following Ortiz de Urbina (1989), that in Basque,
when a wh-phrase is contained within a complex-DP or adjunct-clause, a potential
island effect can be avoided by overtly “pied piping the potentially offending island.”
In (34a), an entire adjunct clause has fronted to clause initial position, whereas
in (34b), only the wh-phrase has moved out of the clause. Fronting of the adjunct
clause is better than movement of the wh-phrase out of the adjunct clause.5
(34) Basque:
(a) ?[T ypP Zer ikusi ondoren] joan ziren hemen-dik?
what see after
go
aux here-from
‘What did they leave [after seeing]?’
5
The original source does not indicate the base position of the adjunct clause in (34a), but it
is most likely base generated in clause-final position, as in (34b). Ortiz de Urbina (1989, 2003)
claims that the basic word order in Basque is generally thought to be SOV, in which case, the base
position of the object wh-phrase in (34b) would most likely be directly preceding the verb inside
of the adjunct clause.
285
(b) *Zer joan ziren hemen-dik [T ypP ikusi ondoren]?
what go
aux here-from
see after
Intended: ‘What1 did they leave [after seeing t1 ]?’ (Ortiz de Urbina
1989:249,252, per Richards 2000:195)
In (34a), I propose that the wh-phrase raises to the specifier of the adjunct clause.
From this position, the wh-phrase’s wh-feature turns the adjunct clause into a whphrase by establishing a Spec-head relation with a Probe in the Typ head. (35)
shows the internal structure of the adjunct clause, which I assume is a PP. The
wh-phrase zer ‘what’ raises to the specifier of the PP to satisfy an EPPXP feature
in the P head. Once the wh-phrase arrives in [Spec, PP], its wh-feature forms a
Spec-head relationship with a probe in the P head, and this wh-feature turns the
entire PP into a wh-phrase.
(35)
PP[+wh]
DP1
zer[Fwh ]
what
P’
TypP
t1 ikusi
see
PEP P XP
ondoren,Prb[Fwh ]
after
Since the PP is now a wh-phrase, it is able to move directly to [Spec, TypP] of
the matrix clause. This latter movement is motivated by an EPPXP feature in the
matrix Typ. There is no movement of a wh-element out of the adjunct clause, and
thus, this example is well-formed. A diagram is shown below.
286
(36)
TypP
PP1[Fwh ]
Typ’
zer ikusi ondoren TP
what see after
TypEP P XP
Prb[Fwh ]
joan ziren hemen-dik t1
go aux here-from
In the ill-formed (34b), the wh-phrase moves overtly out of an adjunct clause that
has been renumerated, thereby resulting in ill-formedness.
Examples (37a-b) show that the same results are obtained when a wh-phrase
is base generated inside of a complex-DP; fronting a complex-DP is better than
moving a wh-phrase out of it.6
(37) Basque:
(a) ?[DP nork
idatzi zuen liburua] irakurri du
Peruk?
who-erg write aux book
read
aux Peter-erg
‘Who1 did Peter read the book that t1 wrote?’
(b) *Nork
irakurri du
Peruk
[DP idatzi zuen liburua]?
who-erg read
aux Peter-erg
write aux book
‘Who1 did Peter read the book that t1 wrote?’ (Ortiz de Urbina
1989:249,252, per Richards 2000:195)
First of all, the wh-phrase raises to the specifier of the TypP of the adjunct clause
contained within the complex-DP, as shown in (38), thus turning the adjunct clause
into a wh-phrase.
6
Assuming an SOV basic word order, per Ortiz de Urbina (1989, 2003), the wh-phrase is most
likely base generated in the position preceding the verb in the complex-DP.
287
(38)
TypP[+wh]
DP1
nork[Fwh ]
who-erg
Typ’
TP
TypEP P XP
Prb[Fwh ]
t1 idatzi zuen liburua
write aux book
Then the adjunct clause is merged with the nominal liburua ‘book’ and with a null
D head. Next, I propose that the TypP raises to [Spec, DP]. By doing so, the whfeature of the newly formed wh-phrase forms a Spec-head relation with the D head,
thereby turning the complex-DP into a wh-phrase.
(39)
DP[+wh]
TypP[Fwh ]
D’
nork idatzi zuen liburua NP
who-erg write aux book
DEP P X 0
Prb[Fwh ]
N’
t
N’
N
liburua
book
Next, the complex-DP moves to [Spec, TypP] of the matrix clause, where it satisfies
an EPPXP feature in the matrix Typ, and its wh-feature values a probe in Typ via
a Spec-head relation, as shown below.
288
(40)
TypP[+wh]
DP1[Fwh ]
nork idatzi zuen liburua
who-erg write aux book
Typ’
TP
Typ
Prb[Fwh ]
t1 irakurri du Peruk
read aux Peter-erg
There is no movement of a wh-element out of a complex-DP, and thus the example is
well-formed. (37b) is ill-formed because the wh-phrase moves out of a renumerated
adjunct clause contained within the complex-DP.
Also of interest is that in Basque, a whether/if -clause cannot be moved to circumvent an island effect, as shown in (41a-b). Ill-formedness results regardless of
whether the island moves (41a), or only the wh-phrase moves out of the island (41b).
(41) Basque:
(a) *[Nor etorrko d-en] galdetu duzu?
who
come
aux-Q asked aux
‘Who have you asked whether t has come?’
(b) *Nor galdetu duzu [etorriko d-en]?
who
asked aux come
aux-Q
‘Who have you asked whether t has come?’ (Ortiz de Urbina
1990:199-200, per Richards 2000:195)
The ill-formedness of this type of construction likely has the same cause as in
Japanese. A wh-phrase cannot also be an interrogative clause (see (24) above).
7.3.2 Tlingit
Another language that appears to have overt pied-piping that can circumvent an
island effect is Tlingit, which, as discussed in chapter 6.4, is a language with overt
289
wh-movement and an overt Qu-morpheme. In (42) (originally given as (44) in
chapter 6.4), I assume that the wh-phrase daa is in [Spec, TypP] and the Qumorpheme sá is in Typ.
(42) Tlingit:
Daa sá kéet
ax/’a?
what Qu killerwhale he.eats.it
‘What do killerwhales eat?’ (Cable 2007:64)
Cable (2007) shows that island effects occur in Tlingit when a Qu-morpheme
is contained within a potential island, such as a complex-DP, but are absent when
the Qu-morpheme is at the edge of the potential island. As Cable notes, this is
similar to what one finds in Sinhala and Okinawan (see chapter 6.3). Example (43a)
demonstrates that a wh-phrase can occur inside of a complex-DP that has fronted
to the beginning of a clause, and that is followed by a Qu-morpheme. (43b-c) show
that the Qu-morpheme cannot occur within the complex-DP, neither adjacent to
nor separated from the wh-phrase.
(43) Tlingit:
tuwáa sigóo?
(a) [DP Wáa kligéiyi
xáat] sá i
how it.is.big.REL fish Qu your spirit it.is.happy
‘How big a fish do you want?’ (A fish that is how big do you want?)
(b) *[DP Wáa sá kligéiyi
xáat] i
tuwáa sigóo?
how Qu it.is.big.REL fish your spirit it.is.happy
tuwáa sigóo?
(c) *[DP [Wáa kligéiyi]
sá xáat] i
how
it.is.big.REL Qu fish your spirit it.is.happy
(Cable 2007:79)
Cable (2007) proposes that a Qu-morpheme is the head of what he refers to as
a QP projection (not to be confused with a Quantifier Phrase), and that the Qumorpheme has a wh-phrasal complement. Thus, a wh-phrase and a Qu-morpheme
290
are base generated together. The Qu element of the QP, not the wh-phrase, is
attracted to [Spec, CP]. The wh-phrase fronts with the Qu-morpheme since it is
contained within the QP.
As I see it, the Tlingit data can be accounted for in another manner. Cable
views the Qu-morpheme as always being base generated in a position adjacent to a
wh-phrase. However, I take the fact that a Qu-morpheme can appear in a position
separate from a wh-phrase to be evidence that this is not the case. Rather, a Qumorpheme by default is Merged directly in Typ, and a wh-phrase raises from a TPinternal position to [Spec, TypP]. Wh-phrasal movement is motivated by an EPPXP
feature in Typ. An advantage of my proposal is that it accounts for both yes/no and
wh-constructions. In both types of interrogatives, a Qu-morpheme is Merged in Typ.
Cable’s proposal that a Qu-morpheme is always Merged in a position adjacent to a
wh-phrase requires that a yes/no Qu-morpheme appear in a different position than
a Qu-morpheme that appears in a wh-construction; i.e., a yes/no Qu-morpheme
is Merged at the edge of the clause, but a Qu-morpheme that appears in a whconstruction is Merged TP-internally. However, as I see it, the Qu-morpheme in
both a yes/no and a wh-construction is the same type of element, and is Merged in
exactly the same position.
Wh-phrasal movement accounts for the lack of island effects in (43a). As shown
in (44), an EPPXP feature in the Typ head of the adjunct clause attracts the whphrase to [Spec, TypP]. A Spec-head relation between the wh-phrase and Typ turns
the TypP into a wh-phrase.
291
(44)
TypP[+wh]
DP1
Typ’
wáa[Fwh ]
how
TP
TypEP P XP
Prb[Fwh ]
t1 kligéiyi
it.is.big.REL
Next, the adjunct clause TypP raises to the specifier of the complex-DP. Again,
a spec-head relation between the moved adjunct clause and the D head turns the
complex-DP into a wh-phrase, as shown in (47).
(45)
DP[+wh]
TypP[Fwh ]
wáa kligéiyi
how it.is.big.REL
D’
NP
DEP P X 0
Prb[Fwh ]
N’
t
N’
N
xáat
fish
Lastly, the complex-DP, which is now a wh-phrase, raises to the specifier of the
matrix TypP.
292
(46)
TypP
DP1[Fwh ]
Typ’
wáa kligéiyi xáat
how it.is.big.REL fish
TP
Typ
sá[FQu ] , Prb[Fwh ]
t1 i tuwáa sigóo
your spirit it.is.happy
In this manner, there is no movement of a wh-element out of a potential island;
rather, the potential island (the adjunct clause contained within the complex-DP)
moves, thus circumventing an island effect.
The fact that the Qu-morpheme must follow the complex-DP, as in (43a), is also
accounted for. The complex-DP is in [Spec, TypP], and the Qu-morpheme is in
Typ. If the Qu-morpheme occurs within the complex-DP, as in examples (43b-c),
ill-formedness results because the Qu-morpheme is not in Typ, but rather within a
wh-phrase that is in [Spec, TypP], and the Qu-morpheme must appear in Typ to
type the clause.
7.3.3 English
English is also a language with overt wh-phrasal movement. However, unlike in
Basque and Tlingit, an island effect cannot be circumvented.
Examples (47a-c) show a wh-phrase that originates in a complex-DP.
(47)
(a) *Who1 did you read [DP books [that t1 wrote]]?
(b) *[DP Books [that who wrote]]1 did you read t1 ?
(c) *[DP Who2 books [that t1 wrote]]2 did you read t1 ?
In (47a), the wh-phrase moves out of a complex-DP, thereby resulting in illformedness. To determine if a wh-feature can turn a complex-DP into a wh-phrase
293
and circumvent an island effect, it is necessary to examine whether or not a complexDP can be moved to clause-initial position. Example (47b) is alright as an echo
question, but ill-formed as a wh-question. It would not be fine if one were asking for
the name of the author of a book. Thus, movement of a wh-feature cannot turn a
complex-DP into a wh-phrase and circumvent an island effect. Example (47c) is one
in which the wh-phrase has moved to the initial position of the complex-DP and the
complex-DP has moved to the initial position of the matrix clause. Unlike in Basque
and Tlingit, in which movement of a wh-phrase within a complex-DP can circumvent a potential island effect, this clearly is not an option in English. Movement
of the wh-phrase within the complex-DP and further movement of the complex-DP
results in utter gibberish. Movement of a wh-phrase within a complex-DP therefore
cannot turn the complex-DP into a larger wh-phrase.
The facts are identical with other types of clauses that function as islands. Examples (48a-c) show constructions with an adjunct clause.
(48)
(a) *What1 did things go well [T ypP because he bought t1 ]?
(b) *[T ypP Because he bought what]1 did things go well t1 ?
(c) *[T ypP What2 because he bought t2 ]1 did things go well t1 ?
Examples (49a-c) show a wh-phrase contained within a whether/if clause.
(49)
(a) *What1 does he remember [T ypP whether she ate t1 ] ?
(b) *[T ypP Whether she ate what]1 does he remember t1 ?
(c) *[T ypP What2 whether she ate t2 ]1 does he remember t1 ?
Movement of the potential island cannot circumvent an island effect in English.
In English, the inability for wh-phrasal movement to circumvent an island effect
indicates that the head of a potential island cannot contain an EPP feature that
attracts a wh-element.
294
7.4 Conclusion
In summary, languages such as Japanese, Korean, Basque, and Tlingit allow movement of a wh-element, either a wh-feature or a wh-phrase, to circumvent a potential
island effect. In Japanese and Korean, movement of a wh-feature to the head of a
clause that is a potential island turns the clause into a wh-phrase. In Basque and
Tlingit, movement of a wh-phrase to the specifier of a clause that is a potential island
turns the clause into a wh-phrase. Once this clause becomes a wh-phrase, there is
either movement of a wh-feature from it (in Japanese and Korean), or movement of
the entire clause (Basque and Tlingit). Since this movement does not pass through
an island, it is well-formed and an island effect is avoided.
Other languages such as Persian and English do not allow movement of a whelement to circumvent an island effect. I proposed that this is because the heads
of these potential islands simply cannot contain an EPP feature that attracts a
wh-element.
I next turn to a discussion of constructions in which there are multiple whphrases.
295
CHAPTER 8
Multiple wh-constructions
8.1 Introduction
Many languages allow multiple wh-constructions in which multiple wh-phrases have
scope and, presumably, multiple wh-features establish a relationship with Typ. The
facts regarding multiple wh-constructions are extraordinarily complicated, with respect to which wh-phrases can occur where, scope relations, island effects, intervention effects, etc. (for example, see Richards (2001)). Due to lack of space, I leave
most of these issues aside and focus on answering the following question: How do
multiple wh-phrases establish scope in a single clause within the system outlined
in this work? I also briefly examine why there are languages which do not allow
multiple wh-constructions.
The organization of this chapter is as follows. Section 8.2 discusses languages in
which all wh-features form an Agree relation with Typ. I next discuss multiple whconstructions in Japanese and English in sections 8.3 and 8.4, respectively. I argue
that in these languages, movement of an initial wh-element to Typ allows a probe
in Typ to form an Agree relation with another wh-element. Section 8.5 discusses
previous analyses of the multiple wh-construction facts in Japanese and English
and compares them with the analysis presented here. Section 8.6 discusses multiple
wh-constructions in Persian, a language that does not pattern with Japanese or
English with respect to multiple wh-constructions. Section 8.7 examines languages
that disallow multiple wh-constructions altogether. Section 8.8 is the conclusion.
296
8.2 Agree languages
Sinhala and Mandarin are languages that allow multiple wh-constructions. (1) is an
example from Sinhala in which there are two wh-phrases in a matrix clause.
(1)
Sinhala:
Kiidenek pot
kiiyak
gatta
[email protected]?
how many books how many bought Qu
‘How many people bought how many books?’ (Sumangala 1992:236, per
Hagstrom 1998:66)
Example (2) from Mandarin also shows that multiple wh-phrases can have scope in
a single clause.
(2)
Mandarin:
Shei mai-le
shenme (ne)?
who buy-asp what
Qu
‘Who bought what?’ (Cheng 1997:106)
In chapter 6.2, I claimed that in these languages, a wh-feature froms an Agree
relation with a probe in Typ. Evidence is the lack of island effects in single whconstructions. In a multiple wh-construction, evidence suggests that the probe in
Typ can form an Agree relation with multiple wh-features.
Evidence that Agree is at work is that in a multiple wh-construction, none of
the wh-phrases are subject to island effects. In the Sinhala example (3) below,
both [email protected] ‘who’ and mokak ‘what’ have matrix scope even though the wh-phrase
mokak ‘what’ is contained within a potential island because of the c-commanding
wh-phrase koheedi ‘where.’ This example is fine as long as the Qu-morpheme is
outside of the embedded clause, indicating that koheedi ‘where’ does not prevent
mokak ‘what’ from obtaining matrix scope.1
1
See chapter 5.5 for discussion of the Qu-morpheme and island effects in Sinhala wh-
constructions.
297
(3)
Sinhala:
[email protected] [api koheedi mokak gatta
kiyala] [email protected]@
[email protected]?
who
we where what
buy-past that
remember Qu
‘Who remembers where we bought what?’ (Sumangala 1992:183)
In the Mandarin (4), according to Richards (2001), the lowest wh-phrase shei3 ‘who,’
which is contained within a potential wh-island, can have matrix scope together with
the highest shei1 ‘who’; an answer refers to both the lowest and highest shei ‘who.’
Therefore, shei2 ‘who’ within the embedded clause does not prevent the lowest shei3
‘who’ from having matrix scope.
(4)
Mandarin:
Shei1 xiang-zhidao [shei2 sha-le
shei3 ]?
who
want-know who
kill-perf who
‘Who wants to know who killed who?’ (Richards 2001:248)
The fact that a wh-phrase can occur in an embedded clause of this sort without
resulting in an island effect indicates that Typ forms an Agree relation with its
wh-feature.
I propose that in multiple wh-constructions in these languages, a Typ head forms
an Agree relation with multiple wh-features. For example, in the Mandarin example
(2) above, the wh-feature associated with shei ‘who’ forms an Agree relation with
the probe in Typ and then the wh-feature associated with shenme ‘what’ also forms
an Agree relation with the probe in Typ. A diagram is shown below. Agree is not
sensitive to island effects because there is no movement of a wh-element.
298
(5)
TypP
Typ’
TP
Typ
Prb[Fwh1 ,Fwh2 ]
shei[Fwh1 ] mai-le shenme[Fwh2 ]
who bought what
In this manner, in certain wh-in-situ languages, a single Typ head is able to
undergo an Agree relation with multiple wh-features. I next turn to languages in
which there is wh-feature movement.
8.3 Wh-feature movement followed by Agree: Japanese
Japanese allows multiple wh-constructions such as (6) below.
(6)
Japanese:
Dare-ga nani-o
katta no?
who-nom what-acc Qu
‘Who bought what?’
In chapter 6.3, I argued that in a single wh-construction in Japanese, a wh-feature
associated with an in-situ wh-phrase raises to Typ. In a multiple wh-construction,
there are two possibilities for how the secondary wh-phrase obtains scope. One is
that its wh-feature moves to Typ and the other is that its wh-feature forms an Agree
relation with a probe in Typ. Evidence indicates that Agree is at work.
Watanabe (1992a) shows that the addition of a wh-phrase outside of a potential
island can eliminate an island effect in Japanese. Following Saito (1994) and Tanaka
(1997), I refer to this phenomenon as an ‘additional wh-effect.’ The marginality of
(7a) is attributed to an island effect since nani-o ‘what-acc’ is contained within a
299
whether/if -clause. When the wh-phrase dare-ni ‘who-dat’ is added outside of the
whether/if -clause in (7b), ill-formedness disappears.
(7)
(a) ??John-wa [Mary-ga nani-o
katta
kadooka] Tom-ni
John-top Mary-nom what-acc bought whether Tom-dat
tazuneta no?
asked
Qu
‘What1 did John ask Tom [whether Mary bought t1 ]?’
(b) John-wa [Mary-ga nani-o
katta
kadooka] dare-ni
John-top Mary-nom what-acc bought whether who-dat
tazuneta no?
Qu
asked
‘Who did John ask [whether Mary bought what]?’ (Watanabe
1992a:263)
The marginality of (7a) is accounted for as discussed in chapter 6.3. An EPPX
0
feature in the matrix Typ is unable to attract the wh-feature associated with nani-o
‘what-acc’ due to the intervening kadooka ‘whether’ in the embedded Typ head,
thereby resulting in an MLC violation.
0
I propose that the well-formedness of (7b) is due to the fact that the EPPX feature in Typ has been satisfied and eliminated. That is, the wh-feature of the matrix
wh-phrase dare-ni ‘who-acc’ moves to the matrix Typ head where it eliminates the
0
EPPX feature, as shown in (8) below; the slashes through the EPP feature indicate
that it has been eliminated.
(8)
TypP
Typ’
TP
TypE/ P/ P/ X 0
Prb+Fwh1,no[FQu ]
John-wa . . . dare-ni+t[Fwh1 ] tazuneta no
John-top . . . who-dat asked Qu
300
0
Once the EPPX feature is eliminated, the Typ head is able to form an Agree relation
with the wh-feature of the wh-phrase contained within the island, as shown in (9).
(9)
TypP
Typ’
TP
...
Typ
Prb[Fwh2 ] +Fwh1,no[FQu ]
TP
Mary-ga nani-o[Fwh2 ] katta kadooka
Mary-nom what-acc bought whether
In this manner, the probe in the matrix Typ is valued by two wh-features, thereby
giving two wh-phrases scope. An initial wh-feature moves to the probe and a second
wh-feature forms an Agree relation with it.
This same additional wh-effect can be seen with respect to intervention effects in
0
Japanese. Example (10a) is ill-formed because the EPPX feature in the matrix Typ
is unable to attract the wh-feature associated with nani-o ‘what-acc’ due to the
intervening sika ‘only,’ thereby resulting in an MLC violation. However, when a whphrase is added to the matrix clause, the intervention effect disappears, as shown in
(10b), in which both the matrix and embedded wh-phrases can have matrix scope.2
(10)
2
(a) ?*Taroo-wa [Hanako-sika nani-o
yoma-nai to]
Taroo-top Hanako-only what-acc read-neg comp
Tomoko-ni
itta no?
Tomoko-dat told Qu
‘What did Taroo tell Tomoko that only Hanako read?’
I gloss sika as ‘only’ but Tanaka (1997) glosses it as ‘Foc.’
301
(b) Taroo-wa [Hanako-sika nani-o
yoma-nai to]
dare-ni
Taroo-top Hanako-only what-acc read-neg comp who-dat
itta no?
told Qu
‘Who did Taroo tell Tomoko that only Hanako read what?’ (Tanaka
1997:165)
In the well-formed (10b), the wh-feature associated with the matrix wh-phrase dare0
ni ‘who-dat’ raises to Typ where it satisfies and eliminates the EPPX feature, as
shown below.
(11)
TypP
Typ’
TP
TypE/ P/ P/ X 0
Prb+Fwh1 ,no[FQu ]
Taroo-wa . . . dare-ni+twhF 1 itta no
Taroo-top . . . who-dat told Qu
0
Since the EPPX feature is no longer present, a probe in the matrix Typ head is
able to Agree with the lower wh-feature associated with nani-o ‘what-acc.’ Since
this is an Agree relation, the intervening sika ‘only’ does not cause an intervention
effect. Agree is sensitive to feature content. There is no MLC violation because the
probe in Typ searches for a wh-feature and is not sensitive to intervening elements
of a similar type. A simplified diagram is shown in (12).
302
(12)
TypP
Typ’
...
Typ
Prb[Fwh2 ] +Fwh1,no[FQu ]
TypP
...
QuantP
Quant’
TP
Hanako-sika+t[FQuant ]
Hanako-only
Quant
Probe[FQuant ]
T’
vP
T
nani-o[whF 2] . . .
what-ACC
0
Thus, in Japanese, movement of a higher wh-feature eliminates an EPPX feature,
thereby allowing a probe in Typ to Agree with a lower wh-feature.
8.4 Wh-phrasal movement followed by Agree: English
In chapter 6.4, I argued that in a single wh-construction in English, an EPPXP
feature in Typ forces movement of a wh-phrase to [Spec, TypP]. As is well-known,
only one wh-phrase undergoes movement in English. Therefore, the EPPXP feature
clearly cannot attract multiple wh-phrases. For example, (13a-b), in which both
wh-phrases have moved (assuming that a subject wh-phrase moves in English - see
chapter 5.3), are ill-formed. The correct way to form this construction requires that
the lower wh-phrase remain in-situ, as in (13c).
303
(13)
(a) *Who1 what2 t1 bought t2 ?
(b) *What2 who1 t1 bought t2 ?
(c) Who1 t1 bought what?
In English, a second wh-phrase then is able to obtain scope without overt movement.
The facts indicate that, as in Japanese, the wh-feature of this second wh-phrase
forms an Agree relation with a probe in Typ.
Mulitiple wh-constructions in English show additional wh-effects, just like in
Japanese. Example (14a) is ill-formed due to the inability of the EPPXP feature in
the matrix Typ to attract what. The intervening wh-phrase who results in an MLC
violation. However, example (14b) with an additional wh-phrase in the matrix clause
is fine.
(14)
(a) *What1 do you wonder [who bought t1 ]?
(b) Who1 t1 wonders [who bought what]? (Brody 1995, per Richards
1998:600)
The well-formedness of (14b) suggests that the wh-feature associated with the matrix
who eliminates an EPPXP feature in Typ, thereby allowing Typ to form an Agree
relation with the wh-feature of what in the embedded clause. Since this is an Agree
relation, what is able to have matrix scope despite being inside of a potential whisland. As shown in (15), the higher wh-phrase raises to [Spec, TypP], where it
satisfies and eliminates an EPPXP feature.
304
(15)
TypP
DP1
Typ’
who[Fwh1 ] TypE/ P/ P/ XP
Prb[Fwh1 ] ,∅F[Qu]
TP
t1
T’
T
vP
t1
v’
v
wonders
...
Since the EPP feature is eliminated, the probe in Typ is able to form an Agree
relation with the wh-feature of the wh-phrase what that is contained within the
embedded wh-island, as shown below.
(16)
TypP
DP1
Typ’
who[Fwh1 ]
Typ
Prb[Fwh1 ,Fwh2 ] ,∅F[Qu]
TP
...
TypP
who bought what[Fwh2 ]
In this manner, the probe in Typ is valued by multiple wh-features.
The following examples demonstrate the same phenomenon with a complex-DP.
In (17a), which car moves out of a complex-DP, resulting in ill-formedness since
the relative clause has been renumerated. In (17b), the matrix clause contains an
305
additional wh-phrase who. Although which car remains in-situ, it can have matrix
scope, indicating that the DP-island effect has disappeared.
(17)
(a) *[Which car]1 did John persuade [DP the man who bought t1 ] to sell
the hubcaps?
(b) Who1 t1 persuaded [DP the man who bought which car] to sell the
hubcaps? (Richards 1998:605)
The well-formedness of (17b) results from the addition of a wh-phrase to the matrix
clause. In this case, the matrix wh-phrase who raises to [Spec, TypP] where it eliminates the EPPXP feature. Then the probe in Typ is able to form an Agree relation
with the wh-feature associated with which car contained within the complex-DP.
8.5 Previous analyses of additional wh-effects
I have argued that in Japanese and English, once an initial wh-element (a whfeature in Japanese, and a wh-phrase in English) moves to TypP, the EPP feature
is eliminated, and a probe in Typ can form an Agree relation with another whfeature. This analysis thus accounts for additional wh-effects, as in (7a-b), (10a-b),
(14a-b), and (17a-b), whereby an additional wh-phrase that c-commands an island or
intervening quantificational element improves the well-formedness of a construction.
My analysis notably differs from other analyses.
Huang (1982) argues that LF movement is not subject to Subjacency (see chapter
2.6.1 for discussion of Subjacency) but overt movement is. For example, (18a) is illformed because there is overt movement of the wh-phrase who out of a complex-DP.
In (18b), the wh-phrase who contained within the complex-DP has matrix scope
together with the matrix who. Huang argues that this is fine because the lower who
moves at LF.
(18)
(a) *Who1 do you like [books that criticize t1 ]?
306
(b) Who likes [books that criticize who]? (Huang 1982:492)
Watanabe (1992b) follows Huang’s (1982) notion that LF movement is not subject to island effects, despite the apparent existence of island effects in wh-in-situ
languages such as Japanese. The difference between Japanese and English, Watanabe argues, is that Japanese wh-phrases have a null wh-operator, whereas English
wh-phrases do not. Both overt wh-movement in English and overt null wh-operator
movement in Japanese are subject to Subjacency.
Constructions such as (7a-b), repeated below, are accounted for by Watanabe as
follows.
(7)
(a) ??John-wa [Mary-ga nani-o
katta
kadooka] Tom-ni
John-top Mary-nom what-acc bought whether Tom-dat
tazuneta no?
asked
Qu
‘What1 did John ask Tom [whether Mary bought t1 ]?’
(b) John-wa [Mary-ga nani-o
katta
kadooka] dare-ni
John-top Mary-nom what-acc bought whether who-dat
tazuneta no?
asked
Qu
‘Who did John ask [whether Mary bought what]?’ (Watanabe
1992a:263)
In (7a), a null operator associated with nani-o ‘what-acc’ moves overtly out of the
embedded clause to [Spec, CP]. This movement passes over two bounding nodes,
and therefore violates Subjacency. In (7b), on the other hand, the null wh-operator
associated with dare-ni ‘who-dat’ in the matrix clause raises to the matrix [Spec,
CP]. Then the wh-phrase nani-o ‘what-acc’ moves at LF out of the embedded
clause and adjoins to the operator in the matrix [Spec, CP] position. Since this
movement is at LF, it is not subject to island effects. Crucially, Watanabe assumes
that if [Spec, CP] is filled at S-structure, then movement of another wh-phrase to
[Spec, CP] can occur at LF.
307
Tanaka (1997) similarly accounts for additional wh-effects in terms of null whoperator movement. Whereas Watanabe discusses island effects, Tanaka focuses on
intervention effects, as in (10a-b), repeated below.
(10) Japanese
(a) ?Taroo-wa [Hanako-sika nani-o
yoma-nai to]
Tomoko-ni
Taroo-top Hanako-only what-acc read-neg comp Tomoko-dat
itta no?
told Qu
‘What did Taroo tell Tomoko that only Hanako read?’
(b) Taroo-wa [Hanako-sika nani-o
yoma-nai to]
dare-ni
Taroo-top Hanako-only what-acc read-neg comp who-dat
itta no?
told Qu
‘Who did Taroo tell Tomoko that only Hanako read what?’ (Tanaka
1997:165)
Tanaka’s analysis relies on the notion that intervention effects arise in a construction such as (10a) because a null operator associated with NPI sika ‘only’ blocks
movement of the null wh-operator associated with nani-o ‘what-acc.’ In (10b) the
null wh-operator associated with dare-ni ‘who-dat’ in the matrix clause moves to
[Spec, CP] at S-structure. Then the wh-phrase in the embedded clause moves at
LF to the matrix [Spec, CP]. Since this movement is at LF, it is not blocked by the
operator associated with sika ‘only.’
My proposal has the advantage of not relying on LF movement nor on null whoperator movement. Huang (1982), Watanabe (1992b), and Tanaka (1997) all rely
on the notion that LF movement of a wh-phrase is not subject to island effects. A
problem for this view is that it is not clear why LF movement should be so special
and not subject to island effects. Under my proposal, there is no recourse to LF
movement. An island effect does not occur when a probe can form an Agree relation with a wh-feature, and this Agree relation occurs before Spell-Out. Watanabe
308
(1992b) and Tanaka (1997) also rely on the notion that in Japanese wh-constructions
there is movement of a null wh-operator. However, the same facts can be accounted
for as resulting from movement of a wh-feature, without recourse to the notion of
null wh-operator movement.
Richards (1998, 2001) takes a different view of these additional wh-effects. He
relies on what he refers to as the “Principle of Minimal Compliance,” which is the
notion ‘that a given constraint only has to be satisfied once in a certain domain
(Richards 2001:197).” The Principle of Minimal Compliance applies to a wide variety of phenomena, such as reflexivity, weak crossover, VP-ellipses, etc. Richards
(2001:607) writes that “it appears to be true quite generally that in cases involving
multiple wh-movement to a single wh- complementizer, only the first-moved wh-word
will have to obey Subjacency; the other wh-movements are free from Subjacency.”
When there is an additional wh-effect, as in (7b) and (10b), movement of an initial
wh-phrase to [Spec, CP] is well-formed and obeys Subjacency. Thus, Subjacency
can be ignored with respect to movement of a lower wh-phrase.
My analysis on the one hand, makes no recourse to the notion that once a constraint is satisfied, it may be ignored. However, it may be that my analysis is
compatible with Richard’s proposal in the sense that it explains why the Principle of Minimal Compliance holds with respect to wh-constructions. According to
Richards, well formed movement of a wh-element that does not violate an island
effect allows movement to occur that is not subject to an island effect. As I see it,
well-formed movement of a wh-element eliminates an EPP feature, and allows an
Agree relation which would not normally be allowed. Thus, my analysis accounts
for why a particular constraint that results in island effects can be ameiliorated.
In summary, unlike my analysis, other proposals rely on the notion that LF
movement is not subject to island effects, movement of a null wh-operator, and the
Principle of Minimal Compliance. Rather, I claim that additional wh-effects result
from movement of a wh-element that eliminates an EPP feature in Typ and from
309
the ability of a probe in Typ to form an Agree relation with another wh-feature.
8.6 Multiple wh-feature movement
Persian also allows multiple wh-constructions, as in (19) below.
(19) Persian
Ki chi-ro xarid?
who what
bought-3sg
‘Who bought what?’ (Karimi 2005:145)
In chapter 6.3, I claimed that Persian is a language in which there is wh-feature
0
movement driven by an EPPX feature in Typ. Evidence suggests that unlike in
Japanese and English, the wh-feature of a secondary wh-phrase undergoes feature
movement, and there is no Agree relation. This is because unlike Japanese and
English, Persian does not show additional wh-effects.
Example (20) demonstrates an island effect in Persian resulting from blocking
of wh-feature movement from chi ‘what’ by ke ‘that’ in the embedded clause.3
This is ill-formed under the interpretation in which the wh-phrase chi ‘what’ has
0
matrix scope.4 Under my analysis, an EPPX feature in the matrix Typ is unable
to attract the wh-feature associated with chi ‘what’ in the embedded clause because
of the intervening complementizer ke ‘that.’
(20) *Parviz az Kimea porsid [ke Maryam chi xarid]?
Parviz of Kimea asked that Maryam what bought
Intended: ‘What did Parviz ask Kimea whether Maryam bought?’ (Simin
Karimi, p.c.)
3
Note that this is not quite identical to a whether/if island effect as ke is a complementizer
that can appear in both statements and interrogative constructions, whereas the English whether
and if cannot be used with statements.
4
This is fine as an echo question if chi ‘what’ has strong stress (Simin Karimi, p.c.).
310
Example (21a) corresponds to (20) above, except that there is an additional whphrase ki ‘who’ outside of the whether/if -clause. Accoring to Simin Karimi (p.c.),
(21a) is well-formed as a wh-construction, but only the matrix wh-phrase is answered, as indicated by the response in (21b). A response such as that in (21c)
is ill-formed because the embedded wh-phrase cannot have matrix scope, which
indicates that the embedded wh-phrase is subject to an island effect.
(21)
(a) Q: Parviz az ki
porsid [ke Maryam chi xarid]?
Parviz of who asked that Maryam what bought
‘Who did Parviz ask whether Maryam bought what?’ (Simin
Karimi, p.c.)
(b) A: Az Kimea porsid.
of Kimea asked
‘He/she asked Kimea’
(c) A: *Be Kimea porsid ke
Maryam book xarid
to
Kimea asked that Maryam book bought
‘He/she asked Kimea if Maryam bought a book.’
Thus, the addition of a wh-phrase outside of an island in Persian does not alleviate
an island effect. I propose that in Persian, once a wh-feature moves to Typ to
0
0
satisfy the EPPX feature, the EPPX feature is still active and it still attracts
other wh-features.
The island effect in (21) can then be accounted for as follows. The wh-feature
associated with the higher wh-phrase ki ‘who’ raises to Typ to satisfy the EPPX
0
0
feature. The EPPX feature remains active so that the wh-feature associated with
the lower wh-phrase chi ‘what’ contained within the whether/if -clause also must
move to the matrix Typ to obtain scope. This wh-feature movement, though, is
blocked by ke ‘that’ in the embedded clause. A diagram of the embedded clause is
shown below in (22).
311
(22)
TypP
Typ’
...
TypEP P X 0
TypP
Typ’
Typ
ke
that
*
TP
Maryam chi+[Fwh ] xarid
Maryam what bought
Evidence from intervention effects similarly shows that all wh-features move in a
multiple wh-construction. Example (23) shows an intervention effect because faghat
‘only’ blocks wh-feature movement from chi ‘what’ to the matrix Typ, thereby
resulting in ill-formedness as a wh-construction. This is fine as a yes/no construction.
(23) *Parviz be Kimea goft [ke faghat Maryam chi xund-e]?
Parviz to Kimea said that only
Maryam what read-has
‘What did Parviz tell Kimea that only Maryam read?’ (Simin Karimi, p.c.)
The addition of a wh-phrase to the matrix clause, as in (24a) below, again does not
allow an intervention effect to be eliminated. The answer in (24b) in which only the
matrix wh-phrase is answered is fine. But the answer in (24c) is ill-formed because
the embedded wh-phrase cannot have matrix scope.
(24)
(a) Q: Parviz be ki
goft [ke faghat Maryam chi xund-e]?
Taro
to who said that only
Maryam what read-has
‘Who did Parviz tell that only Maryam read what?’
(b) A: Be Kimea goft.
to Kimea said
‘He told Kimea.’
312
(c) A: *Be Kimea goft
ke
faghat Arezu book xund-e
to
Kimea asked that only
Arezu book read-has
‘He told Kimea that only Arezu read a book.’ (Simin Karimi, p.c)
Again, movement of the wh-feature associated with the matrix wh-phrase, in this
case ki ‘who,’ to the matrix Typ does not eliminate an EPP feature, and thus
the wh-feature associated with the wh-phrase in the embedded clause is subject to
intervention effects.
The Persian data then show that in a multiple wh-construction, all wh-features
must move to Typ. In a construction such as (19), repeated below, each wh-feature
raises to the matrix Typ, giving its associated wh-phrase scope.
(19) Persian:
Ki chi-ro xarid?
who what
bought-3sg
‘Who bought what?’ (Karimi 2005:145)
This multiple wh-feature movement can be seen in (25).
313
(25)
TypP
Typ’
TP
T’
DP1
ki+tFwh1
who
TypEP P X 0
Prb+Fwh1,Fwh2
vP
t1
T
v’
DP
chi-ro+tFwh2
what
v
xarid
bought-3sg
Evidence for multiple wh-feature movement is supported by the fact that only
a pair-list reading is possible in multiple wh-constructions. For example, Karimi
(2005:146) writes the following about an answer to (19).
The answer to [(19)] is something like: ‘Kimea bought a dress, Rahjue a
hat, and I bought Chomsky’s book on 9-11.’ A single-listing interpretation
is not an option, indicating that all wh-features must move to C.
On this point I agree, all wh-features move to Typ.
In an example such as (25), note that the wh-feature Fwh1 associated with
ki ‘who’ does not block movement of the wh-feature Fwh2 associated with chi-ro
‘what.’ This fact can be accounted for if movement of a wh-feature does not leave a
trace/copy, and thus there is no trace/copy of the higher wh-feature to block movement of the lower wh-feature. Since the higher wh-feature moves to Typ first, and
it does not leave a trace/copy, the lower wh-feature can raise freely to Typ.
314
In summary, in Persian, unlike in Japanese, movement of a wh-feature to Typ
does not ameliorate an EPP feature. An EPP feature remains active even if a
wh-feature has moved to it.
8.7 Languages that disallow multiple wh-constructions
There are languages that completely disallow multiple wh-constructions. In these
languages, multiple wh-phrases cannot have scope in a single clause.5 For example,
multiple wh-constructions do not occur in Scottish Gaelic, as shown below. This is
ill-formed as a wh-construction in which both wh-phrases have scope.
(26) Scottish Gaelic:
*Có a
bha
a’ pógadh có?
who C-rel be-past kissing
who
‘Who kissed who?’ (Adger & Ramchand 2005:183)
Nor do they occur in Irish.
(27) Irish:
(a) *Cé aL
rinne caidé?
who comp did
what
‘Who did what?’
thug sé do cé?
(b) *Caidé aL
comp gave he to who
what
‘What did he give to whom?’
5
Note that multiple wh-phrases may occur in a single clause if a lower wh-phrase receives an
echo question interpretation, but in this case, this is not a multiple wh-construction, as multiple
wh-phrases do not have scope in a single clause.
315
(c) *Cé aL
bhı́ ag caint
le
cé?
who comp was at talking with who
‘Who was talking to who(m)?’ (McCloskey 1979:71)
Similarly, they are banned in Italian.
(28)
(a) *Chi ha scritto che cosa?
Who has written what?’
(b) *Chi è partito quando?
‘Who left when?
(c) *Quale ragazza ha dato un bacio a quale ragazzo?
‘Which girl gave a kiss to which boy?’ (Calabrese 1984, per Dayal 2005)
In languages such as Scottish Gaelic, Irish, and Italian, the fact that multiple
wh-constructions are banned suggests that the probe in Typ is unable to form a
relationship with multiple wh-features. First of all, in these languages, which all
require wh-phrasal movement, the probe in Typ is valued by the wh-feature of a
fronted wh-phrase that moves to satisfy an EPPXP feature. I propose that in these
languages, once the Probe in Typ is valued, it is eliminated. Thus, it is not available
to be valued by any other wh-features.
There are two types of probes that a language may have in Typ, given in (29a-b).
(29)
(a) Probe1 : Remains active once it is valued by a wh-feature.
(b) Probe2 : Eliminated once it is valued by a wh-feature.
Languages such as Sinhala, Mandarin, Japanese, English, and Persian that allow
multiple wh-constructions contain the probe in (29a) which can be valued by multiple wh-features. Languages such as Scottish Gaelic, Irish, and Italian contain the
probe in (29b) which can only be valued once.
Note that all of the languages that disallow multiple wh-constructions are languages with over wh-movement, which I have proposed results from an EPPXP
316
feature in Typ. It is possible that there are other languages that disallow multiple
wh-constructions because the probe in Typ is eliminated once it is initially valued,
but which, unlike the languages discussed here, form wh-constructions via Agree
or wh-feature movement. I suspect that languages of this sort exist, but leave this
issue for further analysis.
8.8 Conclusion
In this chapter. I have examined a few aspects of multiple wh-constructions. I
attempted to show that in languages such as Sinhala and Mandarin, multiple whfeatures can form an Agree relation with Typ. There is no movement of a whelement, and thus, a wh-element is not subject to intervention or island effects even
when there are multiple wh-phrases. In languages such as Japanese and English,
movement of an initial wh-element is driven by an EPP feature in Typ. Once this
movement occurs, the probe in Typ is able to form an Agree relation with another
wh-feature. In languages such as Persian, the EPP feature in Typ remains active
even after an initial wh-feature has moved to it. Therefore, the EPP feature is
able to attract other wh-features. Furthermore, in languages that allow multiple
wh-constructions, I proposed that a probe remains active once it is valued by a whfeature; i.e., a probe can be valued by multiple wh-features. However, in languages
that disallow multiple wh-constructions, such as Scottish Gaelic, Irish, and Italian,
a probe is eliminated once it is valued by a wh-feature.
The table in (30) summarizes the findings regarding wh-feature movement for
multiple wh-constructions in the languages discussed in this chapter.
317
(30)
Languages
(a)
(b)
(c)
Typ
Sinhala, Mandarin
Japanese
English
(d)
Persian
(e)
Scottish Gaelic,
Probe
Fwh ,Fwh
/P
/ P/
E
X0
Fwh ,Fwh
XP
Fwh ,Fwh
X0
EPP
Fwh ,Fwh
EPPXP
*Fwh ,Fwh
/ P/ P/
E
Irish, Italian
In Sinhala and Mandarin multiple wh-constructions (30a), Typ lacks an EPP
feature (or if there is an EPP feature, it does not attract a wh-feature),6 which
results in an Agree relation. The probe Agrees with two wh-features, as shown in
the column labeled Probe. In a Japanese and English multiple wh-construction
(30b) and (30c), respectively, the EPP feature attracts an initial wh-feature
(Japanese) or wh-phrase (English), and then is eliminated, as represented by the
slashes through the EPP feature in the column labeled Typ. A secondary wh-feature
or phrase then can form an Agree relation. In a Persian multiple wh-construction
0
(30d), an EPPX feature is not eliminated, and so every wh-feature must move to
Typ. In all of these languages, multiple wh-features can value a probe in Typ. In
Scottish Gaelic, Irish, and Italian, a probe cannot form a relationship with multiple
wh-features, and thus a multiple wh-construction is banned.
Although no languages of this sort have been discussed here, this analysis also
predicts the existence of languages that do not allow multiple wh-constructions, but
which utilize Agree or wh-feature movement. There could be a language that lacks
an EPP feature in Typ, as in (31a), and thus utilizes Agree, but which does not
allow multiple wh-constructions because a Probe cannot be valued by multiple wh6
Note that in Sinhala, if the arguments in chapters 5.4-5.6 and in the previous section of this
chapter are correct, then when there is a TP-internal Qu-morpheme, there is an EPP feature in
Typ. However, this EPP feature attracts a Qu-feature and not a wh-feature.
318
features. Also, there could be a language that has wh-feature movement resulting
0
from an EPPX feature in Typ, but which does not allow multiple wh-constructions
for the same reason, as shown in (31b).
(31)
Typ
Probe
(a)
(b)
*Fwh ,Fwh
EPPX
0
*Fwh ,Fwh
Further typological investigations would be useful for investigating whether
languages of this sort exist.
319
CHAPTER 9
The weird behavior of why
9.1 Introduction
Cross-linguistic evidence shows that words corresponding to why in various languages have different syntactic distributions from other wh-phrases. (1a-d) state
what some of these differences are.
(1)
(a) Why is not subject to intervention effects in situations in which other
wh-phrases are subject to these effects.
(b) Why is subject to intervention effects even when other wh-phrases are
not subject to these effects.
(c) Why must appear at the clause periphery even though other wh-phrases
can remain in-situ.
(d) Why is subject to island effects even when other wh-phrases are not
subject to these effects.
An example of (1a) can be seen in Japanese. Although a wh-phrase cannot generally
be c-commanded by a scope bearing element, naze ‘why’ can, as shown below in
(2).
(2)
Japanese:
{Taroo-mo/dareka-ga/daremo-ga} naze Boston-o
satta no?
Taroo-also/someone-nom/everyone-nom why Boston-acc left Qu
‘Why did {Taroo also/someone/everyone} leave Boston?’ (Ko 2005:873)
320
An example of (1b) can be seen in Mandarin, a language in which wh-phrases generally can be c-commanded by a scope bearing element without a resulting intervention
effect. However, when weishenme ‘why’ is c-commanded by chang/ye ‘often/also,’
the result is ill-formed, as shown in (3).
(3)
Mandarin:
*Ta {chang/ye} weishenme ma
ta?
he often/also why
scold he
‘What is the reason x such that he often/also scolds/scolded him for x?’
(Soh 2005:146)
An example of (1c) can be seen in Persian. Although wh-phrases can remain insitu, cherâ ‘why’ cannot. (4a) is ill-formed because cherâ ‘why’ is in a TP-internal
position. (4b) is fine because cherâ ‘why’ appears in the clause periphery.
(4)
Persian:
(a) *Ali bâ
Maryam ezdevâj kard
cherâ?
Ali with Maryam marry did.3sg why
(b) Ali cherâ1 bâ
Maryam ezdevâj kard
t1 ?
Ali why
with Maryam marry did-3sg
‘Why did Ali marry Maryam?’ (Kahnemuyipour 2001:47, per Karimi
2005:139)
An example of (1d) can be seen in Japanese. Although wh-phrases generally are not
subject to complex-DP island effects, naze ‘why’ is, as shown in (5).
(5)
Japanese:
*Kimi-wa [DP kare-ga naze kaita hon]-o
yomimashita ka?
you-top
he-nom why wrote book-acc read
Qu
Intended: ‘Why1 did you read [books that he wrote t1 ]?’ (Nishigauchi
1990:41)
321
In this chapter, I discuss the details of this peculiar behavior of why in languages
that fall into each of the categories in (1a-d).
This odd behavior of why has been the subject of much research. Most notably, in
the Government and Binding framework, the behavior of why was often attributed
to the Empty Category Principle (ECP) (cf. Huang (1982) and Lasnik & Saito
(1984), among others). Being an adjunct, why needs to be antecedent governed, as
it cannot be lexically governed. A wh-argument, on the other hand, can be either
lexically or antecedent governed. This requirement that why be antecedent governed
can account for differences between the behavior of why and other wh-phrases. One
problem for the ECP analysis is that it predicts that why and all other adjuncts
should behave in the same manner. But this is not the case. Why often behaves
differently from other wh-adjuncts. Furthermore, the ECP has been abandoned
within the framework of the Minimalist Program.
In this chapter, I examine the behavior of why from the perspective of the Minimalist Program, and attempt to account for the facts described in (1a-d), repeated
below.
(1)
(a) Why is not subject to intervention effects in situations in which other
wh-phrases are subject to these effects.
(b) Why is subject to intervention effects even when other wh-phrases are
not subject to these effects.
(c) Why must appear at the clause periphery even though other wh-phrases
can remain in-situ.
(d) Why is subject to island effects even when other wh-phrases are not
subject to these effects.
I show that the odd behavior of why described in (1a-c) can be accounted for by
adopting a slightly modified version of Ko’s (2005) view that why is base generated
322
in the clause-periphery. I also examine the fact in (1d), but I am unable to offer a
clear solution that accounts for the behavior of why in certain island constructions.
The organization of this chapter is as follows. Sections 9.2 and 9.3 present
evidence that why is base generated in [Spec, TypP] of the clause that it modifies.
In section 9.2, I discuss how the notion that why must appear in [Spec, TypP]
provides a straightforward explanation for the the odd behavior of why with respect
to intervention effects. In section 9.3, I discuss how the notion that why is base
generated in [Spec, TypP] is supported by the fact that in certain languages, why
must appear in the clause periphery in cases in which other wh-phrases can appear
in a TP-internal position. Section 9.4 presents evidence that wh-feature movement
may proceed from why in certain instances. Section 9.5 examines the behavior of
why with respect to island effects. Section 9.6 is the conclusion.
9.2 Why is base generated in [Spec, TypP]
In this section, I discuss Ko’s (2005) proposal, which I adopt with a few minor
changes, that in wh-in-situ languages, why is Merged directly in [Spec, CP]. This
proposal provides a nice explanation for the perplexing facts concerning why given
in (1a-b), repeated below.
(1)
(a) Why is not subject to intervention effects in situations in which other
wh-phrases are subject to these effects.
(b) Why is subject to intervention effects even when other wh-phrases are
not subject to these effects.
Why is not subject to intervention effects in languages such as Japanese, Korean, and
Persian even though other wh-phrases are (1a), whereas why appears to be subject
to intervention effects in languages such as Mandarin in which other wh-phrases are
not (1b).
323
9.2.1 Japanese, Korean, and Persian
In Japanese, Korean, and Persian, wh-phrases are generally subject to intervention
effects, as discussed in chapter 6.3. These intervention effects can be seen in (6ab) (originally (25a-b) in chapter 6.3) from Japanese, (7a-b) (originally (26a-b) in
chapter 6.3) from Korean, and (8a-b) (originally (27a-b) in chapter 6.3) from Persian.
In the (a) examples, a wh-phrase is c-commanded by a scope bearing element,
resulting in an intervention effect. In the (b) examples, the wh-phrase appears
higher than the scope bearing element, and the intervention effect disappears.
(6)
Japanese:
(a) ?*Taroo-sika nani-o
yoma-nai no?
Taroo-only
what-acc read-NEG Qu
‘What did only Taro read?’ (Tanaka 1997:159)
(b) Nani-o Taroo-sika t1 yoma-nai no?
what-acc Taroo-only
read-NEG Qu
‘What did only Taro read?’ (Tanaka 1997:162)
(7)
Korean:
(a) *Amuto muôs-ûl sa-chi
anh-ass-ni?
anyone
what-acc buy-CHI not do-past-Qu
(b) Muôs-ûl1 amuto t1 sa-chi
anh-ass-ni?
what-acc anyone
buy-CHI not do-past-Qu
‘What did no one buy?’ (Beck & Kim 1997:339)
(8)
Persian:
(a) *Hichkas chi-ro
na-xarid?
nobody
what-acc neg-bought
324
(b) Chi-ro1 hichkas t1 na-xarid?
what-acc nobody
neg-bought
‘What was it that no one bought?’ (Karimi & Taleghani 2007:180)
Ko (2005), following work by Miyagawa (1997b, 1999), Cho (1998), Kuwabara
(1998), Watanabe (2000), Lee (2002), and Choi (2003) points out that, surprisingly,
why does not appear to be subject to intervention effects in these languages. For
example, in (9a) (originally presented as (2) above) from Japanese and (9b) from
Korean, why follows a quantificational element, yet this quantificational element
does not cause an intervention effect.
(9)
(a) Japanese:
{Taroo-mo/dareka-ga/daremo-ga} naze Boston-o
satta
Taroo-also/someone-nom/everyone-nom why Boston-acc left
no?
Qu
‘Why did {Taroo also/someone/everyone} leave Boston?’
(b) Korean:
{John-man/John-to/nwukwunka-ka/nwukwuna-ka} way
John-only/John-also/someone-nom/everyone-nom
why
Boston-ul
tenass-ni?
Boston-acc left-Qu
‘Why did {only john/John also/someone/everyone} leave Boston’ (Ko
2005:873)
These examples are also fine if why appears above the quantificational element, as
shown below.
325
(10)
(a) Japanese:
Naze {Taroo-mo/dareka-ga/daremo-ga} Boston-o
satta
why
Taroo-also/someone-nom/everyone-nom Boston-acc left
no?
Qu
‘Why did {Taroo also/someone/everyone} leave Boston?’
(b) Korean:
Way {John-man/John-to/nwukwunka-ka/nwukwuna-ka}
why John-only/John-also/someone-nom/everyone-nom
Boston-ul
tenass-ni?
Boston-acc left-Qu
‘Why did {only john/John also/someone/everyone} leave Boston’ (Ko
2005:873)
Examples (11a-b) also show that why is not subject to intervention effects in Persian. Cherâ ‘why’ can either precede or follow the quantificational element hichkas
‘nobody,’ and the result is fine.
(11)
(a) Hichkas cherâ ketâb-ro
na-xarid?
nobody
why
book-acc NEG-bought
(b) Cherâ1 hichkas t1 ketâb-ro
na-xarid?
why
nobody
book-acc NEG-bought
‘Why didn’t anyone buy the book?’ (Simin Karimi, p.c.)
These facts raise the obvious question of why why is not subject to intervention
effects in these languages, whereas other wh-phrases are subject to these effects.
Ko (2005) provides an answer by arguing that in Korean and Japanese why “is
an adverb which is externally merged (i.e. base-generated) in [Spec,CP] of the clause
it modifies as a CP-modifier (Ko 2005:877).” According to Ko, in (9a-b) there are
no intervention effects because there is no movement of naze ‘why,’ or any element
(such as a feature) associated with it. Rather, why is Merged directly in [Spec, CP]
326
and the clause initial scope bearing element appears in a higher position “either by
A’-movement . . . or by being base-generated above (877)” naze ‘why.’ According to
Ko, in (10a-b), in which why precedes the scope bearing element, why is in [Spec,
CP] and the scope bearing element is below it. This analysis can also be extended
to account for the Persian data in (11a-b) in the same manner; hichkas ‘why’ is base
generated in [Spec, TypP] and thus it can be preceded by a scope bearing element
in a FocP.
Adapting Ko’s (2005) analysis to fit with my proposals, the following picture
emerges. In matrix clauses in Japanese, Korean, and also Persian, why is not subject
to intervention effects because it is base generated in [Spec, TypP], a position above
the base position of why. When why precedes the potential intervener, why is in
[Spec, TypP] and the intervenor is in a lower position within the TypP, as shown in
(12). I refer to the scope bearing potential intervener as a QuantP.
(12)
TypP
AdvP
why[Fwh1 ]
Typ’
TypP
Typ
Prb[Fwh1 ] ,Qu[FQu ]
. . . QuantP
Example (13) shows the structure of a construction in which why is preceded by a
scope bearing element.
327
(13)
FocP
QuantP1[FQuant ]
Foc’
TypP
AdvP
why[Fwh1 ]
Foc
FQuant
Typ’
TP
Typ
Prb[Fwh1 ] ,Qu[FQu ]
. . . t1 . . .
Why is in [Spec, TypP], a position higher than the base position of the QuantP.
The quantifier has scrambled to a position above [Spec, TypP] that I have labeled
as FocP.1
9.2.2 Mandarin
The facts in Mandarin are the exact opposite of those in Japanese, Korean, and
Persian. Although wh-phrases are not generally subject to intervention effects, as
discussed in chapter 6.2, why is an exception. In example (14) (originally (18) in
chapter 6.2), the wh-phrase shenme ‘what’ is c-commanded by the scope bearing
elements zhi ‘only’ or bu ‘not,’ yet there are no intervention effects.
(14) Mandarin:
Ta {zhi/bu} mai shenme?
he only/not sell what
‘What is the thing x such that he {only sells/does not sell} x?’ (Soh
2005:147)
1
In chapter 2.5, I proposed a clause structure in which TypP is above FocP, but it it may be
that a FocP can precede, as well as follow, a TypP in languages that allow a great deal of word
order flexibility, such as Japanese and Korean.
328
Notably, weishenme ‘why,’ unlike other wh-phrases cannot be c-commanded by certain scope bearing elements. When weishenme ‘why’ precedes chang ‘often’ or ye
‘also’ the result is fine, as shown in (15a), but when it follows, the result is ill-formed,
as shown in (15b).2
(15)
(a) Ta weishenme {chang/ye} ma
ta?
he why
often/also scold he
‘What is the reason x such hat he often/also scolds/scolded him for x?’
(b) *Ta {chang/ye} weishenme ma
ta?
he often/also why
scold he
‘What is the reason x such that he often/also scolds/scolded him for x?’
(Soh 2005:146)
Ko (2005) argues that the Mandarin facts are not a result of why being subject
to intervention effects, but rather, they result from Mandarin disallowing ‘scrambling.’ According to Ko, weishenme ‘why’ in Mandarin, just like why in Korean and
Japanese, is Merged in [Spec, CP]. However, whereas in Korean and Japanese, a
scope bearing element can appear in an A’-position above why, there is no position
above [Spec, CP] that can house this type of element. Ko provides evidence that
this is the case. (16a) shows that the definite NP Lisi can be topicalized, and (16b)
shows that it may precede weishenme ‘why.’
(16)
(a) Lisi1 , Zhangsan shuo [(ta1 ) hen congming].
Lisi
Zhangsan said (she) very smart
‘Lisi1 , Zhangsan said that she1 is very smart.
(b) Lisi1 , weishenme t1
kan-le na-ben shu?
Lisi
why
read-asp that-cl book
‘Why did Lisi read that book?’ (Ko 2005:885-886)
However, quantificational elements such as chang ‘often’ and ye ‘also’ cannot be
topicalized, and thus cannot precede weishenme ‘why,’ as in (15b) above. Thus,
there is no position above TypP for these elements to appear in.
2
This example was originally presented as (3) above.
329
The ill-formedness of a construction such as (15b) then is not due to an intervention effect, but rather, according to the analysis presented here, it is due to the
fact that weishenme ‘why’ must appear in [Spec, TypP]. This position cannot be
preceded by a scrambled quantificational element. It can only be preceded by a topicalized element since the periphery of the Mandarin clause-structure is as shown in
(17) below.
(17)
TopP
TypP
AdvP
TP
weishenme
why
There is a TopP projection above TypP, but there is no projection, such as a FocP
above TypP, that a quantificational element can appear in.
9.3 Further evidence that why must appear in [Spec, TypP]
The evidence discussed in the previous sections supports Ko’s (2005) claim that in
certain languages, why is base generated in [Spec, CP] ([Spec, TypP] in my analysis).
In this section, I extend this analysis to account for the behavior of why describe in
(1c), repeated below.
(1)
(c) Why must appear at the clause periphery even though other wh-phrases
can remain in-situ.
In some languages, although a wh-phrase can appear in a TP-internal position, why
cannot. It must occur in the clause periphery. This fact supports the idea that why
is base generated in [Spec, TypP].
330
As discussed in section 6.3, wh-phrases generally may remain in-situ in Persian
or undergo movement; a form of ‘scrambling,’ as shown in (18a-c).
(18)
(a) Kimea diruz
ketâb-ro
be ki
dâd?
Kimea yesterday book-acc to who gave
‘Who did Kimea give the book to yesterday?’
(b) Kimea be ki
diruz
ketâb-ro
dâd?
Kimea to who yesterday book-acc gave
‘Who did Kimea give the book to yesterday?’
(b) Be ki
Kimea diruz
ketâb-ro
dâd?
to who Kimea yesterday book-acc gave
‘Who did Kimea give the book to yesterday?’ (Karimi 2005:136)
Karimi (2005:139) notes, following an observation by Kahnemuyipour (2001),
that cherâ ‘why’ “is subject to obligatory movement from its base position.” For
example, (19a) is a statement with an adjunct clause and (19b) is a corresponding
interrogative construction in which the adjunct clause in (19a) has been replaced
with cherâ ‘why.’ The ill-formedness of this construction shows that cherâ ‘why’
cannot occur in a TP-internal position. Rather, cherâ ‘why’ must appear at the
edge of the clause as in (19c), a position I assume to be [Spec, TypP].3
(19) Persian:
(a) Ali bâ
Maryam ezdevâj kard
[con
dust-esh dâsht].
Ali with Maryam marry did.3sg because friend-her had-3sg
‘Ali married Maryam because he loved her.’
(b) *Ali bâ
Maryam ezdevâj kard
cherâ?
Ali with Maryam marry did.3sg why
(c) Ali cherâ1 bâ
Maryam ezdevâj kard
t1 ?
Ali why
with Maryam marry did-3sg
‘Why did Ali marry Maryam?’ (Kahnemuyipour 2001:47, per Karimi
2005:139)
3
(19b-c) were originally presented as (4a-b).
331
In (19c), if cherâ ‘why’ is in [Spec, TypP], then the preceding subject Ali could be
in a higher position, such as [Spec, FocP].4 In chapter 6.3, I argued that in Persian,
a wh-feature associated with an in-situ wh-phrase raises to Typ. Although Karimi
claims that the clause-peripheral cherâ ‘why’ moves from the base position of the
adjunct clause in (19a), another possibility, consistent with the evidence presented
in the previous section, is that cherâ ‘why’ actually never moves. Rather, it is simply
Merged directly into [Spec, TypP].
As discussed in chapter 6.6, Malay allows wh-phrases to remain in-situ, as in
(20), which was originally presented as (70) in chapter 6.6.
(20) Malay:
Ali memberitahu kamu tadi
[Fatimah baca apa]?
Ali informed
you
just now Fatimah read what
‘What did Ali tell you Fatimah was reading?’ (Cole & Hermon 1998:224)
Yet, according to Cole & Hermon (1998), the non-DP wh-phrase kenapa ‘why’ is an
exception. It cannot occur in a non-clause-peripheral position. When it appears in
clause-initial position in (21a), the result is fine, but when it appears TP-internally
in (21b), ill-formedness results.
(21) Malay:
(a) Kenapa Fatimah menangis?
why
Fatimah cry
‘Why did Fatimah cry?’
(b) *Fatimah kenapa menangis?
Fatimah why
cry
‘Why did Fatimah cry?’ (Cole & Hermon 1998:226)
The wh-phrase kenapa ‘why’ can also appear in a partial wh-movement construction, as shown in (22), originally presened as (71) in chapter 6.6.
4
See Karimi (2005) for discussion of the positions of scrambled phrases in Persian.
332
(22) Malay:
Jon fikir [kenapa1 (yang) Mary rasa [Ali dipecat
t1 ]]
John think why
(that) Mary feel Ali was fired
‘Why does John think (that) Mary felt Ali was fired?’ (Cole & Hermon
1998:225)
Even though kenapa ‘why’ does not appear in its scope position in (22), it is in a
clause-peripheral position that I assume to be [Spec, TypP]. Wh-feature movement
from kenapa ‘why’ thus is permitted as long as kenapa ‘why’ appears in [Spec,
TypP]. Therefore, if kenapa ‘why’ is in [Spec, TypP] of the clause that it modifies,
its wh-feature can move to a higher clause.
In the wh-movement language of English, notably, why also must appear in
a clause-peripheral position when it has scope in a wh-construction, even in situations in which a wh-phrase normally can appear TP-internally. In a single whconstruction, a wh-phrase must move, and therefore, the appearance of why in [Spec,
TypP] in this type of construction is expected. This example can be accounted for
if why moves to [Spec, TypP] from a position corresponding to the adjunct clause
in (23b).
(23)
(a) Why1 did he eat dinner t1 ?
(b) Harry ate dinner [because he was hungry].
However, as discussed in chapter 8, in a multiple wh-construction, only one whphrase moves to [Spec, TypP]. What is interesting is that why cannot appear in a
non-clause-peripheral position (and have scope) in a multiple wh-construction. For
example, whereas the lower wh-phrase what may remain in-situ and have matrix
scope in the multiple wh-construction in (24a), this is not the case in (24b), which
is ill-formed because why does not appear at the beginning of the clause. (24c) is
fine because why appears in clause-initial position.
333
(24)
(a) Who bought what?
(b) *Who e arrived why? (Reinhart 1998:30)
(c) Why did who arrive?
Note that whereas why must appear clause-peripherally, other adverbials, such as
when, as in (25) need not. Although (25) may be a bit marginal under the interpretation in which both who and when have scope, I think that it is much better than
(24b).
(25) ?Who arrived when?
The fact that why must appear in clause-initial position in a multiple whconstruction suggests that unlike with other wh-phrases, why must appear in [Spec,
TypP]. Furthermore, the fact that in (24c), why appears in clause-initial position
even though there is another wh-phrase, is consistent with why being base generated
in this position. If why were base-generated within the TP below the subject who,
then it should not be able to move to clause-initial position without an MLC effect
arising.
In Persian, Malay, and English, the wh-adjunct why thus does not pattern with
other wh-phrases. Although Persian and Malay allow wh-phrases to appear in-situ
in single wh-constructions, and English allows a secondary wh-phrase to remain insitu in multiple wh-constructions, why must appear in the clause periphery. This
fact can be accounted for straightforwardly if why is base generated in [Spec, TypP].
9.4 Feature movement may proceed from why in some languages
Ko (2005) demonstrates that in Japanese and Korean, when why appears in an
embedded clause but has matrix scope, it is subject to intervention effects. Examples
(26a-b) from Japanese and (27a-b) from Korean demonstrate this.
334
(26) Japanese:
(a) *John-wa [Mary-sika naze sono hon-o
yoma-nakat-ta to]
John-top Mary-only why that book-acc read-not-past C
itta no?
said Qu
‘What is the reason x such that John said that only Mary read that
book for x?
(b) John-wa [naze Mary-sika sono hon-o
yoma-nakat-ta to]
John-top why
Mary-only that book-acc read-not-past C
itta no?
said Qu
‘What is the reason x such that John said that only Mary read that
book for x? (Ko 2005:875)
(27) Korean:
(a) *John-un [amwuto way ku
chayk-ul ilk-ci-anh-ass-ta-ko]
John-top anyone
why that book-acc read-Ci-not-past-dec-C
malhayss-ni?
said-Qu
‘What is the reason x such that John said that no one read that book
for x?’
(b) John-un [way amwuto ku
chayk-ul ilk-ci-anh-ass-ta-ko]
John-top why anyone
that book-acc read-Ci-not-past-dec-C
malhayss-ni?
said-Qu
‘What is the reason x such that John said that no one read that book
for x?’ (Ko 2005:875)
In (26-27a), why is contained within an embedded declarative clause, yet it must
have matrix scope because the Qu-morpheme appears in the matrix clause. According to Ko, LF movement of why is blocked by Mary-sika ‘Mary-only’ in Japanese
and amwuto ‘anyone’ in Korean. When why appears above these scope bearing
elements in (26-27b), the intervention effects disappear.
335
Ko (2005) accounts for the intervention effects in constructions such as (26-27a)
as resulting from the blocking of LF movement of why. Specifically, why needs to
move at LF to the matrix interrogative clause to have scope, yet this movement is
blocked by the intervening scope bearing element. In (26-27b), why proceeds freely
to the matrix [Spec, CP] at LF.
Unlike Ko, I propose that the facts concerning embedded clauses with why can
be resolved without resorting to LF movement of why. Rather, there is movement
of a wh-feature from why and this feature movement is motivated by an an EPPX
0
feature (see chapter 6). In constructions such as (26-27a) above, why is Merged
0
in [Spec, TypP] of the embedded clause. Then an EPPX feature contained in the
matrix Typ attracts the wh-feature from why. This wh-feature movement is blocked
by certain c-commanding scope bearing element. A simplified diagram is shown
below, where the inability of the EPP feature to attract the wh-feature is signified
by an arrow that is blocked.
336
(28)
TypP
Typ’
...
TypEP P X 0
Prb
ForceP
FocP
Foc’
...
TypP
Typ’
AdvP
why+t[whF 1]
Foc
TP
FQuant1
*
Typ
...
When there is no intervening scope bearing element, feature movement may proceed
uninhibited, and the result is well-formed, as shown in (29).
337
(29)
TypP
Typ’
...
TypEP P X 0
Prb+Fwh1
ForceP
FocP
Foc’
...
TypP
Typ’
whP1
why+t[Fwh1 ]
Foc
TP
Typ
...
In this manner, why can be subject to intervention effects, but only when wh-feature
movement proceeds from it.
Persian also shows these same effects in an embedded clause; why is subject
to intervention effects. Remember that cherâ ‘why’ is not subject to intervention
effects in a matrix clause (see (11a) above). However, (30) shows that cherâ ‘why’
cannot have matrix scope when it occurs in an embedded clause and is c-commanded
by faghat ‘only.’
338
(30) Persian:
*pro fekr
mi-kon-i
[(ke) faghat cherâ Kimea in xuna-ro
thought dur-do-s2sg that only
why
Kimea this house-acc
dar Tusân sâxte]?
in Tucson built-3sg
‘Why1 do you think [that Kimea only built this house in Tucson t1 ]?’ (Simin
Karimi, p.c.)
The fact that cherâ ‘why’ cannot have matrix scope can be accounted for if whfeature movement is blocked by the intervening faghat ‘only.’ Note that wh-feature
movement may proceed from cherâ ‘why’ in an embedded clause to a matrix clause if
there is no intervener. In (31) below, cherâ ‘why’ modifies the verb of the embedded
clause, but it has matrix scope, as it turns the matrix clause into a wh-construction
(Simin Karimi, p.c.).
(31) Persian:
pro fekr
mi-kon-i
[(ke) Kimea cherâ in xuna-ro
dar
thought dur-do-s2sg that Kimea why
this house-acc in
Tusân sâxte]?
Tucson built-3sg
‘Why1 do you think Kimea [built this house in Tucson t1 ]?’ (Karimi
2005:138-139)
The facts discussed in this section support the notion that why is Merged in
[Spec,TypP] of a clause that it modifies, although, at least in some languages, it
need not be Merged in [Spec, TypP] of the clause in which it has scope. When why
is Merged in a non-scopal TypP, wh-feature movement can proceed from it, and
this wh-feature movement is subject to intervention effects. In this manner, while
adopting Ko’s proposal that why is Merged in the clause-periphery, I account for
the facts without resorting to LF movement.
339
9.5 Why and island effects
In this section I examine the odd behavior of why described in (1d), repeated below.
(1)
(d) Why is subject to island effects even when other wh-phrases are not
subject to these effects.
Japanese and Mandarin shown this phenomenon described in (1d). This perplexing
fact, as discussed below, leaves a number questions unresolved.
As discussed in chapter 7, wh-phrases in Japanese are not generally subject to
complex-DP island and adjunct island effects. (32a-b) (originally (4a) and (5a)
presented in chapter 9.4) show that a wh-phrase can occur within a complex-DP
and within an adjunct clause in Japanese.
(32) Japanese:
(a) Kimi-wa [DP [dare-ga kaita] hon]-o
yomimashita ka?
you-top
who-nom wrote book-acc read
Qu
‘Who1 did you read books that t1 wrote? (Nishigauchi 1990:40)
(b) Taroo-ga [T ypP doko-ni
itta kara]
umaku itta no?
Taro-nom
where-dat went because well
went Qu
‘Where1 did things go well [because Taroo went t1 ?’ (Richards 2000:187)
However, naze ‘why’ does does not pattern with other wh-phrases with respect to
complex-DP and adjunct island effects. When naze ‘why’ occurs within a complexDP, as in (33a) or in an adjunct clause, as in (33b), the result is ill-formed.
(33) Japanese:
(a) *Kimi-wa [DP kare-ga naze kaita hon]-o
yomimashita ka?
you-top
he-nom why wrote book-acc read
Qu
Intended: ‘Why1 did you read [books that he wrote t1 ]?’ (Nishigauchi
1990:41)
340
(b) *Taroo-ga [T ypP naze itta kara]
umaku itta no?
Taro-nom
why went because well
went Qu
Intended: ‘Why1 did things go well [because Taroo went t1 ]?’ (Junko
Ginsburg, p.c.)
In chapter 7, I argued, following Nichigauchi (1986, 1990, 1999a, 1999b), that in
Japanese, a wh-phrase can avoid DP- and adjunct island effects via wh-feature movement; the wh-feature raises to turn the potential island into a wh-phrase and thus
circumvent an island effect. These examples in (33a-b) clearly show that movement
of a wh-feature associated with naze ‘why’ cannot circumvent an island effect.
One possibility, proposed by Nishigauchi (1990) is that a mismatch between the
syntactic category of naze ‘why’ and the ‘island’ that contains it is the cause of
these effects. According to Nishigauchi (1990:89), “a WH-phrase must be identical
in syntactic category with the dominating node in order for the [+WH]-feature to
be percolated to the latter.” In order for a wh-phrase to circumvent a complex-NP
island effect, the wh-phrase must be a nominal. For example, since naze ‘why’ is not
a nominal, its wh-feature is incompatible with a complex-NP. However, a problem
for this analysis is that naze ‘why’ is subject to adjunct island effects, as in (33b),
even though it is an adjunct. There is also the problem of accounting for why
adjuncts other than naze ‘why’ are not subject to island effects, even though these
adjuncts do not appear to be nominals (for example, see (4b) and (5a) in chapter
7.2).
I have taken the position that naze ‘why’ is Merged directly in [Spec, TypP]. As
shown above, when Merged in [Spec, TypP], the wh-feature of naze ‘why’ can move
to a higher clause. Therefore, even if naze ‘why’ were Merged in [Spec, TypP] within
an adjunct clause, the inability of its wh-feature to undergo further movement is
not accounted for. The facts thus indicate that the wh-feature of naze ‘why,’ unlike
the wh-features of other wh-phrases, simply cannot turn a larger phrase into a whphrase; the wh-feature of naze ‘why’ cannot turn a complex-DP or an adjunct clause
341
into a wh-phrase. At this point, I do not have any solution that can account for this
behavior of the wh-feature of naze ‘why.’
I next turn to Mandarin, in which, unlike other wh-phrases in this language,
weishenme ‘why’ is subject to island effects. Examples (34a-b) (originally (19b-c)
in chapter 6.2) show that a wh-phrase can occur within a complex-DP or adjunct
clause in Mandarin.
(34) Mandarin:
(a) Ni xihuan [DP shei xie
de
shu]?
you like
who write gen book
‘Who1 do you like [the book t1 wrote]?’
(b) Ta [T ypP yinwei ni shuo shenme hua] hen shengqi?
he
because you say what
word very angry
‘What1 was he angry [because you said t1 words]?’(Aoun & Li 1993:203)
In chapter 6.2, I argued that these facts result from a probe in Typ being able
to form an Agree relation with the wh-feature of a wh-phrase in these types of
constructions. Yet, examples (35a-b) show that weishenme ‘why’ cannot scope
outside of a complex-DP or adjunct clause, as shown in the following examples.
(35) Mandarin
(a) *Ni xihuan [DP ta weishenme xie
de
shu]?
you like
he why
write gen book
Intended: ‘Why1 do you like [the book he wrote t1 ]?’
(b) *Ta [T ypP yinwei ni shuo weishenme hua] hen shengqi?
he
because you say why
word very angry
Intended: ‘Why1 was he angry [because you said words t1 ]?’(Aoun & Li
1993:203-204)
These facts suggest that a Typ head cannot form an Agree relation with the whfeature of weishenme ‘why’ when weishenme ‘why’ is in a complex-DP or adjunct
clause. Why exactly this is the case is not clear to me.
342
9.6 Conclusion
In conclusion, many of the facts concerning the syntactic behavior of why are accounted for straightforwardly if Ko’s (2005) analysis is adopted; why is base generated in [Spec, TypP] (Ko uses [Spec, CP]). The odd fact that why appears to be
immune to intervention effects in Japanese, Korean, and Persian is accounted for as
resulting from why being base generated in a position above the base position of a
potential intervener. As long as there is no wh-feature movement from why, there
are no intervention effects. In this chapter, I showed how Ko’s proposal that why is
base generated in the clause periphery straightforwardly accounts for the fact that
in languages such as Persian, Malay, and English, why clearly cannot occur in a TPinternal position (and have scope) even though other wh-phrases can. However, as
pointed out by Ko, when why appears in [Spec, TypP] of an embedded clause, it is
subject to intervention effects. I proposed that when why is Merged in [Spec, TypP]
of an embedded clause in languages such as Japanese, Korean, or Persian, and has
matrix scope, its wh-feature raises to the matrix Typ. This wh-feature movement
is subject to intervention effects. Unlike Ko, I account for these intervention effects
as resulting from feature movement, and not from LF movement of a wh-phrase.
The facts regarding island effects and why are perplexing. Whereas wh-phrases
are not generally subject to complex-DP and adjunct island effects in Japanese, why
is subject to these effects. I suggested that this fact could be an indication that
the wh-feature associated with why is unable to move to the relevant position in
Japanese to turn a larger phrase into a wh-phrase and circumvent an island effect.
In Mandarin, the fact that why is subject to island effects appears to result from the
inability of an Agree relation to be formed between Typ and the wh-feature of why
when why is within an island. Why exactly a wh-feature behaves in these ways with
respect to island effects, though, is not clear to me, and requires further analysis.
In conclusion the facts concerning why raise a few perplexing questions, given
343
below.
(36)
(a) Why does why have to be Merged in [Spec, TypP]?
(b) Why can’t wh-feature movement from why circumvent an island effect
in languages such as Japanese?
(c) Why can’t the wh-feature of why form an Agree relation when why is
contained within an island in languages such as Mandarin?
I suspect that investigation of the semantics of why may lead to some answers to
these questions.
344
CHAPTER 10
Conclusion
In this dissertation, I have accounted for cross-linguistic and language internal variation in the formation of yes/no and wh-constructions. In this chapter, I summarize
the main findings of this work concerning the formation of interrogative constructions and I discuss remaining issues.
My analysis began in chapter 3, where I argued that a Qu-feature is required to
type a clause as an interrogative construction, and it does so by appearing in the
head of a clausal typing projection that I refer to as TypP. I discussed how some
languages contain identical overt Qu-morphemes that appear in both yes/no and
wh-constructions, and I claimed that this is evidence that a Qu-feature is required
in both yes/no and wh-constructions. I argued that when an overt Qu-morpheme
is absent in an interrogative construction, a covert Qu-morpheme is present.
In chapter 4, I focused on yes/no constructions. I argued that a certain amount
of cross-linguistic and language internal variation in yes/no constructions is dependent on the properties of a Qu-morpheme; specifically whether or not it is overt, and
whether or not it is an affix. I also discussed the fact that a yes/no construction,
depending on the language, may be formed with a Qu-morpheme in a clause peripheral position, or in a TP-internal position. I argued that by default a Qu-morpheme
is Merged directly in Typ in the clause periphery. However, a Qu-morpheme appears in a TP-internal position when it contains a Focus-feature that forces it to
be Merged into a derivation in a position adjacent to a focused phrase. Evidence
for this proposal is that a TP-internal phrase with an adjacent Qu-morpheme is
interpreted as an identificational focus. When a Qu-morpheme appears in a TPinternal position, its Qu- and Focus-features move to the heads of a TypP and a
345
FocP, respectively, and this movement can be detected via the existence of island
effects.
Next, in chapter 5, I examined the positions of Qu-morphemes in whconstructions. As in a yes/no construction, in a wh-construction, a Qu-morpheme
by default is Merged directly in Typ. I also extended my analysis of TP-internal
Qu-morphemes in yes/no constructions to account for TP-internal Qu-morphemes in
wh-constructions. In some languages, a Qu-morpheme may contain a Focus-feature
which forces it to be Merged in a position adjacent to a wh-phrase in a TP-internal
position, that is, if the wh-phrase is in-situ. In these cases, a wh-phrase appears
to receive a special identificational focus interpretation, although there is a possible exception in Sinhala that requires further examination. I also showed that the
properties of a Qu-morpheme, specifically, whether or not it is an affix, influence
the form of a wh-construction in certain English dialects.
In chapter 6, I examined how a wh-feature establishes a relationship with a probe
in Typ. I argued that depending on the language, 1) a wh-feature can form an Agree
relation with a probe in Typ, 2) a wh-feature can move to Typ, or 3) a wh-phrase
can move to [Spec, TypP]. Which option is chosen is dependent on the Typ head.
When Typ lacks an EPP feature, then there is an Agree relation. When Typ con0
tains an EPPX feature, then there is wh-feature movement, and when there is an
EPPXP feature, there is wh-phrasal movement. In order to distinguish between
Agree and wh-feature movement, I examined the presence and absence of intervention and island effects. I also examined languages with partial wh-movement, which
I argued results when a lower Typ head contains an EPPXP feature and a higher
0
c-commanding Typ head contains an EPPX feature. The lower Typ head with the
EPPXP feature attracts a wh-phrase to its specifier position and the higher Typ
0
head with an EPPX feature attracts a wh-feature.
In chapter 7, I investigated instances in which movement of a wh-element within a
larger non-wh-phrase can turn that phrase into a wh-constitutent. This phenomenon
346
can lead to island circumvention effects. In certain languages, movement of a whfeature or a wh-phrase to the head or specifier, respectively, of a potential island
turns the potential island into a wh-phrase. Then either wh-feature movement can
proceed from the newly formed wh-phrase, or the wh-phrase itself can undergo
movement. I found that island circumvention effects can occur when a wh-phrase
is contained within a complex-DP or an adjunct clauses, although not when it is
contained within a whether/if -island. I proposed that this wh-movement that turns
a larger phrase into a wh-phrase is motivated by an EPP feature, which attracts
the wh-element. I also discussed how some languages do not allow movement of a
wh-element to circumvent island effects. In these languages, the head of a potential
island cannot contain an EPP feature.
In chapter 8, I examined multiple wh-constructions. I argued that in certain
languages such as Japanese and English, a probe in Typ can form an Agree relation
with a secondary wh-feature once movement of an initial wh-element to TypP eliminates an EPP feature. However, in some languages such as Persian, Agree is not an
option for secondary wh-features because movement of a wh-element to TypP does
not eliminate the EPP feature. Also, in some languages which disallow multiple
wh-constructions, a Typ head simply cannot form any relation with a secondary
wh-phrase. I proposed that this was the result of the probe in Typ being eliminated after it is valued by a single wh-feature. In languages which allow multiple
wh-constructions, a probe in Typ can be valued by multiple wh-features.
In chapter 9, I examined the behavior of why, which does not pattern with other
wh-phrases. I argued that why’s odd behavior results from a requirement that it
appear in [Spec, TypP], even in situations in which a wh-phrase generally can remain
in-situ. Thus, the Typ head of a clause containing why must have an EPPXP feature.
Interestingly, there are instances in languages that allow wh-feature movement, such
as Malay, Japanese, Korean, and Persian, in which why can appear in [Spec, TypP]
of a clause in which it does not have scope. In these cases, the wh-feature associated
347
with why undergoes feature movement from why in the specifier of the non-scopal
TypP to the Typ head where it has scope.
In summary, if the arguments presented in this dissertation are on the right track,
then cross-linguistic variation in interrogative constructions boils down to several
important parameters given in (1a-f).
(1)
(a) A Qu-morpheme has either a [+OVERT] or a [-OVERT] feature.
(b) A Qu-morpheme has either a [+AFFIX] or a [-AFFIX] feature.
(c) A Qu-morpheme may or may not have a Focus-feature.
(d) In a wh-constructions, Typ contains an EPPXP feature that attracts a
0
wh-phrase, an EPPX feature that attracts a wh-feature, or no EPP
feature that attracts a wh-elment.
(e) A probe may or may not be eliminated by a wh-feature.
(f) An EPP feature may or may not be eliminated by movement of a
wh-feature.
(1a) accounts for whether or not a Qu-morpheme is overtly pronounced in an interrogative construction. (1b) accounts for whether or not a Qu-morpheme attaches
onto another lexical item. (1c) accounts for whether a Qu-morpheme appears in a
TP-internal position or in a clause-peripheral position. (1d) accounts for whether
there is wh-phrasal movement, wh-feature movement, or no movement at all in
a wh-construction. (1e) accounts for whether or not a language allows multiple
wh-constructions; if a probe is eliminated by a wh-feature (via Agree, wh-feature
movement, or wh-phrasal movement), then multiple wh-constructions do not occur.
(1f) accounts for whether or not a secondary wh-phrase forms an Agree relation
with a probe in Typ (in languages that allow multiple wh-constructions); if an EPP
feature is eliminated and a probe remains active, then a secondary wh-feature forms
an Agree relation with the probe in Typ. I illustrate how these parameter settings
348
account for the wh-construction facts in some of the languages that I have discussed
in this work.
In English the facts are as shown in (2).
(2)
English
Clause
Qu[±OV ERT ]
Qu[±AF F IX]
Matrix
[-OVERT]
[+AFFIX]
Embedded
[+OVERT]
[-AFFIX]
QuFF oc
∅
Typ
EPPXP
Mult-wh:
Mult-wh:
Prb
EPP
Fwh ,Fwh
/ P/ P/ XP
E
In a matrix clause, there is a [-OVERT] and [+AFFIX] Qu-morpheme, whereas
an embedded clause contains a [+OVERT] and [-AFFIX] Qu-morpheme. English
does not allow a Qu-morpheme to contain a Focus-feature. A Typ head (in a
wh-construction) contains an EPPXP feature that attracts a wh-phrase. A probe is
not eliminated by movement of a wh-element; multiple wh-phrases can have scope.
Lastly, movement of a wh-feature eliminates an EPPXP feature from Typ, and thus
secondary wh-features form an Agree relation with a probe in Typ.
The Japanese and Korean facts are shown in (3).
(3)
Japanese/Korean
Qu[±OV ERT ]
[+OVERT]
Qu[±AF F IX]
[-AFFIX]
QuFF oc
∅
Typ
X0
EPP
Mult-wh:Prb
Mult-wh:EPP
Fwh ,Fwh
/ P/ P/ X
E
0
There is a [+OVERT] and [-AFFIX] Qu-morpheme that lacks a Focus-feature.
0
Typ contains an EPPX feature. A probe can be valued by multiple wh-features
0
and movement of an initial wh-feature eliminates the EPPX feature from Typ; i.e.
secondary wh-features form an Agree relation with a probe in Typ.
The Persian facts are summarized in (4).
349
(4)
Persian
Qu[±OV ERT ]
Qu[±AF F IX]
[-OVERT]
QuFF oc
∅
[-AFFIX]
Typ
X0
EPP
Mult-wh:Prb
Mult-wh:EPP
Fwh ,Fwh
EPP
The Persian Qu-morpheme is null and is not an affix.
It also lack a Focus-
feature. Typ contains an EPPX 0 feature. A probe can be valued by multiple
wh-features, and movement of a wh-feature does not eliminate the EPP feature; a
secondary wh-feature must move to Typ .
The Sinhala facts are summarized in (5).
(5)
Sinhala
Qu[±OV ERT ]
Qu[±AF F IX]
[+OVERT]
?
QuFF oc
∅,X
Typ
X0
(EPP
Mult-wh:Prb
)
Fwh ,Fwh
In Sinhala there is a [+OVERT] Qu-morpheme.
I am not sure whether or
not it is an affix. Note that if it is an affix, this does not appear to have an important influence on the structure of interrogative constructions (the Qu-morpheme
does not appear to force movement of a tense element, etc., for the purpose of
affixation). Sinhala has Qu-morphemes both with and without a Focus-feature, as
signified by the ∅ and X marks. The Typ head generally lacks an EPP feature.
It lacks an EPP feature in a wh-construction. However, Typ can contain an EPP
feature that attracts a TP-internal Qu-feature. Since this EPP feature attracts
a feature, it would have to be an EPPX 0 feature. In other cases in which the
Qu-morpheme is base generated in Typ, then there is no EPP feature. Lastly,
multiple wh-features can value a probe in Typ.
The Mandarin facts are summarized in (6).
(6)
Mandarin
Qu[±OV ERT ]
Qu[±AF F IX]
QuFF oc
[+OVERT]/[-OVERT]
[-AFFIX]
∅
Typ
Mult-wh:Prb
Fwh ,Fwh
350
The Mandarin Qu-morpheme can be overt or covert.
It is not an affix, and
it lacks a Focus-feature. The Typ head lacks an EPP feature and a probe in Typ
can be valued by multiple wh-features.
I have also discussed various other languages here, although not in enough detail
to be able to provide the settings of all of the parameters listed above. Notable
though, is that German, Albanian, and Malay require both EPPXP and EPPX
0
features to surface in partial wh-movement constructions. Scottish Gaelic, Irish,
and Italian are languages in which movement of a wh-feature appears to eliminate
a probe in Typ, thereby preventing the occurrence of multiple wh-constructions.
The analyses presented here also raise a number of unresolved issues.
I have argued that a Qu-morpheme can contain a Focus-feature based on evidence that TP-internal elements with an adjacent Qu-morpheme receive an identificational focus interpretation in some languages such as Sinhala, Okinawan, Premodern Japanese, etc. These languages allow a single lexical item to both type a
clause and focus a particular element in a clause. In other languages such as English and Japanese, a Qu-morpheme and focus are not connected in this manner, as
a Qu-morpheme does not contain a Focus-feature. If this analysis is correct, then
some languages have a close connection between a Qu- and a Focus-feature, whereas
others do not. Further examination of interrogative interpretations and their connections, or disconnections, from focus might provide clues as to why languages
differ in this manner.
My analyses of wh-constructions rely on the notion that there is cross-linguistic
variation in EPP features; with some languages having an EPPXP feature in Typ
0
(resulting in wh-phrasal movement), some an EPPX feature (resulting in wh-feature
movement), some allowing both types of EPP feature (resulting in partial whmovement), and some disallowing an EPP feature altogether (resulting in Agree).
What exactly the EPP feature is, and why there are different types of EPP feature,
if there really are different types, is an issue that requires further investigation.
351
In this work, I have taken the position that there is a difference between Agree
and movement triggered by the EPP. Specifically, Agree is a relation that is not
subject to blocking effects (island effects, intervention effects, etc.) because it is
sensitive to feature content. When α forms an Agree relation with β, α searches
specifically for β. Intervening elements that are of the same type as β do not block
Agree. For example if γ and β are quantificational, and γ intervenes between α
and β, γ does not block Agree, because γ has a different content from β. The EPP
feature on the other hand, is sensitive to type. Thus, if α has an EPP feature that
attracts β, but γ intervenes and is of the same type as β, attraction of β is blocked.
However, further investigation of the nature of Agree and the EPP feature may be
needed to clarify why Agree and the EPP differ in these ways.
There are also other unresolved issues concerning wh-constructions. Although
a wh-phrase may remain in-situ in a matrix clause with a null complementizer
in French, in which case there is wh-feature movement, why exactly wh-in-situ is
confined to this one particular environment is not clear. Lastly, evidence indicates
that why must appear in a specifier of a TypP, even in languages that generally
allow a wh-phrase to remain in-situ. Why exactly this is the case requires further
examination.
In conclusion, I hope that further research will shed light on the the proposed
parameters in (1a-f) and on the various other issues brought up in this dissertation.
352
References
Adger, David, & Gillian Ramchand.
2005.
Merge and move: Wh-
dependencies revisited. Linguistic Inquiry 36.161–193.
Agbayani, Brian. 2000. Wh-subjects in English and the vacuous movement
hypothesis. Linguistic Inquiry 31.703–713.
Alexiadou, Artemis, & Elena Anagnostopoulou. 1998. Parameterizing
word order, V-movement, and EPP-checking. Natural Language and Linguistic
Theory 16.491–539.
Anderson, Stephen R. 1993. Wackernagel’s revenge: Clitics, morphology, and
the synax of second position. Language 69.68–98.
Aoun, Joseph, & Yen-hui Audrey Li. 1993. Wh-elements in situ: Syntax or
LF. Linguistic Inquiry 24.199–238.
——, & Dominique Sportiche. 1981. On the formal theory of government.
Linguistic Review 2.211–236.
Baker, C. L. 1970. Notes on the description of English questions: The role of
an abstract question morpheme. Foundations of Language 6.197–219.
Barrett, Rusty. 2008. Is it any way might you could tell me how come am I not
a English speaker?: Subject-auxiliary inversin and introspective methodologies in
linguistics and anthropology. Arizona Linguistics and Anthropology Symposium .
Beck, Sigrid. 1996. Quantified structures as barriers for LF movement. Natural
Language Semantics 4.1–56.
353
——. 2006. Intervention effects follow from focus interpretation. Natural Language
Semantics 14.1–56.
——, & Shin-Sook Kim. 1997. On wh- and operator scope in Korean. Journal
of East Asian Linguistics 6.339–384.
Benzing, Johannes. 1955. Lamutische grammatik . Wiesbaden: Franz Steiner
Verlag.
Bobaljik, Jonathan David. 1994. What does adjacency do? In MIT Working
Papers in Linguistics 22: The morphology-syntax connection, ed. by Heidi Harley
& Collin Phillips, 1–32. MIT Working Papers in Linguistics.
Bos̆ković, Z̆eljko. 1998. LF movement and the Minimalist Program. In NELS
28 , ed. by Pius N. Tamanji & Kiyomi Kusumoto, 43–57. Amherst, MA: University
of Massachusetts, GLSA.
——. 2000. Sometimes in [Spec, CP], sometimes in situ. In Step by step: Essays
on minimalist syntax in honor of Howard Lasnik , ed. by Roger Martin, David
Michaels, & Juan Uriagereka, 53–87. Cambridge, MA: MIT Press.
——. 2002. On multiple wh-fronting. Linguistic Inquiry 33.351–383.
——, & Daiko Takahashi. 1998. Scrambling and last resort. Linguistic Inquiry
29.347–366.
Brandon, Frank Roberts, & Lucy Ferreira Seki. 1984. Moving interrogatives without an initial +WH node in Tupı́. In Syntax and semantics, volume
16: The syntax of Native American Languages, ed. by Eung-do Cook & Donna B.
Gerdts, 77–103. Orlando, Florida: Academic Press, Inc.
Bresnan, Joan, & Sam A. Mchombo. 1987. Topic, pronoun, and agreement
in Chichewa. Language 63.741–782.
354
Brody, Michael. 1995. Lexico-logical form. A radically minimalist theory.
Cambridge, MA: MIT Press.
Bruening, Benjamin. 2007. Wh-in-situ does not correlate with wh-indefinites
or question particles. Linguistic Inquiry 38.139–166.
Cable, Seth, 2007. The grammar of Q: Q-particles and the nature of Whfronting, as revealed by the wh-questions of Tlingit. Massachusetts Institute of
Technology dissertation.
Calabrese, Andrea. 1984. Multiple questions and focus in Italian. In Sentential complementation, ed. by Wim de Geest & Yvan Putseys, 67–74. Foris.
Chang, Lisa. 1997. Whin-situ phenomena in French. MA thesis, University of
British Columbia .
Cheng, Lisa. 2000. Moving just the feature. In Wh-scope marking, ed. by Uli
Lutz, Gereon Müller, & Arnim von Stechow, 77–99. Amsterdam: John Benjamins.
Cheng, Lisa Lai-Shen, 1991. On the typology of wh-questions. MIT dissertation.
——. 1997. On the typology of wh-questions. New York and London: Garland
Publishing, Inc. Published version of Cheng (1991).
——, & Johan Rooryck. 2000. Licensing wh-in-situ. Syntax 3.1–19.
Cho, Eun. 1998. Why, contrastive topic and LF movement. In Japanese/Korean
Linguistics 8 , ed. by D.J. Silva, 403–415. Stanford University, CA: CSLI Publications.
Choi, Young-sik. 2003. Intervention effect in Korean wh-questions revisited .
Paper presented at Glow in Asia 4, Seoul National University, Seoul.
355
Chomsky, Noam. 1967. Remarks on nominalization. In Readings in English
Transformational Grammar , ed. by Roderick Jacobs & Peter Rosenbaum, 184–
221. Waltham, MA: Ginn: MIT Press.
——. 1973. Conditions on transformations. In A festchrift for Morris Halle, ed.
by Stephen P. Anderson & Paul Kiparsky, 232–286. New York: Holt, Rinehart,
& Winston.
——. 1977. On wh movement. In Formal syntax , ed. by Peter W. Culicover,
Thomas Wasow, & Adrian Akmajian, 71–132. New York: Academic Press.
——. 1981. Lectures on Goverment and Binding. Dordrecht: Foris.
——. 1986. Barriers. Cambridge, MA: MIT Press.
——. 1993. A minimalist program for linguistic theory. In The view from building
20: Essays in Linguistics in honor of Sylvain Bromberger , ed. by Ken Hale &
Samuel Jay Keyser, 1–52. Cambridge, MA: MIT Press.
——. 1995a. Bare phrase structure. In Evolution and revolution in linguistic
theory, ed. by Héctor Campos & Paula Kempchinsky, 51–109. Washington, D.C.:
Georgetown University Press.
——. 1995b. The Minimalist Program. Cambridge, MA: MIT Press.
——. 1999. Derivation by phase. MIT Working Papers in Linguistics: MIT
occasional papers in linguistics number 18.
——. 2000. Minimalist inquiries: The framework. In Step by step, ed. by Roger
Martin, David Michaels, & Juan Uriagereka, 89–155. Cambridge, MA: MIT Press.
——. 2001a. Beyond explanatory adequacy. Cambridge, MA: MIT Ocassional
Papers in Linguistics 20.
356
——. 2001b. Derivation by phase. In Ken Hale: a life in language, ed. by Michael
Kenstowicz, 1–52. Cambridge, MA: MIT Press. Originally published as Chomsky
(1999).
——. 2004. Beyond explanatory adequacy. In Structures and beyond: The cartography of syntactic structures, volume 3 , ed. by Adriana Belletti, 104–131. Oxford,
UK: Oxford University Press. Originally published as Chomsky (2001a).
——. 2006. On phases. In Foundational issues in linguistic theory; essays in
honor of Jean-Roger Vergnaud , ed. by Robert Freidin, Carlos Otero, & MariaLuisa Zubizaretta, 133–166. Cambridge, MA: MIT Press.
——, & Howard Lasnik. 1977. Filters and control. Linguistic Inquiry 8.425–
504.
——, & Howard Lasnik. 1993. The theory of principles and parameters. In
Syntax: An international handbook of contemporary research, ed. by Joachim
Jacobs, Arnim von Stechow, Wolfgang Sternefeld, & Theo Vennemann, 506–569.
Berlin: de Gruyter.
——, & Howard Lasnik. 1995. The theory of principles and parameters. In The
Minimalist Program, 13–127. Cambridge, MA: MIT Press. Originally published
as Chomsky & Lasnik (1993).
Cole, Peter, & Gabriella Hermon. 1995. Is wh-in-situ really in-situ: Evidence from Malay and Chinese. In Proceedings of the thirteenth West Coast
Conference on Formal Linguistics, ed. by R. Aranovich et al, 189–204. Stanford
University, Palo Alto, California: Center for the Study of Lanuage and Information.
——, & ——. 1998. The typology of wh-movement: Wh-questions in Malay.
Syntax 1.221–258.
357
Cover, Norbert, & Jairo Nunes (eds.) 2007. The copy theory of movement.
Amsterdam/Philadelphia: John Benjamins Publishing Company.
Dayal, Veneeta. 2005. Multiple wh-questions. In Blackwell Companion to
Syntax , ed. by Martin Everaert, Henk Van Riemsdijk, Rob Goedmans, & Bart
Hollebrandse. Malden, MA: Blackwell.
Denham, Kristin. 2000. Optional wh-movement in Babine-Witsuwit’en. Natural Language and Linguistic Theory 18.199–251.
Enrico, John. 1986. Word order, focus, and topic in Haida. International
Journal of American Linguistics 52.91–123.
——. 2003. Haida syntax . Lincoln, NE: University of Nebraska Press.
Epstein, Samuel David, Erich M. Groat, Ruriko Kawashima, & Hisatsugu Kitahara. 1998. A derivational approach to syntactic relations. Oxford:
Oxford University Press.
Frantz, Donald G. 1973. On question word movement. Linguistic Inquiry
4.531–534.
Gair, James, & Lelwala Sumangala. 1991. What to focus in Sinhala. In
The proceedings of the Eastern States Conference on Linguistics (ESCOL) ’91 ,
93–108. Coumbus, OH: The Ohio State University.
Gair, James W. 1970. Colloquial Sinhalese clause structures. The Hague:
Mouton.
George, Leland, 1980. Analogical generalization in natural language syntax .
Massachussetts Institute of Technology dissertation.
358
Greenberg, Joseph H. 1966. Some universals of grammar with particular
reference to the order of meaningful elements. In Universals of language: second
edition, ed. by Joseph H. Greenberg, 73–113. Cambridge, MA: MIT Press.
Hagstrom, Paul, 1998. Decomposing questions. Massachusetts Institute of
Technology dissertation.
——. 2004. Particle movement in Sinhala and Japanese. In Clause structure
in South Asian languages, ed. by Veneeta Dayal & Anoop Mahajan, 227–252.
Dordrecht: Kluwer.
——. 2006. A-not-a questions. In The Blackwell Companion to Syntax: Volume 1.,
ed. by Martin Everaert & Henk van Riemsdijk, 174–213. Malden, MA: Blackwell.
Hale, Ken. 1983. Papago (k)c. International Journal of American Linguistics
49.299–327.
——. 2002. On the position of the auxiliary in O’odham. Annuario del seminario
de filogia vasca “Julio de Urquijo 44.303–309.
Hale, Kenneth, & Samuel J. Keyser. 1993. On argument structure and
lexical expression of syntactic relations. In The view from building 20: Essays
in linguistics in honor of Sylvain Bromberge, ed. by Kenneth Hale & Samuel J.
Keyser, 53–109. Cambridge, MA: MIT Press.
——, & Elisabeth Selkirk. 1987. Government and tonal phrasing in Papago.
In Phonology Yearbook 4 , ed. by Colin J. Ewen & John M. Anderson, 151–183.
Cambridge, UK: Cambridge University Press.
Halle, Morris, & Alec Marantz. 1993. Distributed morphology and the
pieces f inflection. In The view from building 20: Essays in honoro of Sylvain
Bromberger , ed. by Kenneth Hale & Samuel Jay Keyser, 111–176. Cambridge,
MA: MIT Press.
359
Harley, Heidi. 2004. Merged, conflation, and head movement: The First Sister
Principle revisited. In Proceedings of NELS 34 , ed. by Keir Moulton & Mathew
Wolf, 239–254. BookSurge Publishing.
Henry, Alison. 1992. Infinitives in a for-to dialect. Natural Language and
Linguistic Theory 10.279–301.
Hiraiwa, Ken, 2005. Dimensions of symmetry in syntax: Agreement and clausal
architecture. Massachusetts Institute of Technology dissertation.
Hornstein, Norbert, Howard Lasnik, & Juan Uriagereka. 2007. The
dynamics of islands: Speculations on the locality of movement. Linguistic Analysis
33.149–175.
Horvath, Julia. 1986. Focus in the theory of grammar and the syntax of
Hungarian. Dordrecht: Foris.
Huang, C. -T. James, 1982. Logical relations in Chinese and the theory of
grammar . Massachusetts Institute of Technology dissertation.
Hyde, Viliana. 1971. An introduction to the Luiseño language. Banning, CA:
Malki Museum Press.
Iwasaki, Shoichi, & Preeya Ingkaphirom. 2005. A reference grammar of
Thai . Cambridge, UG: Cambridge University Press.
Jackendoff, Ray. 1974. Introduction to the X̄ convention. Bloomington,
Indiana: Indiana University Linguistics Club.
——. 1977. X̄ Syntax: A study of Phrase Structure. Cambridge, MA: MIT Press.
Jang, Youngjun. 1999. Two types of question and existential quantification.
Linguistics 37.847–869.
360
Johnson, Kyle. 2002. Towards an etiology of adjunct islands. Ms., University
of Massachusetts at Amherst .
Kahnemuyipour, Arsalan. 2001. On wh-questions in Persian. Canadian
Journal of Linguistics 46.41–61.
Kalectaca, Milo. 1978. Lessons in Hopi . Tucson, AZ: University of Arizona
Press.
Karimi, Simin. 2005. A Minimalist approach to scrambling: Evidence from
Persian. Berlin: Mouton de Gruyter.
——, & Azita Taleghani. 2007. Wh-movement, interpretation, and optionality
in Persian. In Phrasal and clausal architecture: Syntactic derivation and interpretation. In honor of Joseph E. Emonds, ed. by Simin Karimi, Vida Samiian, &
Wendy K. Wilkins, 167–187. Amsterdam: John Benjamin Publishing Company.
Katz, Jerrold, & Paul Postal. 1964. An integrated theory of linguistic
descriptions. Cambridge, MA: MIT Press.
Kayne, Richard S. 1991. Romance clitics, verb movement, and PRO. Linguistic Inquiry 22.647–686.
—— 1994. The antisymmetry of syntax . Cambridge, MA: MIT Press.
Kendall, Martha B. 1976. Selected problems in Yavapai syntax: The Verde
Valley dialect. New York: Garland Publishing, Inc.
Keyser, Samuel Jay. 1975. A partial history of the relative clause in English. In University of Massachusetts occasional papers in linguistics 1: Papers in
the history and structure of English, ed. by Jane Grimshaw. University of Massachusetts Amherst, MA: GLSA.
361
Kim, Shin-Sook. 2002. Intervention effects are focus effects. In Japanese/Korean
Linguistics 10 , ed. by Noriko Akatsuka & Susan Strauss, 615–628. Center for the
Study of Language and Information.
——. 2006. More evidence that intervention effects are focus effects. In Proceedings
of the 8th Seoul International Conference on Generative Grammar: Minimalist
views on language design, ed. by Changguk Yim, 161–180. The Korean Generative
Grammar Circle.
Kishimoto, Hideki. 2005. Wh-in-situ and movement in Sinhala questions.
Natural Language and Linguistic Theory 23.1–51.
Kiss, Katalin E. 1998. Identificantional Focus versus Information Focus.
Language 74.245–273.
Kitagawa, Yoshihisa. 2005. Prosody, syntax and pragmatics of wh-questions
in Japanese. English Linguistics 22.2.302–346.
——. 2006. Wh-sope puzzles. In Proceedings of the thirty-fith annual meeting of
the North Eastern Linguistic Society, ed. by Leah Bateman & Cherlon Ussery.
University of Massachusetts, Amherst, MA: Graduate Linguistics Student Association.
——, & Janet Dean Fodor. 2006. Prosodic influence on syntactic judgments.
In Gradience in grammar: Generative perspectives, ed. by Gisbert Fanselow, Caroline Fery, Matthias Schlesewsky, & Ralf Vogel. New York: Oxford University
Press.
——, Dorian Roehrs, & Satoshi Tomioka.
2004.
Multiple wh-
interpretations. In Generative Grammar in a broader perspective: Proceedings
of the 4th GLOW in Asia 2003 , ed. by Hang-Jin Yoon, 209–233. Seoul: Hankook.
362
Klima, Edward. 1964. Negation in English. In The structure of language:
Readings in the philosophy of language, ed. by Jerry A. Fodor & Jerrold J. Katz.
Englewood Cliffs, New Jersey: Prentice-Hall.
Ko, Heejong. 2005. Syntax of why-in-situ: Merge into [Spec,CP] in the overt
syntax. Natrual Language and Linguistic Theory 23.867–916.
Koizumi, Masatoshi, 1995. Phrase structure in Minimalist Syntax . Massachussetts Institute of Technology dissertation.
Koopman, Hilda. 1983. ECP effects in main clauses. Linguistic Inquiry 14.346–
351.
——. 1984. The syntax of verbs. Dordrecht: Foris.
——. 1996. The Spec head configuration. Ms., UCLA.
——, & Dominique Sportiche. 1991. The position of subjects. Lingua 85.211–
258.
Kuwabara, Kazuki. 1998. Overt wh-movement and scope-fixing scrambling:
A preliminary study. In Report (2) researching and verifying an advanced theory
of human language, 115–127. Chiba, Japan: Kanda University of International
Studies.
Langacker, Ronald. 1969. On pronominalization and the chain of command.
In Modern studies in English: Readings in transformational grammar , ed. by
David Reibel & Sanford Schane. Englewood Cliffs, New Jersey: Prentice-Hall.
Larson, Richard K. 1985. On the syntax of disjunction scope. Natural
Language and Linguistic Theory 4.217–264.
—— 1988. On the double object construction. Linguistic Inquiry 19.335–391.
363
Lasnik, Howard. 1976. Remarks on coreference. Linguistic Analysis 2.1–22.
——. 2003. Derivation and representation. In The handbook of contemporary syntactic theory, ed. by Mark Baltin & Chris Collins, 62–88. Malden, MA: Blackwell.
——, & Mamoru Saito. 1984. On the nature of proper govenment. Linguistic
Inquiry 15.235–289.
——, & Mamoru Saito. 1992. Move α: Conditions on its application and
output. Cambridge, MA: MIT Press.
Lee, Mina. 2002. Why is “why” different? . Paper presented at Chicago Linguistic Society 38.
Mahajan, Anoop, 1990. The A/A-bar distinction and movement theory. Massachusetts Institute of Technology dissertation.
Maki, Hideki, 1995. The syntax of particles. University of Connecticut dissertation.
——, & Masao Ochi. 1998. Scrambling of wh-phrases and the Move-F hypothesis. In Japanese/Korean linguistics: Volume 8 , ed. by David J. Silva, 487–500.
Stanford, CA: Center for the Study of Language and Information.
Matsushansky, Ora. 2006. Head movement in linguistics theory. Linguistic
Inquiry 37.69–109.
McCloskey, James. 1979. Transformational syntax and model theoretic semantics. Dordrecht: D. Reidel Publishing Company.
——. 1991. Clause strucure, ellipsis and proper government in Irish. Lingua
85.259–302.
364
McDaniel, Dana. 1989. Partial and multiple wh-movement. Natural Language
and Linguistic Theory 7.565–604.
Mesthrie, Rajend. 1992. English in language shift: The history, structure and
sociolinguistics of South African Indian English. Cambridge, UK: Cambridge
University Press.
Miyagawa, Shigeru. 1997a. Against optional scrambling. Linguistic Inquiry
28.1–26.
——. 1997b. On the nature of wh-scope. Ms., Massachusetts Insitute of Technology
.
——. 1999. Reconstruction and ECP effects. Ms., Massachusetts Institute of
Technology .
——. 2001. EPP, scrambling and wh-in-situ. In Ken Hale: A life in language,
ed. by Michael Kenstowicz, 293–338. Cambridge, MA: MIT Press.
——. 2003. A-movement scrambling and options without optionality. In Word
order and scrambling, ed. by Simin Karimi, 177–200. Malden, MA: Blackwell.
Miyara, Shinsho. 2001. Wh-questions in Okinawan. Linguistic Analysis 30.25–
66.
Moro, Andrea. 2000. Dynamic antisymmetry. Cambridge, MA: MIT Press.
Müller, Gereon. 1997. Partial wh-movement and Optimality Theory. The
Linguistic Review 14.249–306.
——, & Wolfgang Sternefeld. 1996. A-bar chain formation and economy of
derivation. Linguistic Inquiry 27.280–511.
365
Nichols, Lynn. 2003. Attitude evaluation in complex NPs. In Formal approaches to function in grammar: In honor of Eloise Jelinek , ed. by Andrew
Carnie, Heidi Harley, & Mary Ann Willie, 155–164. Amsterdam: John Benjamins.
Nicholson, V. C. 1978. Aspectos da lı́ngua Assurinı́. Brası́lia: Summer Institute of Linguistics.
Nishigauchi, Taisuke, 1986. Quantification in syntax . University of Massachusetts, Amherst dissertation.
——. 1990. Quantification in the theory of grammar . Dordrecht: Kluwer.
——. 1999a. Quantification and wh-constructions. In The handbook of Japanese
linguistics, ed. by Natsuko Tsujimura, 269–296. Malden, MA: Blackwell.
——. 1999b. Ronri-kouzou to bunpou-riron - nichieigo no wh-genshou. Tokyo:
Kurosio.
Nunes, Jairo, & Juan Uriagereka. 2000. Cyclicity and extraction domains.
Syntax 3.20–43.
Omondi, Lucia Ndong’a. 1982. The major syntactic structures of Dholuo.
Berlin: Dietrich Reimer Verlag.
Ortiz de Urbina, Jon. 1989. Parameters in the grammar of Basque. Dordrecht:
Foris.
——. 1990. Operator feature percolation and clausal pied-piping. In MITWPL
13: Papers on wh-movement, ed. by Lisa Cheng & Hamida Demirdash, 193–208.
Cambridge, MA: MITWPL.
——. 2003. Word order. In A grammar of Basque, ed. by José Ignacio Hualde &
Jon Ortiz de Urbina, 448–458. Berlin: Mouton de Gruyter.
366
Payne, Doris L. 1987. Information structure in Papago narrative discourse.
Language 63.783–804.
Pesetsky, David. 1987. Wh-in-situ: Movement and unselective binding. In
The representation of indefiniteness, ed. by E. Reuland & A. ter Meulen, 98–129.
Cambridge, MA: MIT Press.
——. 2000. Phrasal movement and its kin. Cambridge, MA: MIT Press.
——, & Esther Torrego. 2000. T-to-C movement: Causes and consequences.
In Ken Hale: A life in language, ed. by Michael Kenstowicz, 355–426. Cambridge,
MA: MIT Press.
Plann, Susan. 1982. Indirect questions in Spanish. Linguistic Inquiry 13.297–
312.
Reinhart, Tanya, 1976. The syntactic domain of anaphora. Massachusetts
Institute of Technology dissertation.
——. 1993. Wh-in-situ in the framework of the minimalist program. Utrecht: OTS
Working Papers, Research Institute for Language and Speech, Utrecht University.
——. 1995. Interface strategies. Utrecht: OTS Working Papers, Research Institute
for Language and Speech, Utrecht University.
——. 1998. Wh-in-situ in the framework of the Minimalist Program. Natural
Language Semantics 6.29–56.
Richards, Norvin. 1998. The Principle of Minimal Compliance. Linguistic
Inquiry 29.599–629.
——. 2000. An island effect in Japanese. Journal of East Asian Linguistics
9.187–205.
367
——. 2001. Movement in language: Interactions and architectures. New York:
Oxford University Press.
Rivero, Maria-Luisa. 1991. Long head movement and negation: SerboCroation vs. Slovak and Czech. The Linguistic Review 8.319–351.
Rizzi, Luigi. 1980. Violations of the Wh island constraint and the subjacency
condition. Journal of Italian Linguistics 5.157–195.
——. 1990. Relativized minimality. Cambridge, MA: MIT Press.
——. 1994. Argument/adjunct (a)symmetries. In Paths towards universal grammar: Studies in honor of Richard S. Kayne, ed. by Gugliemo Cinque, Jan Koster,
Jean Yves Pollock, Luigi Rizzi, & Raffaela Zanuttini, 361–376. Washington, DC:
Georgetown University Press.
——. 1997. The fine structure of the left periphery. In Elements of grammar , ed.
by Liliane Haegeman, 281–337. Dordrecht: Kluwer.
——. 2001. On the position “Interrogative” in the left periphery of the clause.
In Current studies in Italian syntax: Essays offered to Lorenzo Renzi , ed. by
Guglielmo Cinque & Giampaolo Salvi, 287–296. Amsterdam: Elsevier.
Robbek, V. A. 1989. Iazyk evenov berezovki . Lenningrad: Nauka.
Rochemont, Michael. 1978. A theory of stylistic rules in English. New York:
Garland.
——. 1986. Focus in Generative Grammar . Amsterdam/Philadelphia: John
Benjamins.
Ross, John, 1967. Constraints on variables in syntax . Massachusetts Institute
of Technology dissertation.
368
Sabel, Joachim. 2000. Partial wh-movement and the typology of wh-questions.
In Wh-scope marking, ed. by Uli Lutz, Gereon Muller, & Arnim von Stechow,
409–446. Amsterdam: John Benjamins Publishing Company.
Saddy, Douglas. 1990. Wh-scope mechanisims in Bahasa Indonesia. MIT
Working Papers in Linguistics 15.183–218.
Saito, Mamoru, 1985. Some asymmetries in Japanese and their theoretical
implications. Massachusetts Institute of Technology dissertation.
——. 1994. Additional-wh effects and the adjunction site theory. Journal of East
Asian Linguistics 3.195–240.
——, & Naoki Fukui. 1998. Order in phrase structure and movement. Linguistic
Inquiry 29.439–474.
Scott, Eugene, & Donald G. Frantz. 1974. Sharanahua questions and
proposed constraints on question movement. Linguistics 132.75–86.
Soh, Hooi Ling. 2005. Wh-in-situ in Mandarin Chinese. Linguistic Inquiry
36.143–155.
Sohn, Ho-min. 1999. The Korean Language. Cambridge, UK: Cambridge
University Press.
Sportiche, Dominique. 1981/82. Bounding nodes in French. The Linguistic
Review 1.219–246.
Stein, Gail. 2005. Webster’s new world Spanish grammar handbook . Hoboken,
NJ: Wiley Publishing, Inc.
Stowell, Timothy, 1981. Origins of phrase structure. Massachusetts Institute
of Technology dissertation.
369
Sugahara, Mariko. 1996. Shuri Okinawan kakari-musubi and movement.
In Formal approaches to Japanese linguistics 2, MITWPL 29 , ed. by Masatoshi
Koizumi, Masayuki Oishi, & Uli Sauerland. MIT Working Papers in Linguistics.
Sumangala, Lelwala, 1992. Long distance dependencies in Sinhala: The syntax of focus and WH questions. Cornell University dissertation.
Summer Institute of Linguistics. 1978. Livro de frases Sateré-Mawé Português. Brası́lia: Summer Institute of Linguistics.
Svenonious, Peter. 2004. On the edge. In Peripheries: Syntactic edges and
their effects, ed. by David Adger, Cécile de Cat, & George Tsoulas, 261–287.
Dordrecht: Kluwer.
Szabolcsi, Anna. 1981. The semantics of topic-focus articulation. In Formal
methods in the study of language, ed. by Jan Groenendijk, Theo Janssen, & Martin
Stokhof, 513–541. Amsterdam: Matematisch Centrum.
Tanaka, Hidekazu. 1997. Invisible movement in sika-nai and the linear crossing
constraint. Journal of East Asian Linguistics 6.143–188.
Tiee, Henry Hung-Yeh, & Donald M. Lance. 1986. A reference grammar
of Chinese sentences with exercises. Tucson, AZ: The University of Arizona Press.
Tomioka, Satoshi. 2007. Pragmatics of LF intervention effects: Japanese and
Korean wh-interrogatives. Journal of Pragmatics 39.1570–1590.
Torrego, Esther. 1984. On inversion in Spanish and some of its effects.
Linguistic Inquiry 15.103–129.
Travis, Lisa, 1984. Parameters and effects of word order variation. Massachusetts Institute of Technology dissertation.
370
Tsai, Wei-tien, 1994. On economizing the theory of A-bar dependencies. Massachusetts Institute of Technology dissertation.
Turano, Guiseppina. 1998. Overt and covert dependencies in Albanian. Studia
Linguistica 52.149–183.
Uriagereka, Juan. 1999a. Minimal restrictions on Basque movements. Natural
Language and Linguistic Theory 17.403–444.
——. 1999b. Multiple Spell-Out. In Working Minimalism, ed. by Samuel David
Epstein & Norbert Hornstein, 251–282. MIT Press.
Vovin, A. 2006. Tungusic languages. In Encyclopedia of language and linguistics:
Second edition, ed. by Keith Brown, Anne Anderson, Laurie Bauer, Margie Berns,
Jim Miller, & Graeme Hirst, 143–145. Boston: Elsevier.
Watanabe, Akira. 1992a. Subjacency and S-structure movement of wh-in-situ.
Journal of East Asian Linguistics 1.255–291.
——. 1992b. Wh-in-situ, Subjacency and chain formation. MIT Occasional Papers
in Linguistics 2.
——. 2000. Absorption: Interpretability and feature strength. In Grant-in-Aid
for COE research report (4): Researching and verifying an advanced theory of
human language: Explanation of the human faculty for constructing and computing sentences on the basis of lexical conceptulal features, 253–296. Chiba, Japan:
Kanda University of International Studies.
——. 2002. Loss of overt wh-movement in Old Japanese. In Syntactic effects of
morphological change, ed. by David Lightfoot, 179–195. Oxford: Oxford University Press.
371
Whitman, John. 1997. Kakarimusubi from a comparative perspective. In
Japanese/Korean linguistics 6 , ed. by Ho min Sohn & John Haig, 161–178. Stanford, CA: Center for the Study of Language and Information.
Yoon, Jeong-Me. 2006. Selective island effects revisited: Against the null Q
movement approach. In Proceedings of the 8th Seoul International Conference
on Generative Grammar: Minimalist views on language design, ed. by Changguk
Yim, 387–400. Seoul: The Korean Generative Grammar Circle.
Zepeda, Ofelia. 1983. A Papago grammar . Tucson, AZ: University of Arizona
Press.
Ziervogel, D., & E. J. Mabuza. 1976. A grammar of the Swati language
(siSwati). Pretoria: J. J. Van Schaik Limited.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Related manuals

Download PDF

advertisement