Professional Documents
Culture Documents
The Higher Functional Field: Evidence from The Structure of CP and IP: The Cartography
North Italian Dialects of Syntactic Structures, Volume 2
Cecilia Poletto Edited by Luigi Rizzi
The Syntax of Verb-Initial Languages The Syntax of Anaphora
Edited by Andrew Carnie and Eithne Guilfoyle Ken Safir
Parameters and Universals Principles and Parameters in a VSO Language:
Richard Kayne A Case Study in Welsh
Ian G. Roberts
Portuguese Syntax: New Comparative Studies
Edited by João Costa Structures and Beyond: The Cartography
of Syntactic Structures, Volume 3
XP-Adjunction in Universal Grammar:
Edited by Adriana Belletti
Scrambling and Binding in Hindi-Urdu
Ayesha Kidwai Movement and Silence
Richard S. Kayne
Infinitive Constructions: A Syntactic Analysis
of Romance Languages Restructuring and Functional Heads: The
Guido Mensching Cartography of Syntactic Structures, Volume 4
Guglielmo Cinque
Subject Inversion in Romance and the Theory
of Universal Grammar Scrambling, Remnant Movement and
Edited by Aafke Hulk and Jean-Yves Pollock Restructuring in West Germanic
Roland Hinterhölzl
Subjects, Expletives, and the EPP
Edited by Peter Svenonius The Syntax of Ellipsis: Evidence from Dutch
Dialects
A Unified Theory of Verbal and Nominal
Jeroen van Craenenbroeck
Projections
Yoshiki Ogawa Mapping the Left Periphery: The Cartography
of Syntactic Structures, Volume 5
Functional Structures in DP and IP: The
Edited by Paola Benincá and Nicola Munaro
Cartography of Syntactic Structures, Volume 1
Edited by Guglielmo Cinque Mapping Spatial PPs: The Cartography
of Syntactic Structures, Volume 6
Syntactic Heads and Word Formation
Edited by Guglielmo Cinque and Luigi Rizzi
Marit Julien
The Grammar of Q: Q-Particles, Wh-Movement,
The Syntax of Italian Dialects
and Pied-Piping
Christina Tortora
Seth Cable
The Morphosyntax of Complement-Head
Sequences: Clause Structure and Word Order
Patterns in Kwa
Enoch Oladé Aboh
The Grammar of Q
Q-Particles, Wh-Movement,
and Pied-Piping
Seth Cable
2010
Oxford University Press, Inc., publishes works that further
Oxford University’s objective of excellence
in research, scholarship, and education.
9 8 7 6 5 4 3 2 1
First and foremost, I wish to thank David Katzeek, Anita Lafferty, John Marks, and
Fred White, the Tlingit language consultants for this project. Their generosity,
patience, and energy are truly exceptional, and I thank them for all the time and
help they have provided me in my study of their language. Learning from them has
been a great privilege and a great pleasure, one for which I will always be deeply
grateful.
Most of the Tlingit data presented here were gathered from interviews conducted
at the Sealaska Heritage Institute (SHI) in 2005, 2006, and 2007. Special thanks are
owed to Rosita Worl, Yarrow Vaara, Jordan Lachler, and everyone else at SHI. The
time I have spent at SHI has always been remarkably productive, enjoyable, and
memorable. It is my hope that the information contained here might be of some use
to SHI and its programs, so that the resources and knowledge that I have received
from SHI may be repaid in some small way.
In the course of this work’s development, it has been greatly improved by the
criticisms and corrections of the world’s leading scholars of the Tlingit language. For
their many contributions (particularly their corrections), I wish to thank Jeff Leer,
Richard Dauenhauer, Nora Marks Dauenhauer, Keri Edwards, and James Crippen.
Finally, I wish to thank Roby Littlefield for inquiring about certain Tlingit forms with
Mary Anderson, and Mary Anderson for teaching both Roby and me these forms.
Aatlein gunalchéesh!
For providing me their Japanese judgments, I thank the following persons:
Sachiko Kato, Shigeru Miyagawa, Junri Shimada, and Shoichi Takahashi. For
providing me data on Korean, I thank Dong-Whee Yang. Many thanks are owed to
Hideki Kishimoto for providing important data and information regarding Sinhala.
I also thank Kai von Fintel for providing his German judgments, and Kirill Shklovsky
vii
viii ACKNOWLEDGMENTS
for his Russian judgments. Finally, Jessica Coon is owed special thanks for all the data
she has provided regarding Chol, Tzotzil, and Mayan in general.
This work is a revised and greatly reduced version of my doctoral dissertation,
submitted to MIT in 2007. In addition to many corrections and rewrites, several sig-
nificant portions of the dissertation have been removed. For example, the present
work does not discuss pronominal resumption in Tlingit, nor does it provide an in-
depth review of the semantic puzzles surrounding pied-piping. The removal of this
material was not for its inadequacies, but simply in order to streamline the argumen-
tation. For this reason, I sometimes refer the reader to Cable (2007) for greater ex-
pansion on certain points.
Although I claim authorship of this work, its content has been greatly influenced
by the comments and criticisms of a number of outstanding individuals. First and
foremost among these is David Pesetsky, my principal doctoral advisor, as well as
the other members of my doctoral committee: Danny Fox, Irene Heim, and Norvin
Richards. The following individuals also provided critically important comments
upon earlier versions of this work: Mark Baker, Henry Davis, Keri Edwards, Kai von
Fintel, Paul Hagstrom, Fabian Heck, Sabine Iatridou, Angelika Kratzer, Jeff Leer,
Lisa Matthewson, Shigeru Miyagawa, and Keren Rice. I would also like to briefly
acknowledge here the outstanding work of Fabian Heck (Heck 2004, 2008, 2009),
whose 2004 dissertation provided much of the background to my own, and whose
2008 book (particularly its critical discussion of my thesis) spurred many of the
revisions and improvements herein.
I would like to thank the audiences at the following workshops and conferences
for their comments on earlier versions of this work: NELS 37, WSCLA 11, SULA 5,
and the 2006 ECO5 Student Syntax Workshop. I would also like to thank the audi-
ences and organizers of the colloquia where this work has been presented: University
of Ottawa; University of Massachusetts, Amherst; UCLA; University of British
Columbia; University of Victoria; Cornell University; University of Maryland;
McGill University; and University of Tromsø.
I gratefully acknowledge the financial support of the National Science Foundation.
This book is based on work supported under a National Science Foundation Graduate
Research Fellowship and a National Science Foundation Dissertation Improvement
Grant (BCS-0632431).
The most critical support, however, has come from my wife, Summer. Without
her, none of this could have come into being.
CONTENTS
Abbreviations xi
American Tlingit Orthography xiii
1. Introduction 3
1.1 The Central Claim and the Main Character 3
1.2 Some Classic Assumptions in the Theory of Wh-Questions 4
1.3 Tlingit Wh-Questions Force a New Approach 6
1.4 Two Broader Consequences for Grammatical Theory 9
1.5 The Overarching Research Project, and Further Major
Consequences 11
2. Wh-Fronting and Q-Movement in Tlingit 13
2.1 Introduction 13
2.2 Relevant Background Regarding the Tlingit Language 14
2.3 The Behavior of Wh-Words in Tlingit Wh-Questions 21
2.4 Q-Particles in Tlingit Wh-Questions: The Formal Status of Sá 30
2.5 Wh-Fronting in Tlingit as a Consequence of Q-Movement 36
2.6 The QP-Intervention Condition 43
2.7 A Semantics for Tlingit Wh-Words and Q-Particles 63
3. Applications to Wh-In Situ Languages 84
3.1 Introduction 84
3.2 The Nature of Wh-In Situ Languages 85
3.3 The Semantics of Wh-Indefinites and Wh-Questions in Wh-In Situ
Languages 93
3.4 The Theory of LF/Focus Intervention Effects 96
x CONTENTS
Grammatical Categories
xi
xii ABBREVIATIONS
FA Function Application
LC Lambda Conversion
Institutions
Consonants
xiii
xiv AMERICAN TLINGIT ORTHOGRAPHY
Vowels
ee ìޝ long low-toned high front
ée íޝ long high-toned high front
i ì short low-toned high front
í í short high-toned high front
ei èޝ long low-toned mid front
éi éޝ long high-toned mid front
e è short low-toned mid front
é é short high-toned mid front
aa ܤҒޝ long low-toned low back
áa ޝ́ܤ long high-toned low back
a ܤҒ short low-toned low back
á ́ܤ short high-toned low back
oo uҒޝ long low-toned high back rounded
óo úޝ long high-toned high back rounded
u uҒ short low-toned high back rounded
ú ú short high-toned high back rounded
The Grammar of Q
This page intentionally left blank
1
Introduction
The central claim of this book is that, for a variety of phenomena surrounding
wh-operators, the proper locus of explanation is not those wh-operators them-
selves, but rather a distinct element bearing a special semantic (and sometimes
syntactic) relationship to the wh-operator. In many languages, this distinct ele-
ment is phonologically empty, and for this reason its existence and importance
have not been widely recognized. I follow certain previous authors in referring to
this special element as a ‘Q(uestion)-particle’ (Hagstrom 1998; Kishimoto 2005),
but it should be noted that this label is somewhat misleading. As we will see, this
particle is not restricted simply to questions, and it is a distinct entity from the
interrogative complementizer head, which also has a tradition of being referred to
as ‘Q’.
To help unpack these claims, it will be useful to briefly introduce the ‘main char-
acter’ of this book, the wh-questions of Tlingit (Na-Dene; Alaska, British Columbia,
Yukon). As we will see, it is these structures, illustrated in (1) and schematized in (2),
that provide the most direct evidence for the proposals that follow.
3
4 THE GRAMMAR OF Q
Since at least the 1960s, a fundamental question in the theory of wh-questions has been
“Why do wh-words have to front in the wh-questions of some languages?” Although
there are many specific answers to this question, they all share the following form:
That is, across many different frameworks, theoreticians generally agree that wh-
words front in some languages because the wh-word has a ‘special property’ that
requires it to be located at the position that it fronts to. To illustrate, I outline in (4)
several theories of wh-fronting, each characterized in terms of its particular value for
X, its value for Y, and its particular story regarding why X must be located at Y.
Although it is not often explicitly said, sentences like these directly challenge the
analytic structure in (3). After all, if it is a property of the wh-word that motivates the
fronting, how did this property come to appear on the larger, fronted phrase, a phrase
that does not otherwise inherit the properties of the wh-word? For example, contrasts
like those in (6) show that a possessive DP does not inherit the number properties of
a wh-possessor. How, then, does the DP inherit the ‘wh-property’ that supposedly
triggers the fronting in (5a)?
6 THE GRAMMAR OF Q
There is, of course, a commonly accepted answer, one that allows the hypothesis
in (3) to ‘preserve the phenomenon’ in (5): the structures in (5) all exhibit something
called “pied-piping”. Although details of implementation vary across frameworks,
generally speaking, the term “pied-piping” describes cases where an operation tar-
geting the features of a particular lexical item applies to a phrase properly containing
the maximal projection of that item. This definition is highlighted in (7).2
(7) Pied-piping occurs when an operation that targets the features of a lexical item L
applies to a phrase properly containing LMAX.
We might, then, contrast the analytic term “pied-piping” with the more descriptive
and theory-neutral label “pied-piping structure”, defined in (8).
(8) A pied-piping structure is one where a phrase properly containing the maximal
projection of a wh-word (or related operator) has undergone fronting ‘typically
associated’ with that operator.
Thus, to claim that pied-piping exists is to claim that it is possible for an operation
targeting the features of L to apply to a phrase properly containing the projections
of L.
Of course, what makes such cases possible—what mechanisms serve to derive
pied-piping—is a separate, subsequent question, one that has received much attention
(Ross 1967; Sells 1985; Webelhuth 1992; Kayne 1994; Ginzburg & Sag 2000; Grim-
shaw 2000; Heck 2004, 2008, 2009; Watanabe 2006; Horvath 2007a). Curiously, how-
ever, the more fundamental question of whether pied-piping actually exists has not
yet received serious attention. This is largely due to the ubiquity of the explanatory
structure in (3). After all, if the only analytic option is that the fronting in wh-ques-
tions targets a property of wh-words, then the sentences in (5) clearly show that pied-
piping exists. Indeed, in some introductory discussions of pied-piping, pied-piping is
presented as an observable phenomenon, a datum that must be explained, rather than
as a technical solution to an empirical challenge faced by a particular analysis.
In summary, despite significant disagreement over more specific issues, the lit-
erature on wh-fronting does exhibit a common, classic consensus: (a) the fronting of
wh-words in wh-questions directly results from a property borne by the wh-word,
and (b) wh-questions where there is fronting of a phrase properly containing the
projections of the wh-word reveal the existence of pied-piping.
I will argue that the wh-questions of Tlingit strongly challenge this classic consen-
sus. Specifically, we will see that these structures require a model where wh-fronting
is not directly triggered by any properties of the wh-word itself. Rather, such fronting
INTRODUCTION 7
DP VP QP1 IP
I éesh QP V
Daa sá i éesh t1 al’óon
DP Q al’óon
daa sá
QP-Fronting
In outline, the analysis runs as follows. First, as shown in the structure above, the
Q-particle sá must c-command the wh-word. Moreover, this Q-particle heads its own
projection, labeled ‘QP’. Note that because of the c-command relation between the
Q-particle and the wh-word, this QP projection necessarily contains the wh-word.
Finally, and most importantly, the ‘rule’ for forming wh-questions in Tlingit is that
the QP is fronted, and nothing about the wh-word specifically enters into the rule at
all. Nevertheless, because the QP necessarily contains the wh-word, such obligatory
fronting of the QP has as a secondary consequence the obligatory appearance of the
wh-word in the left periphery as well.
Thus, although it is true that a wh-word must appear in the left periphery of a
Tlingit wh-question—and so the language is, descriptively speaking, a ‘wh-fronting’
language—this word order is not due to an operation directly triggered by the fea-
tures of the wh-word. Rather, the movement targets the features of the Q-particle.
The left-peripheral position of the wh-word simply follows from its being contained
inside the Q-particle’s phrasal projection. (Note that, although the analysis as
sketched in (9) is phrased informally in terms of a ‘rule’ of QP-fronting, the actual
analysis developed in chapter 2 will be formalized within the framework of ‘probes
and goals’, as employed in recent work within the Minimalist Program.)
As we will see, some core evidence in support of this analysis is the fact that the
well-formedness of a Tlingit wh-question depends only upon the locality of the
Q-particle to the left periphery; the locality of the wh-word is irrelevant. This fact is
illustrated by patterns like that in (10).
(10) Wh-Operators May Be Inside Islands Iff Q-Particle Is Outside the Island
a. [ [ Wáa kwligeyi CP] xáat NP] sá i tuwáa sigóo?
how it.is.big.REL fish Q your spirit it.is.glad
How big a fish do you want?
(A fish that is how big do you want?)
8 THE GRAMMAR OF Q
As witnessed in (11), the key property of Tlingit pied-piping structures is the follow-
ing: the particle sá always marks the right edge of whatever has been ‘pied-piped’.
For this reason, we can adopt the following as our analysis of these structures.
DP VP QP1 IP
Aadóo yaagú
INTRODUCTION 9
Under this analysis, the pied-piping structures in (11) are all simply cases where the
Q-particle has as its sister a phrase larger than the maximal projection of the wh-
word. For example, a sentence like (11a) possesses a structure where the Q-particle
sá is sister to the complex DP aadóo yaagú ‘whose boat’, which properly contains
the maximal projection of the wh-word aadóo ‘who’. Consequently, as we see in
(12), the structures in (11) can all be derived by normal phrasal movement of the QP,
exactly as in the case of simple wh-questions like (9).
Crucially, since it is the QP—and not the wh-word—that bears the features trig-
gering ‘wh-fronting’ in Tlingit, we find that the pied-piping structures of Tlingit are
not cases of true pied-piping, as defined in (7). That is, under the analysis in (9), the
structures in (11) are not cases where movement applies to a phrase larger than the
projection of the lexical item that triggers it. Furthermore, since the projections of sá
are never properly contained within the fronted constituent of a Tlingit wh-question,
we find that there simply are not any true cases of pied-piping in Tlingit. For this
reason, the special concept of ‘pied-piping’ can be eliminated without cost from our
theory of Tlingit grammar, thus simplifying the overall theory. By adopting the
analysis in (9), we need not deviate from the null hypothesis that if an operation (in
Tlingit) targets the features of a given lexical item, then it applies only to the maxi-
mal projection of that lexical item.
Besides the advantages that the analysis in (9) brings to the theory of Tlingit grammar,
we will see that it also advances a number of broader issues in the syntax of wh-
constructions. Of course, for this to be the case, the analysis in (9) must not simply
be peculiar to Tlingit, but rather must underlie the structure of wh-questions in many
other languages. Indeed, a central claim of this book is that the Q-based analysis in
(9) actually holds for all wh-fronting languages.
Some initial motivation for this ‘universalist’ position can be found in the fol-
lowing thought experiment. Suppose we were to delete the particle sá from the sur-
face structures of Tlingit. The result would be a language that would not look
significantly different from more widely studied wh-fronting languages. This sug-
gests that we might think of these latter languages as having the Q-based structure in
(9), but lacking phonologically overt Qs.3 Furthermore, in as much as Tlingit wh-
questions require the analysis in (9), general typological and learning-theoretic con-
siderations would suggest we extend the analysis to similar wh-fronting structures in
other languages.
Beyond these general considerations, extending the ‘Q-based’ analysis in (9) to
other wh-fronting languages brings a number of other potentially valuable results.
One of the most immediate is that the concept of ‘pied-piping’ may be eliminated
from the theory of grammar. It will be shown that such elimination comes at rela-
tively little cost, as many of the subtler facts that theories of pied-piping seek to
capture receive interesting analysis under a Q-based approach, where there is no true
pied-piping.
10 THE GRAMMAR OF Q
Importantly, when the Q-particle sá appears with wh-indefinites, there are certain
conditions on where the particle can appear. For example, it cannot appear between
a postposition and its complement; compare (13) and (14).
We will see in chapter 2 that, as one might suspect, the QP does not undergo move-
ment in declarative sentences like (13) and (14). Therefore the impossibility of sen-
tences like (14) must reflect a pure condition on the placement of the Q-particle, and
not any property of the movement relation itself.
On its own, this condition on the placement of Q-particles might be of rather
limited interest. However, combined with the analysis in (9), it has intriguing conse-
quences. Note that, under the analysis in (9), the condition against structures like (14)
alone rules out postposition stranding in Tlingit. As illustrated in (15), postposition
stranding must be derived from a structure where a Q-particle appears between a P
and its DP complement. However, as we have just seen, such structures are ruled out
by independent constraints governing the placement of Q-particles. Consequently,
we find that the ban on postposition stranding in Tlingit is due to a property of the
Q-particle, and not a property of the movement relation itself.
CP
QP1 IP
DP Q
…wh-word… PP
Impossible PP,
QP P Ruled Out by Constraints on
Q-Placement
t1
INTRODUCTION 11
We will see in chapter 2 that similar explanations account for the ill-formedness in
Tlingit of various left-branch extractions, phenomena which in other languages are
also commonly thought to reflect a property of the movement relation.
In its inability to strand adpositions and form left-branch extractions, wh-fronting
in Tlingit is very similar to that in other wh-fronting languages. Given the evidence
that the account in (15) is correct for Tlingit, it follows that we should pursue such an
account for those phenomena in all other wh-fronting languages as well. We find,
then, that the nature of postposition stranding and left-branch extractions in Tlingit
provides additional support for extending the analysis in (9) to all wh-fronting lan-
guages. Moreover, under this approach, we see that phenomena commonly attributed
to constraints on movement are ultimately due to more basic conditions on the place-
ment of Q-particles.4
The two results described above illustrate and motivate the broader research project
undertaken in this book. As mentioned, the perspective of this project is that the formal
element pronounced as ‘sá’ in Tlingit is a structural component of the wh-questions of
all human languages, including all wh-fronting languages. Because of its phonologi-
cal invisibility in the best-studied wh-fronting languages, the important role played by
the Q-particle in a variety of phenomena has not been recognized. Consequently, these
phenomena have in previous treatments been somewhat misanalyzed, often in terms
of the movement relation between the wh-word and the left periphery. However, the
overt appearance of the Q-particle in Tlingit wh-questions and wh-indefinites yields
an invaluable empirical tool, which can factor out three possible sources of explana-
tion. These are listed in (16).
As these results indicate, a general message of this book is that the introduction of the
‘QP’ projection in (9) brings with it a new and highly versatile analytic tool, one that
allows many classic puzzles to be reconceived and approached in novel ways.
Finally, the work detailed in this book might also offer yet another object lesson
in the importance of endangered and understudied languages in the development of
linguistic theory. Often, when linguists are asked to explain the importance of
research into such languages, it is noted that their study can provide crucial evidence
to decide between competing analyses. In this way, careful documentation of these
languages is seen to ‘broaden the empirical database’ that theories of language must
cover. However, it is often overlooked in these discussions that the study of such
languages can also serve to introduce new analyses, ones that may offer entirely new
approaches and perspectives to older, seemingly settled issues (cf. Matthewson 1998;
Rullmann et al. 2008). That is, rather than shrink the space of potential analyses,
careful study of these languages can reveal that the current hypothesis space is too
narrow and fails to include hypotheses that would have otherwise never been imag-
ined for more well-studied languages.
2
2.1 Introduction
In this chapter I argue that the ‘Q-based’ analysis from chapter 1, repeated below,
provides the best account of several features of the wh-questions of Tlingit.
DP VP QP1 IP
I éesh QP V
Daa sá i éesh t1 al’óon
DP Q al’óon
daa sá
QP-Fronting
I begin in the following section by providing the reader with some very brief
background regarding the Tlingit language, as well as the methodology employed by
the present study. Although comparatively long for a section of its type, this section
greatly condenses the expansive background chapter of Cable (2007), to which the
reader is referred for further information.
Following these preliminaries I undertake my defense of the analysis in (1)
for the wh-questions of Tlingit. I begin in section 2.3 by arguing that Tlingit is a
13
14 THE GRAMMAR OF Q
‘wh-fronting language’, in the sense that wh-words in its wh-questions must ap-
pear at left peripheral positions. Following this, I argue in section 2.4 that the
Tlingit particle sá should be categorized as a ‘Q-particle’ alongside the Japanese
particle ka and the Sinhala particle da.1 The importance of this categorization
is clarified in section 2.5, where the analysis in (1) is contextualized among cer-
tain recent proposals regarding the syntax of Japanese and Sinhala wh-questions
(Hagstrom 1998; Kishimoto 2005). Section 2.5 also introduces a more formal
implementation of the analysis in (1), couched within the framework of ‘probes’
and ‘goals’, as developed within the Minimalist Program (Chomsky 2000). Finally,
section 2.5 presents the core arguments supporting (1) over a more traditional
analysis where Tlingit wh-words (rather than QPs) are the ‘targets’ for movement.
These arguments show that the interrogative C head of a Tlingit wh-question
probes and Agrees with only the (projection of the) Q-particle sá, and not with the
wh-word itself.
Having presented the core arguments for the analysis in (1), I then discuss in
section 2.6 some further Tlingit-internal applications of the analysis. I argue that
several additional conditions in Tlingit governing the placement of sá would follow
from a single generalization, the ‘QP-Intervention Condition’, but only under the
assumption that both sá and its sister are contained within a QP projection. Further-
more, I show that, under the analysis in (1), these independent conditions on the
placement of sá would alone account for several additional constraints on the form of
Tlingit wh-questions. Interestingly, in many other languages, these ‘additional con-
straints’ are thought to result from conditions on movement. This fact foreshadows
later arguments that the analysis in (1) is not unique to Tlingit, but holds for all wh-
fronting languages.
Finally, in section 2.7 I provide a semantics for Tlingit wh-questions and wh-
indefinites. This semantics, based on earlier semantic research into wh-words and
Q-particles (Hagstrom 1998; Beck 2006), is shown to correctly interpret the perhaps
exotic-looking syntactic structures that the analysis in (1) employs. Interestingly, this
semantics also captures certain properties of Q-particles crucial to the argumentation
from section 2.4. Most interesting of all, however, is that the system can be shown to
assign the correct interpretation to the pied-piping structures of Tlingit, without ap-
pealing to any special machinery beyond that used for plain wh-questions without
pied-piping.
In this section I provide the reader with some elementary background regarding the
Tlingit language. Section 2.2.1 covers key sociohistorical features of the language,
including its geographic distribution, genetic classification, and current vitality. Sec-
tion 2.2.2 provides the reader with a brief summary of prior scholarship on this rather
understudied language, and section 2.2.3 introduces the reader to certain of its major
grammatical features. Finally, section 2.2.4 describes the methodology employed in
this study.
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 15
Dene-Yeniseian
Yeniseian Na-Dene
Ket
Yugh Tlingit Eyak-Athabaskan
Pumpokol Tlingit
Eyak Athabaskan
Eyak Navajo
Apache
Slave
Dogrib, etc.
speakers has not been done, though scholars privately estimate the number as now
being at most between 300 and 400 (James Crippen, personal communication).
The youngest native speakers of Tlingit are in their 50s, and there is no known
native speaker of the language under the age of 40 (Dauenhauer & Dauenhauer
1987). Consequently, Tlingit is considered a highly endangered language and has
been categorized by some as ‘moribund’ (Leer 1991).
It should be noted, however, that there is extensive ongoing work aimed at docu-
menting, maintaining, and revitalizing the Tlingit language (Sealaska Heritage Insti-
tute 2003). The Tlingit community generally has a positive attitude toward their
language, and community interest in the language continues to grow (Gordon 2005).
Courses in Tlingit are presently taught in Alaskan public schools and at the Univer-
sity of Alaska, Southeast, and a number of successful immersion camps have been
held (Sealaska Heritage Institute 2003).
Thanks in part to this activity, some younger adults have acquired a significant
degree of fluency and have made efforts to reintroduce the language into family and
public life. Particularly strong L2 communities appear to be emerging in areas
throughout Southeast Alaska: Sitka, Klukwan, Hoonah, Juneau, and Wrangell.
The ever-growing visibility of the language is illustrated by the recent translation
of Macbeth into Tlingit, which was performed at the Smithsonian’s National
Museum of the American Indian (Quinn 2007). All of this helps to dispel absolute
pessimism concerning the survival of the language and builds a certain amount of
guarded optimism amongst specialists.
Following their study of Tlingit grammar, Naish and Story collaborated on both
a verb dictionary (Story & Naish 1973) and a noun dictionary (Story et al. 1976).
Besides its value as a dictionary, Story and Naish (1973) contains a grammatical
sketch and numerous example sentences. A later and more extensive noun dictionary
for Tlingit is Leer et al. (2001). Story and Naish (1973) remains the only verb dictio-
nary for the language, although a more ‘learner-friendly’ verb dictionary is in devel-
opment (Edwards, forthcoming).
The only other purely grammatical study of the Tlingit language is the PhD disser-
tation of Jeff Leer (Leer 1991), whose treatment of Tlingit verbal morphology and pho-
nology improves immeasurably upon the work of Story (1966). Although it clarifies a
great many issues in the grammar of Tlingit, Leer (1991) focuses primarily on verbal
morphology and phonology, and only 20 pages in 500 are given to syntactic descrip-
tion. It should be noted, however, that those 20 pages contain important, novel insights
regarding the structure of the language’s noun phrase and clausal architecture.
Since the 1960s, a sizeable amount of Tlingit textual material has been collected
and published. Worthy of special mention are the four volumes of texts edited by
Nora and Dick Dauenhauer (Dauenhauer & Dauenhauer 1987, 1990, 1994; Black,
Dauenhauer, & Dauenhauer 2008). Williams, Williams, and Leer (1978) and Nyman
& Leer (1993) are also deserving of special mention.
Educational materials are available for the study of Tlingit as a second language.
Besides the aforementioned dictionaries, there is an introductory textbook with audio
CDs (Dauenhauer & Dauenhauer 2000), a phrase book (Dauenhauer & Dauenhauer
2002), and a phrase book with accompanying CDs (Edwards et al. 2005). An inter-
mediate textbook is presently under development (Dauenhauer & Dauenhauer, forth-
coming), as well as a more comprehensive verb dictionary (Edwards, forthcoming).
All the aforementioned educational materials (except for those yet to be published)
are either published or distributed by the Sealaska Heritage Institute (SHI).
2.2.3.1 Phonology
Throughout this book I represent the sounds of Tlingit using the ‘American orthog-
raphy’ for the language (Dauenhauer & Dauenhauer 2000). This orthography is used
in the Alaskan school system and in all publications by SHI. A different, ‘Canadian
orthography’ is used by Tlingit living in Canada and by the Yukon Native Language
Center (Nyman & Leer 1993). The front material of this book provides a chart match-
ing the characters of the American Tlingit orthography to their equivalents in IPA.
This chart also summarizes the phonemic inventory of the language.
Although the phonemic inventory and phonological alternations of the Tlingit
language are of much interest, they will not be of direct relevance to the syntactic
study that follows. Interested readers are referred to Leer (1991) and Cable (2006b).
18 THE GRAMMAR OF Q
2.2.3.2 Morphology
Like nearly all languages of North America, Tlingit may be described as a ‘head-
marking’ language (Nichols 1986). Consequently, nominal morphology in Tlingit is
comparatively poor—consisting primarily of possessive marking and optional plural
marking—while verbal morphology is quite rich and contributes a wide variety of
information about the event described.
The verbal morphology of Tlingit is strikingly similar in form to that of its
Athabaskan relatives. That is, the verbal morphology is almost exclusively prefixal.
Moreover, the order of the prefixes in a Tlingit verb is not easily derivable from general
principles, and so it is standardly described by use of a stipulative ‘morphological tem-
plate’. For descriptions and analyses of the ‘templatic verbal morphology’ of Tlingit,
see Story (1966), Leer (1991), Dauenhauer & Dauenhauer (2000), and Cable (2006b).
The prefixal template of a Tlingit verb consists of approximately 16 different
positions. Consequently, a particular surface verbal form of Tlingit may underlyingly
contain a sizeable number of prefixes. As in its Athabaskan relatives, these underly-
ing prefixes are greatly affected by various phonological processes, which serve to
drastically reduce the pronounced surface form of the verb.
These points are illustrated by the verbal forms in (3). Note that, although the
underlying forms of the verbal prefix strings differ in only one syllable, the surface
forms differ remarkably in their appearance.
Given the complexities of Tlingit verbal morphology, and the fact that we are
only concerned with its phrasal syntax, I will only provide the roughest of glosses for
the Tlingit verbs exemplified throughout. I will not provide a full morphological
breakdown of every verbal form, but will instead gloss only the ‘propositional con-
tent’ of a given verb, as illustrated in (4).
Note that these ‘propositional glosses’ are merely a notational convenience, and do
not represent any serious proposals regarding the morphosyntax of Tlingit. Thus,
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 19
although these glosses contain English pronouns, I do not seriously adopt the ‘Pro-
nominal Argument Hypothesis’ (Jelinek 1984) for Tlingit. Rather, I assume that full
DPs in Tlingit can function as verbal arguments, and are not necessarily mere clausal
adjuncts.
2.2.3.3 Syntax
As mentioned previously, the only works providing focused discussion of Tlingit
syntax are Naish (1966) and Leer (1991). Although many aspects of Tlingit syntax
have yet to be studied, certain general features are well documented. Since many of
the more specific details of Tlingit syntax will be introduced as they become relevant,
I discuss here only the most basic facts about its gross syntactic structure.
In its syntax, the Tlingit language is not very different from its Athabaskan rela-
tives. Like the other Na-Dene languages, Tlingit is a head-marking language with
extensive null anaphora. Also like its Athabaskan relatives, Tlingit largely displays a
head-final alignment: the language employs postpositions, and no prepositions; pos-
sessors and other nominal complements precede the head noun; demonstratives, ad-
jectives, relative clauses, and other nominal modifiers precede the head noun;
auxiliary verbs follow main verbs. In addition to this, the most frequent word order
in Tlingit texts is typically OV (Dryer 1985).
Unlike its Athabaskan relatives, however, Tlingit has a rather free word
order, and freely permits the positioning of major constituents after the verb.3
Generally speaking, any permutation of S, V, O is an allowable sentence of
Tlingit, although there are of course discourse-structural effects associated
with particular orders (see Leer 1991, chapter 2). This freedom of word order is
illustrated in (5).4
These interview sessions included both group sessions and sessions with indi-
vidual speakers. In general, at a given time I worked with all the people who were
available.
With the permission of the speakers, all interview sessions were recorded on a
Tascam digital 8-track recorder. I later made transcriptions based on these record-
ings. For each session, each speaker in that session was later given a copy of the re-
cording and of the transcription. Furthermore, copies of every recording and
transcription were given to SHI for inclusion in their Tlingit language archives.
Having provided a basic introduction to the Tlingit language, I will now seek to
establish the principle claim of this chapter, that the structure in (1) is the correct
analysis of wh-questions in Tlingit.
I begin, in this section, by arguing that Tlingit can be described as a ‘wh-fronting
language’, in that the wh-words of a Tlingit wh-question obligatorily appear within
the left periphery of the clause. Given the paucity of descriptive work on Tlingit
22 THE GRAMMAR OF Q
syntax, this is an original claim regarding the structure of the language’s wh-ques-
tions, and so I take care to defend it at length. I begin in section 2.3.1 by introducing
various basic properties of Tlingit wh-questions. Then, in section 2.3.2, I outline a
variety of word order constraints found operable in Tlingit wh-questions, and argue
that they are best explained by the generalization that wh-words in such questions
must be located in the left periphery of the clause.
remaining material of the sentence typically follows the wh-word, with a strong ten-
dency to follow the verb.
Because of the freedom of word order in Tlingit, it is not obvious upon casual
examination whether the language requires wh-words to occupy a left-peripheral
position in wh-questions. Indeed, this issue has not yet been addressed in the
published grammatical descriptions of Tlingit. Nevertheless, certain facts indicate
that such wh-words are left peripheral in Tlingit wh-questions.10
As the table shows, all the wh-questions in the selected corpus containing an overt
predicate place the wh-operator before the predicate.
This pattern is also confirmed by the grammaticality judgments offered by native
speakers. As the following data show, speakers reject as ill-formed any wh-question
where the wh-operator follows the main predicate. Such sentences are consistently
corrected by speakers to ones in which the wh-operator precedes the predicate.
Of course, one might justifiably wonder whether the ill-formedness of the starred
sentences above is due not to a rule of obligatory wh-fronting, but to independent
semantic conditions on postpredicate NPs. Perhaps postpredicate NPs must possess
qualities that wh-words inherently lack, such as definiteness? Recall, however, that
wh-words in Tlingit can function as indefinites in declarative clauses (chapter 1).
When a wh-word is used as an indefinite, it is not required to appear before the pred-
icate of the clause. This fact is clearly indicated both by textual examination and by
the well-formedness judgments of native speakers. The table in (16) demonstrates
that the selected corpus of texts supports this grammatical generalization.
(16) Wh-Indefinites May Freely Follow the Main Predicate of the Clause
Sentences containing Of such sentences, those in
wh-indefinite and an which wh-indefinite precedes
Text overt predicate the predicate
As the table shows, not all wh-indefinites in the selected corpus precede the main
predicate of their clause. The following two sentences illustrate these textually attested
cases of postpredicative wh-indefinites.
In this table, the left-most column indicates the number of wh-questions in the corpus
containing some major constituent besides the wh-word and the predicate. The
middle column reports how many, from the questions represented in the left column,
place the wh-word initially in the clause. Finally, the right column indicates the
number of questions in the middle column in which the initial position of the wh-
word does not follow from more general word-order frequencies in Tlingit, such as
the fact that subjects tend to precede objects in the language (Dryer 1985). The totals
at the bottom indicate an overwhelming preference for wh-questions to begin with
wh-words.
Consonant with their textual rarity, sentences like (11c) are occasionally judged
by speakers to be marginal or ill-formed, a classification that is sometimes revised
upon further reflection. The textual rarity of sentences like (11c) would, of course,
follow from their possessing special discourse properties, ones that place strong
limits on the kind of context in which such structures might be embedded. Such spe-
cial discourse properties would also account for their occasional rejection by speakers,
rejection occurring when the licensing context is difficult for the speaker to imagine
or strikes them as far-fetched.
It seems likely, then, that sentences like (11c) possess some special discourse-
structural property. That this property is the ‘topichood’ of the material preceding the
wh-word comports well with a number of other facts. First, in all the naturally occur-
ring instances of noninitial wh-operators I have encountered, the material preceding
the wh-operator is a referential expression. The following two examples illustrate the
general pattern.
Note that this pattern is also evident in sentences (11c), (12b), (13b), and (14b).14
Indeed, speakers do not allow fully nonreferential material to precede the wh-operator
of a wh-question.
These data indicate that only referential XPs may precede the wh-operator of a
wh-question. Of course, one of the core properties of ‘topics’ is that they can only be
28 THE GRAMMAR OF Q
denoted by referential expressions (Li 1976), and so these data support the notion
that any material preceding the wh-operator of a Tlingit wh-question must be
construed as a discourse topic.
A final suggestive piece of evidence is the translations offered by speakers for
sentences like (11c). When these sentences are accepted by native speakers, they are
regularly translated into English using hanging topic left dislocation structures, such
as (22).
That speakers use English left dislocation to translate these sentences supports their
having a special discourse structure that is not possessed by a simple wh-question
and that only left dislocation in English is able to simulate.15
There is, then, reason to conclude that any material preceding the wh-operator of
a Tlingit wh-question must be interpreted as a discourse topic. This fact itself would
most naturally follow from a syntax in which wh-operators are fronted into the left
periphery of Tlingit wh-questions. Under such a syntax, any material occurring to the
left of a Tlingit wh-operator would either have to occupy a left-peripheral topic posi-
tion (Rizzi 1997), or else would have to simply be a dislocated, hanging topic.
main verb. In such sentences, the interrogative word must appear to the left of the
main verb, and cannot appear downstairs in its base position.
b. * Wáa sá aa sá kuyawsikaa?
how Q who Q they.said.to.someone
I conclude from the grammatical patterns described in the previous section that the
wh-operator of a Tlingit wh-question must occupy a left peripheral position within
the clause. In this section I argue that the Tlingit particle sá is most plausibly catego-
rized as a Q-particle.
This argument, however, must be rather indirect, as there are no widely accepted
diagnostics for applying the term ‘Q-particle’. I will therefore argue that sá is a
Q-particle on the basis of its similarity to the particle da in Sinhala and the particle
ka in Japanese. Given that da and ka are uncontroversial instances of Q-particles, the
overwhelming parallels between sá, da, and ka will demand that sá receive the same
categorization.
In section 2.4.1 I demonstrate that sá, da, and ka share the property that they must
appear both with wh-operators in wh-questions and wh-indefinites in declarative
clauses. In section 2.4.2 I show that these particles all must c-command the wh-word
they are paired with. In section 2.4.3 I show that they all have the property that they
cannot be separated from the edge of the clause by a syntactic island. Finally, in sec-
tion 2.4.4 I show that sá and da share the property that they cannot appear at the end
of matrix clauses, although they can appear clause-finally in subordinate clauses.
The data in (31) demonstrate that sá is required not only by the interrogative
force of the clause, but by the wh-word itself. Although this may seem to undercut
the force of the label “question particle”, this property also holds for such prototypi-
cal ‘Q-particles’ as Japanese ka and Sinhala da.18, 19
The data in (32) and (33) led Hagstrom (1998) to develop a semantics for
Q-particles under which they are expected to appear both within wh-questions and
with wh-words interpreted as indefinites in declarative clauses. We will see in section
2.7 that this semantic analysis may, with minor modification, be extended to the
Tlingit particle sá, and would similarly predict its parallel grammatical behavior.
Such a shared semantics would constitute one reason to apply the label “Q-particle”
to Tlingit sá.21
a. [ Goodéi ] sá kkwagóot?
where.to Q I.will.go
Where will I go to?
32 THE GRAMMAR OF Q
Upon examination of just the sentences in (34), one might form the simple hy-
pothesis that the particle sá can be freely placed anywhere to the right of the interrog-
ative word. Although this would be the simplest conclusion, the ill-formedness of
sentences (35b) and (36b) demonstrates that it cannot be correct. Rather, the correct
generalization is that sá must appear either directly to the right of the wh-word, or
directly to the right of a phrase containing the wh-word. In other words, the particle
sá has to c-command the wh-word.
The condition that the Q-particle c-command the wh-word also holds for Sin-
hala da and Japanese ka. This c-command condition is stated explicitly by Kishimo-
to (2005: 13) for Sinhala da and by Yatsushiro (2001: 182) for Japanese ka. The
sentences in (37) illustrate structures where the Sinhala particle da appears detached
from its associated wh-word.23
Such identity of distribution further emphasizes the formal similarity between sá, da,
and ka. Moreover, it will be shown in section 2.7 that this apparently syntactic
condition on the placement of these particles follows from a particular semantic
theory of Q-particles and wh-words.
The same condition can be observed in Tlingit. The wh-operator of a Tlingit wh-
question may be contained inside an island if and only if the particle sá is merged
outside the island. When this occurs, the entire island is pied-piped into the left pe-
riphery of the interrogative clause. In the case of relative clause islands, the particle
sá must be merged to the right of the head of the relative clause.
(41) a. [ [ Wáa yateeyí CP] sháax’w sáani NP] sá ash tuwáa gaa yatee?
how they.are.REL girls Q his.spirit for they.are
What kind of girls are pleasing to his eye? (= Girls that are how are pleasing to
his eye?)
b. * [ [ Wáa sá yateeyí CP] sháax’w sáani NP] ash tuwáa gaa yatee?
how Q they.are.REL girls his.spirit for they.are
c. * [ [ Wáa yateeyí CP] sá sháax’w sáani NP] ash tuwáa gaa yatee?
how they.are.REL Q girls his.spirit for they.are
The speaker judgments indicated in (39)–(41) are consistent with the available
textual data as well. The table in (42) demonstrates that the selected corpus of texts
supports this generalization.
As the totals here clearly indicate, whenever a wh-operator in the selected corpus
is contained inside a relative clause, the particle sá appears to the right of both the
relative clause and the noun it modifies. Thus the selected corpus supports the gener-
alization that in Tlingit wh-questions, the wh-operator can be located inside of
an island if and only if the Q-particle sá is outside the island. To further illustrate
the content of this generalization, the following are some textually attested examples
of the Tlingit structure witnessed in (39)–(41); note that the relative clauses in the
structures below are ‘reduced relatives’.
Both the speaker judgments and the textual data above further emphasize the
syntactic parallels between Tlingit sá and Sinhala da. In section 2.5 we will see that
a uniform syntactic account can be provided for these facts, just so long as both these
particles share a syntactic categorization as Q-particles.
(44) Sinhala Da Cannot Appear at the Right Edge of a Matrix Clause (Kishimoto 2005)
a. Chitra monawa da gatte?
Chitra what Q buy
What did Chitra buy?
b. * Chitra monawa gatta da?
Chitra what buy Q (Kishimoto 2005: 3–4)
The speaker judgment data in (45)–(47) demonstrate that this property also holds for
Tlingit sá.
Despite the inability for da and sá to appear at the right edge of matrix clauses,
both particles may freely appear at the right edge of subordinate clauses, following a
subordinated predicate. This is illustrated for Sinhala da below.
Furthermore, sentences like (49) illustrate that this property also holds of Tlingit sá.
Thus we find that both Sinhala da and Tlingit sá share the property that, although
they can appear at the right edge of subordinate clauses, they cannot appear at the
right edge of a matrix clause. Again, it will be shown in section 2.6 that a uniform
account can be provided for these facts, but only if Tlingit sá and Sinhala da are as-
sumed to be the same formal entity, a Q-particle.
I conclude that Tlingit sá should most likely receive the same analysis as Sinhala da
and Japanese ka. As these latter two particles are classically glossed as ‘Q-particles’
(Hagstrom 1998; Kishimoto 2005), I conclude that Tlingit sá should therefore also
be glossed as a Q-particle.
In the preceding sections we have seen that (i) wh-operators occupy a left-peripheral
position in Tlingit wh-questions, and that (ii) wh-words in Tlingit are obligatorily
c-commanded by a Q-particle. In this section I will argue that the left-peripheral
position of these wh-operators is ultimately due to the movement of the Q-particle
that c-commands them. That is, I will argue that Tlingit does not actually possess a
rule of ‘wh-fronting’ per se. Rather, the generalization that the wh-operator is fronted
in a Tlingit wh-question is merely an epiphenomenal consequence of an operation
that targets the Q-particle.
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 37
However, before I can present evidence for this proposal, earlier summarized in
(1), I must clarify a number of technical points. The analysis sketched in (1) is vague
on several details. This informality was intended both to ease the discussion and to
show that the core proposals rely on few architectural assumptions, and so can be
exported into a variety of more specific frameworks. However, to make our argumen-
tation here more rigid, we must adopt a more precise statement of our syntactic
hypotheses than what appears in (1). In this context, it will help to begin with some
recent proposals concerning wh-in situ languages, which will provide our own pro-
posals with some clarifying context.
In recent work, an operation of ‘Q-movement’ has been argued to underlie the
formation of wh-questions in several wh-in situ languages. For example, various
lines of evidence lead Hagstrom (1998) and Kishimoto (2005) to propose the follow-
ing analysis of wh-questions in Sinhala, couched in the framework of ‘probes’ and
‘goals’, as developed within the Minimalist Program.
CP
CP Q1
IP CQ
Agree /
XP Attract
Adjunction
XP Q1
The structure in (51) represents the following claims. As we have seen, a wh-word in
Sinhala is obligatorily c-commanded by a Q-particle, the particle da. Under the
analysis in (51), this particle is adjoined to a phrase containing the wh-word. Further-
more, under this analysis it is the Q-particle da—not the wh-word—which is probed
by and Agrees with the interrogative C head of the wh-question. More concretely, the
interrogative C head bears an uninterpretable instance of the interpretable Q-feature
borne by da. The interrogative C must therefore probe for an interpretable instance
of the Q-feature. Upon reaching the adjoined Q-particle, the interrogative C Agrees
with the particle, eliminating its own uninterpretable instance of Q. This Agreement
then triggers movement of the goal, the Q-particle da, into the projection of C.
Because the Q-particle is adjoined to its sister, it may freely detach from its base
position. Therefore its movement into the CP, which in Sinhala is typically (though
not always) covert, leaves the wh-word and the phrases containing it in their base
positions at LF.28
Other lines of evidence lead Hagstrom (1998) to extend the ideas underlying this
analysis to wh-questions in Japanese. Hagstrom proposes (52) as the derivation of
wh-questions in Japanese.
38 THE GRAMMAR OF Q
CP
CP Q1
IP CQ
XP Agree /
Adjunction Attract
XP Q1
Overt Movement
…wh-word…
CP
QP1 CP
Complementation
XP Q
CQ IP
Agree/
… wh-word… Attract
QP1
Overt Movement
Under this analysis, the wh-questions of Tlingit receive a derivation nearly identical
to those of Japanese. The principal difference is that, in Tlingit, the Q-particle sá is
not adjoined to the phrase containing the wh-operator. Rather, it takes that phrase as
complement, thus projecting the category of the phrase minimally containing the
Q-particle and its sister. As a projection of Q, it would be natural to assume that this
QP also bears the Q-feature probed for by the interrogative C. Furthermore, because
this QP properly contains the Q-particle, it is the first node bearing the Q-feature to be
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 39
probed by the interrogative C. The standard algorithm for probing therefore entails
that interrogative C in Tlingit must Agree with this QP projection. As before, this
Agreement requires the goal—in this case, the QP—to move into the projection of the
interrogative C. Thus the entire QP is attracted into the left periphery of the wh-question.
Since this constituent necessarily contains the wh-operator of the wh-question, it
follows that such wh-words must occupy left-peripheral positions in wh-questions.29
We see, then, that the analysis in (53) links together the wh-questions of Tlingit, Sin-
hala, and Japanese in a typology of wh-question formation. Besides this, there are a number
of empirical considerations that support the analysis in (53) for Tlingit wh-questions.
First, it should be noted that Tlingit wh-questions require both the wh-word and
the Q-particle to be fronted. For example, sentence (54a) becomes ill-formed if sá is
left downstairs in its base position, as in (54b). Note that the ill-formedness of (54b)
is not due simply to a condition against the ‘stranding’ of sá. Such a condition, after
all, would not alone rule out the ill-formed (54c). In sentence (54c), the Q-particle sá
is not ‘stranded’ since its complement is the unmoved subordinate CP, a possibility
that is independently witnessed in sentences like (54d).
Moreover, we see below that the well-formed (55a) becomes ill-formed if the particle
sá is fronted without the wh-word, as in (55b). The ill-formedness of (55b) is
not simply due to a (prosodic) condition that sá follow some phrasal material in the
sentence, as (55c) illustrates.
Importantly, all the data in (54) and (55) follow naturally from the analysis in (53).
Under this analysis, a Tlingit wh-question must have the Q-particle sá within its left
periphery, thus ruling out (54b, c). Furthermore, the fronting of the QP necessarily
brings with it the wh-word associated with sá, as that wh-word is contained within
the QP. Thus sentences (55b, c) are also ruled out under the analysis.
40 THE GRAMMAR OF Q
Of course, one might conclude from (54) and (55) that both the wh-operator and
the Q-particle are directly attracted into the left periphery, perhaps by separate heads.
CP
DP1 CP
Daa
C CP
Q2 CP
sá
C IP
i éesh t1 t2 aawaxáa?
An immediate problem for the structure in (56), however, arises in the context of
multiple wh-questions. Sentences like (57a, b) demonstrate that all the wh-words of a
Tlingit multiple wh-question may front together into the left periphery of the
clause.
We can also see from (57a, b) and (27)–(29) that the order of wh-words and Q-parti-
cles in multiple wh-questions is such that each Q-particle immediately follows the
wh-word it is associated with. Therefore, if there were separate C heads attracting
wh-words and Q-particles of Tlingit, then the left periphery of a Tlingit multiple
wh-question must appear as in (58), where the CWH heads are those attracting
wh-words and the CQ heads are those attracting Q-particles.
(58) Structure Required for Multiple Wh-Fronting, Under the Analysis in (56)
[CP CWH-1 [CP CQ1 [CP CWH-2 [CP CQ2 . . . ] ] ] ]
We have already seen that the order of wh-words in a Tlingit multiple wh-question is
constrained by superiority; as shown again in (59), wh-subjects must precede
wh-objects.
However, if the left periphery of a multiple wh-question has the structure in (58),
then the simplest algorithm for probing would incorrectly derive the ill-formed,
superiority-violating orders in (59). The lowest CWH head would probe first, attract-
ing the highest wh-word in the IP. Only later will the higher CWH head probe for a
wh-word, and by this time, the only ‘visible’ wh-word left in the clause will be the
lower wh-word not probed by the first CWH head. The structure in (60) illustrates.
(60) Derivation of the Incorrect Orders in (59), via the Analysis in (56)
CPWH-1
Daa2 CPWH-1
CWH-1 ...
CPWH-2
aadóo1 CPWH-2
CWH-2 …
IP
On the other hand, the analysis proposed in (53) can derive the targeted word
order, assuming a theory of ‘Tucking-In’, as in Richards (1997). Under this analysis,
a single CQ head probes for both QPs in the multiple wh-question. Following the
standard algorithm for probing, this CQ first probes and attracts the highest QP in the
clause. Following this attraction, the CQ then continues to probe for additional QPs.
It subsequently probes and attracts the lower QP, requiring that the QP front into the
CQ projection. However, because of the constraint ‘Shortest Move’, this lower QP has
to be merged to as close a position to the CQ as possible. Consequently, the QP ‘tucks
in’, and moves to a Spec position lower than that occupied by the higher wh-word.
This derivation is sketched in (61).
(61) Derivation of the Correct Orders in (57), via the Analysis in (53) [with ‘Tucking In’]
CP
QP1 CP
Aadóo sá QP2 CP
daa sá CQ IP
Thus the view that there is a single head attracting the entire wh-word + Q complex
as a whole is necessitated by the word order facts in (57) and (59).
Finally, one might yet resist the analysis in (53) by suggesting that the single C
head attracting the wh-word + Q complex also probes for features of the wh-word.
That is, we have not yet ruled out that a single “Cwh/Q” head Agrees with both the
Q-particle and the wh-operator. In response, however, one might also point out that
there is yet no evidence that the C head does Agree with the wh-operator. After all,
the left-peripheral position of the wh-operator could very well result from the already
demonstrated relationship between the C head and the Q-particle, as proposed in
(53). In the absence of evidence that a relation holds between C and the wh-word, it
is simplest to assume that it does not.
We can, however, press the issue even further, and show that Agreement does
not hold between the C and the wh-operator. First, let us entertain a comparatively
strong view of syntactic islands, under which they are domains that no syntactic
relations may cross, not even probing and Agree. Assuming this view of islands, the
acceptability of sentence (62)—where the wh-word is contained within an island—
indicates that there is no probing/Agreement between it and the matrix C.
Now, one might counter by proposing that the wh-word in (62) is somehow accessible
to matrix C. Indeed, one might reject the notion that relative clauses are islands
to probing and Agree. However, any such proposal is immediately subject to the
following problem: recall the contrast between (62) and (63).
Sentence (63) differs from (62) only in that the Q-particle sá is directly adjacent to
the wh-word. Therefore any analysis which holds that the wh-word is syntactically
accessible to matrix C in (62) must equally well hold that the Q-particle is accessible
to matrix C in (63). Therefore the impossibility of (63) must follow from something
other than the fact that the Q-particle is located inside a syntactic island. What this
could be, however, remains unclear.30
The analysis in (53), however, predicts the contrast between (62) and (63), under
the assumption that no syntactic relationship may cross into an island.31 The impos-
sibility of (63) follows directly from the placement of the Q inside a relative clause
island. When the Q-particle is located outside the island, as in (62), it is accessible to
matrix C, and the sentence is well formed. The fact that the wh-word in (62) remains
inside the island has no bearing on the well-formedness of the sentence, given that
the matrix C bears no syntactic relationship to the wh-operator itself. We find, then,
that the contrast between (62) and (63) supports what is, perhaps, the most unusual
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 43
feature of the analysis in (53): the existence of a relationship between the interroga-
tive C and the Q-particle, but not between the C and the wh-word.32
The preceding arguments demonstrate the empirical support held by the analysis
of Tlingit wh-questions in (53), an analysis independently motivated by their similar-
ity to the wh-questions of Sinhala and Japanese. I conclude that (53) is, in essence,
the correct analysis of Tlingit wh-questions.
In the preceding section I presented arguments supporting the analysis in (1)/(53) for
Tlingit wh-questions. None of those arguments, however, directly addresses one cru-
cial aspect of (1)/(53): the claim that the Q-particle sá takes its sister as complement
rather than being adjoined to its sister as in (51) and (52). Of course, this assumption
is far simpler than the alternative, in as much as it treats the concomitant fronting of
the wh-word as resulting from normal phrasal movement of the QP. Nevertheless, it
would be optimal to find some independent evidence that the wh-word is dominated
by a QP-projection in Tlingit. We will see in this section that the behavior of sá in
certain environments provides this evidence.
In outline, this section concerns some further constraints on the placement of sá
within the clause. We will see that the particle sá cannot appear (i) between a postpo-
sition and its complement, (ii) between a wh-possessor and the possessed NP, or (iii)
between a wh-determiner and its NP complement. We will then first consider the
possibility that these constraints follow from the fact that QPs must undergo move-
ment in Tlingit, given that the three aforementioned environments tend to be islands
for movement.
Although an attractive possibility, I will ultimately reject such a movement-based
analysis of the patterns in (i)–(iii). In brief, we will see that the generalizations in (i)–(iii)
also hold for QPs interpreted as indefinites. Therefore a movement-based analysis of
(i)–(iii) must assume that such indefinite QPs also undergo obligatory movement, pre-
sumably obligatory QR. However, there is evidence that indefinite QPs do not undergo
obligatory QR: indefinite QPs can scope out of islands. Therefore QR is not necessary
to fix their scope, and so the putative obligatory movement remains mysterious.
In place of a movement-based account, I propose a condition governing the po-
sition of QPs, which I dub the ‘QP-Intervention Condition.’ I show that this condi-
tion can capture the observed patterns, as well as avoid the problems inherent in a
movement-based account.
Importantly, this account of the patterns in (i)–(iii) requires that a QP-projection
dominate the wh-word in Tlingit. Thus the success of the account provides some
additional evidence that Q in Tlingit takes its sister as complement. More impor-
tantly, however, we will see that, given the analysis in (1)/(53), the QP-Intervention
Condition alone derives various constraints on the form of Tlingit wh-questions, con-
straints which in other languages are typically analyzed in terms of conditions on
the operation of movement. Therefore the discussion here foreshadows my arguments
in later chapters that the analysis in (1)/(53) applies not simply to Tlingit, but to all
wh-fronting languages.
44 THE GRAMMAR OF Q
The judgments above correspond with the patterns found in our Tlingit corpus, as
illustrated in (66).
As the numbers here clearly indicate, whenever a wh-operator in the selected corpus
appears (in a phrase) marked by a postposition, the particle sá appears to the right of
that postposition. Thus we find that the order ‘[wh . . . Q . . . P]’ is textually unattested.
Only the order ‘[wh . . . P]Q’ is found in natural speech.
A second constraint on the position of sá is that it cannot, in a Tlingit wh-
question, intervene between a possessor and the possessed NP. This is illustrated by
speaker judgments like (67)–(70).
As before, the judgments here correspond with the patterns found in our Tlingit
corpus.
As the numbers here clearly indicate, whenever a wh-operator in the selected corpus
functions as a possessor modifying a possessed NP, the particle sá appears to the
right of that possessed NP.
Finally, a third constraint on the position of sá is that it cannot, in a Tlingit wh-
question, intervene between a determiner and its NP complement. This is illustrated
by speaker judgments like (72)–(75).
Once again, the speaker judgments here correspond with what we find in our Tlingit
corpus.
As the numbers in (76) show, whenever a wh-operator in the selected corpus functions
as a determiner, the particle sá appears to the right of its NP complement.
In summary, then, we can see from the array of data above that in a Tlingit wh-
question, the particle sá cannot intervene (i) between a postposition and its comple-
ment, (ii) between a possessor and a possessed NP, and (iii) between a determiner
and its NP complement.
Of course, when these restrictions are observed in the context of wh-questions,
as they are here, they might not seem very surprising, particularly given our analysis
in (1)/(53). Under that analysis, a Tlingit wh-question requires that the QP be fronted
into the left periphery. Therefore placement of the Q-particle between, for example,
a postposition and its complement, would, in a wh-question, result in extraction of
the complement of P, as illustrated in (77).
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 47
CP
QP1 C´
DP Q CQ IP
…wh-word…
PP
QP P
t1
Extractions of this form, however, are ill-formed in many languages of the world, a
pattern that leads many to view such movements as cross-linguistically marked (Ross
1967; Abels 2003; Heck 2008). Similarly, placement of the Q-particle between a
WH-possessor and the possessed NP in a wh-question would entail extraction of the
specifier of DP, as shown in (78).
(78) The Order [ WH-POSS . . . Q . . . NP ] in Wh-Questions Entails Possessor Extraction
CP
QP1 C
DP Q CQ IP
…wh-word…
DP
QP D
t1 D NP
POSS possessum
CP
QP1 C
D Q CQ IP
wh-word
DP
QP NP
t1
48 THE GRAMMAR OF Q
Once more, though, such extractions of D from DP as in (79) are in many languages
impossible, which has led linguists to conclude that they are also in violation of
certain general movement constraints (Ross 1967; Corver 1990, 2007; Bošković
2005a).
It is apparent, then, that each of the ill-formed structures in (64)–(75) contains
an extraction that independently appears to be ill-formed across languages. Therefore
the ill-formedness of those structures may simply follow from whatever constraints
on movement disallow such extraction. Given that this is a rather natural and compel-
ling line of explanation, I will dignify it with a name.
The table in (85) is identical to that in (66), except that it tracks the position of sá
in declarative clauses containing wh-indefinites. As we saw before, the totals here
demonstrate that whenever a wh-indefinite in the selected corpus appears (in a phrase)
marked by a postposition, the particle sá appears to the right of that postposition.
Thus we find that with wh-indefinites too, only the order ‘[ wh . . . P ] Q’ is found in
naturally occurring speech.
Similarly, the constraint against placement of sá between a possessor and a pos-
sessed NP holds even for wh-indefinites. This is illustrated by speaker judgments like
(86)–(90).
Again, textual analysis independently supports the generalizations gained from the
speaker data in (86)–(90).
The table in (91) is again nearly identical to the table in (71), the only difference
being that it tracks the position of sá in declarative clauses containing wh-indefinites.
As the totals here again indicate, whenever a wh-indefinite in the selected corpus
functions as a possessor modifying a possessed NP, the particle sá appears to the
right of that possessed NP, just as in wh-questions.
Finally, further investigation reveals that the inability of sá to appear between a
wh-D and its NP complement also holds for wh-D’s functioning as indefinites in
declarative clauses.
Additional support for this generalization can again be found in the selected corpus
of Tlingit texts.
appear inside extraction islands. Following this, I challenge the claim, crucial for
the movement-based analysis, that wh-indefinites in Tlingit undergo obligatory
QR.
To begin, let us note that according to the movement-based account of (81)–(94),
the inability for QPs to occupy either CompPP or SpecDP follows partly from the
fact that these positions are islands for extraction. Under the logic of this account,
then, it would follow that QPs containing wh-indefinites should generally be unable
to appear inside extraction islands.
This prediction, however, is not true. There is no general constraint in Tlingit
against sá appearing within an island. Recall that relative clauses in Tlingit are islands
for extraction, and that in a wh-question the Q-particle cannot appear inside of a
relative clause.
(96) QPs Containing Wh-Indefinites Can Appear Inside Relative Clause Islands
Wáa sá yatee [ wé [ l goodéi sá wugoodi ] káa ]?
how Q he.is that not where.to Q he.went.REL man
How is the man who didn’t go anywhere?
The well-formedness of (96) suggests that the ill-formedness of the illicit structures
in (81)–(94) is not due simply to their QPs appearing within an extraction island.
Beyond this, however, there are some deep difficulties for one core assumption
of the movement-based account in (80), the idea that wh-indefinites in Tlingit oblig-
atorily undergo QR.
Let us begin here by asking why QR of Tlingit wh-indefinites should be obliga-
tory. Such obligatory QR would seem to imply that wh-indefinites in Tlingit cannot
obtain their scope in situ. After all, if wh-indefinites in Tlingit could obtain their
scope in situ, there would be no reason for QR to obligatorily target all such indefi-
nites (Reinhart 1997).
Therefore the movement-based analysis predicts that wh-indefinites in Tlingit
always move to their scope positions via QR. Given the assumption that QR is sensi-
tive to (adjunct) islands (Chomsky 1975; Reinhart 1997), we therefore predict that
Tlingit wh-indefinites should—like strong quantifiers—be unable to scope out of
such islands (Reinhart 1997).
Let us, then, seek to test this prediction of the movement-based account. As
shown in the work of Ruys (1992, 2000) and Matthewson (1999), an effective test
can be gained by looking to the interpretation of indefinites contained within con-
ditionals. For example, the well-known fact that English indefinites can scope out
of islands is demonstrated by the consistency and coherency of the discourse in
(97).
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 53
As noted by Ruys (1992, 2000), in order for the discourse in (97) to be coherent,
sentence (97e) must be interpreted so that the existential force of the indefinite “one
of Dave’s in-laws” has scope outside the conditional antecedent. If the existential
force were interpreted internal to the conditional antecedent, the final sentence would
be equivalent to “if any of Dave’s in-laws win the lottery, he’ll get a house”, which is
inconsistent with the prior discourse.
Thus the coherence of discourses like (97) shows that indefinites in English can
scope out of a conditional antecedent, a structure that otherwise constitutes a move-
ment island. This fact, in turn, has led many to abandon the notion that English indef-
inites must undergo movement (QR) to fix their scope (Reinhart 1997; cf.
Schwarzschild 2002). Consequently, if it can be shown that Tlingit allows for dis-
courses comparable to (97), where the final sentence contains a wh-indefinite in
the conditional antecedent, then we may conclude that such wh-indefinites need not
undergo QR.
In order to apply this test, we should first confirm that Tlingit wh-indefinites al-
low the ‘specific readings’ that are required for indefinites to obtain the exceptional
scope seen in (97e) (Fodor & Sag 1982; Kratzer 1998; Matthewson 1999). First, the
reader may note that in many of the examples we have seen thus far, the Tlingit wh-
indefinite is apparently interpreted as an NPI or free-choice item ((18a, b), (81)–(92)).
Nevertheless, it is possible for wh-indefinites in Tlingit to be interpreted as plain
existentials, outside the scope of any other logical operators.35 Listed in (98) are some
textually attested examples.
Furthermore, there is at least one passage within our Tlingit corpus where such a
‘specific wh-indefinite’ appears to take scope outside of an adjunct island.
The context from which sentence (99) was taken makes clear that the wh-indefinite
here is to be given a specific reading. The speaker does not mean to state that the baby
began to fuss as soon as the mother had gone any distance at all, which would be the
interpretation if the wh-indefinite had scope inside the temporal adjunct. Rather, the
speaker is stating that the baby’s fussing began only after a particular distance had
been crossed—that there is a certain distance after which the baby began to fuss.
Such an interpretation requires that the wh-indefinite have wide scope, outside the
temporal adjunct.
From the previous data, it is apparent that Tlingit wh-indefinites can receive a
‘specific’ reading, and so we should be able to obtain speaker judgments regarding
whether discourses comparable to (97) are possible. Consider, then, the following
Tlingit discourse.
The Tlingit discourse in (100) was constructed with the help of a native speaker, who
recognized the discourse as a sensible story and an accurate translation of the English
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 55
original in (97).36 Note that if the wh-indefinite in (100e) could only have narrow
scope inside the antecedent of the conditional, then the discourse in (100) would be
neither internally consistent nor an accurate translation of the original English story.
Rather, the consistency and faithfulness of (100) require that the existential force of
the wh-indefinite be located outside the antecedent of the conditional.
Sentences like (100e) therefore demonstrate that, contrary to the predictions of
the movement-based account, it is possible for Tlingit wh-indefinites to scope out of
(adjunct) islands. More acutely, these sentences show that, in Tlingit, there are mech-
anisms other than movement (QR) that determine the scope of an indefinite’s existen-
tial force. This, however, entails that the scope of a wh-indefinite in Tlingit need not
be fixed by movement (QR), contrary to the assumptions of the movement-based ac-
count. That is, because wh-indefinites in Tlingit can obtain their scope without under-
going QR, there is simply no reason for QR to obligatorily target all such indefinites.
Of course, one could always maintain the movement-based account of (81)–(94)
by simply giving up the notion that QR is the movement operation that obligatorily
targets Tlingit wh-indefinites. However, since we have already eliminated all other
plausible candidates, we find that the movement-based account must appeal to an as
yet unknown form of covert movement. It is therefore most reasonable to conclude
that the movement-based account in (80) is simply incorrect.
PP
P
QP
DP Q
…wh-word…
56 THE GRAMMAR OF Q
DP
QP D
DP Q D NP
POSS
possessum
…wh-word…
DP
QP NP
D Q
wh-word
QP
PP Q
DP P
…wh-word…
QP
DP Q
DP D
D NP
…wh-word… POSS
possessum
QP
DP Q
D NP
wh-word
L QP F QP
Q XP Q XP
XXXX
Selection: OK QP-Intervention Condition Violated
If sá were to occur to the right of the matrix predicate, then there are two logical
possibilities concerning its exact position in the clause, neither of which is consistent
with the stated properties of Q in Tlingit. First, it could be the case that sá takes the
entire matrix CP as complement, as illustrated in (109).
QP
CP Q
sá
Daa iyatéen
However, we will see in the next section that our semantics for Tlingit wh-questions
would rule out such a structure. Briefly, Q-particles are analyzed as variables that
must be bound by higher operators. In structures like (109), however, there is no
position for a higher, binding operator to appear. Thus the Q-particle goes unbound,
and the sentence cannot be interpreted.
The second possibility regarding the position of a clause-final sá is that it takes
as complement either the VP or one of the higher projections along the ‘functional
spine’ of the clause. In either case, however, the QP-Intervention Condition would be
violated. If sá were to take VP as complement, then the QP it projects would inter-
vene between VP and the higher Infl head, as illustrated in (110).
IP
QP I
VP Q
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 59
However, Infl head is a functional head, and selects for the VP complement of Q.
Therefore (110) violates the QP-Intervention Condition. Similarly, if Q were to take
as complement any higher projection, F1, along the ‘functional spine’ of the clause,
the QP it projects would intervene between F1P and the next higher functional
projection, F2P.
F2 P
QP F2
F1 P Q
Again, though, F2 is a functional head, and selects for the F1P complement of Q.
Therefore (111) also violates the QP-Intervention Condition.
We have thus ruled out the ability for Tlingit sá to appear anywhere to the right
of the matrix predicate.40 Nevertheless, our account does correctly predict that sá can
appear to the right of a subordinate clause, as in (112).
CP
QP1 C
DP Q CQ IP
…wh-word…
PP Impossible PP,
Ruled Out by
QP P QP-Intervention Condition
t1
We have already seen, however, that such base structures are impossible in Tlingit,
and are ruled out by the QP-Intervention Condition. Since the QP-Intervention Con-
dition rules out the base structure that necessarily underlies P-stranding, it thereby
rules out P-stranding in Tlingit, and so no special condition against such extractions
need be appealed to in the grammar of Tlingit.
Next, let us turn to the ill-formed possessor-extraction sentences in (67)–(70).
Note that, according to our core analysis in (1)/(53), each of these structures would
at earlier stages of their derivation have a QP intervening between a possessor and
the possessive D that selects the possessor. Again, however, such configurations are
independently ruled out by the QP-Intervention Condition.
CP
QP1 C’
DP Q CQ IP
…wh-word…
DP
Impossible DP,
QP D’ Ruled Out by
QP-Intervention
t1 D NP Condition
POSS possessum
CP
QP1 C
D Q CQ IP
wh-word
DP Impossible DP,
Ruled Out by
QP NP QP-Intervention Condition
t1
In summary, we find that all the ill-formed sentences in (64)–(76) could only be
derived from structures that violate the QP-Intervention Condition. Thus the
QP-Intervention Condition in (107) is alone sufficient to rule out the ill-formed
sentences in (64)–(76), and therefore does provide a uniform account for all the data
in (64)–(94). Furthermore, we have seen that it also predicts the fact, observed
earlier, that the Tlingit sá can appear to the right of subordinate predicates, but not to
the right of matrix predicates. Consequently, I conclude that the QP-Intervention
Condition offers the best analysis of the data we have seen in this section.
At this point, let us note two further consequences of this analysis. First, it should
be observed that, given the success of this analysis, we have further confirmation that
the Q-particles of Tlingit take their sisters as complements and project a QP. If we
were to assume that Tlingit Q-particles, like those in Japanese, simply adjoined to
their sisters, and projected no higher phrasal category, then the statement of the QP-
Intervention Condition becomes much more difficult. It would have to be rephrased
to something like “no Q-particle can be immediately dominated by a node occurring
between a functional head F and a phrase F selects for”, a condition which must ap-
peal to more complex tree-geometric relations. Thus the assumption that Q in Tlingit
takes its sister as complement not only simplifies the theory of Tlingit wh-fronting, it
is also crucial for understanding the wider distributional properties of Tlingit sá.
Finally, let us note a potential consequence for the general theory of movement;
my comments here will be brief, as I will more fully treat this subject in chapter 4.
Under our analysis, the impossibility of the extractions in (77), (78), and (79) is ulti-
mately due not to any constraint on extraction per se, but to independently visible
constraints on the placement of Q. Such constraints serve to limit wh-extraction by
limiting the structural preconditions for it, ruling out the base structures from which
the ill-formed extractions must be derived. Thus, rather than explain the impossibil-
ity of such extractions in terms of the ‘islandhood’ of the base positions, we explain
the apparent islandhood of those positions in terms of independently visible con-
straints on the position of Q.
Recall, however, that the impossible extractions of the kind seen in (77)–(79) are
found to be ill-formed in many languages of the world. Just as we should seek a uni-
form Tlingit-internal explanation of the facts in (64)–(94), we should likewise seek a
uniform account of these facts across languages. Given the evidence supporting the
62 THE GRAMMAR OF Q
When we combine this semantics for QPs with our hypothesis concerning selection,
we effectively derive the ‘QP-Intervention Condition’ in (107).
First, let us show that QPs will be permitted to intervene between a lexical head
and any phrase selected by that lexical head. Let L be a lexical head s-selecting for
a phrase of semantic type τ. By assumption, L has no c-selectional requirements.
Therefore the selectional requirements of L will be satisfied as long as its complement
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 63
is of type τ, regardless of its syntactic category. Now, recall that a QP where Q takes
as sister a phrase of type τ will itself be of type τ. Consequently, such a QP will
satisfy the selectional requirements of L. Thus we see that a QP will be able to intervene
between any lexical head L and a phrase of the semantic type selected by L. This
general result is schematized in (117).
LP
On the other hand, our system predicts that QPs will not be able to intervene
between a functional head F and a phrase selected by F. Recall that, by assumption,
a functional head F also c-selects for its arguments. Consequently, unless F actually
c-selects for a QP, the selectional requirements of F will not be satisfied by any QP.
Consequently, a QP will be unable to intervene between any functional head F and a
phrase of the category selected by the functional head.
FP
F XXX QP
c-selection: XP c-selection not
s-selection: satisfied by QP
Q XP
Thus we find that the content of our QP-Intervention Condition in (107) could
follow from a broader difference in the selectional requirements held by lexical and
functional categories. Of course, whether or not (107) can be shown to follow from
more general principals of grammar does not impact its truth as a surface generaliza-
tion. Since our ensuing discussion hinges only upon the descriptive adequacy of
(107), I leave to future work a more thorough study of what accounts for the gener-
alization as stated.
In this section I provide a semantics for the wh-words and Q-particles of Tlingit that
assigns the correct interpretations to the somewhat exotic tree structures employed
by our Q-based account in (1)/(53). Furthermore, we will see that this semantics can
alone account for various observed properties of Q. In chapter 3 we will see that this
64 THE GRAMMAR OF Q
semantics can also be applied to the wh-words and Q-particles of Japanese and Sin-
hala, further emphasizing the formal identity between Tlingit sá, Japanese ka, and
Sinhala da.
For reasons of space, the discussion here will be rather compact, and will be
most comprehensible to those with some familiarity with current work on the seman-
tics of wh-words and Q-particles, particularly Hagstrom (1998), Shimoyama (2001),
Kratzer and Shimoyama (2002), and Beck (2006).
The rules outlined in (119) indicate that the value of [[ X ]]F depends upon whether
or not X bears intonational focus. If X is not focused, then its focus-semantic value
is simply the singleton set of its normal-semantic value. If X is focused, however,
then its focus-semantic value is its set of ‘focus alternatives’, the set of all the entities
of its semantic type.
Note that, since the rule in (119a) applies only to unfocused heads, the rules in
(119) do not yet provide a focus-semantic value for unfocused phrases. In order to
rectify this, we introduce a special rule known as ‘pointwise semantic composition’.
It is defined in (120).
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 65
Finally, let us review our assumptions regarding the meaning of so-called focus-
sensitive operators, such as “only”. The core property of a focus-sensitive operator is
that its semantic contribution depends upon the focus-semantic value of its sister. For
example, the meaning of “only” might be represented as in (122).
According to this semantics, the operator “only” contributes the following information:
(a) the normal-semantic value of its sister is true, and (b) for all the propositions in
the focus-semantic value of its sister, if they are not equal to the normal semantic
value of its sister, then they are false. Given this semantics, we can compute the truth
conditions of a sentence like “Only DAVEF smokes”.
The definitions introduced in (119), (120), and (122) will constitute the formal
theory of focus semantics employed throughout our discussion. For a more complete
66 THE GRAMMAR OF Q
background into systems of this sort, I refer the reader to Rooth (1985, 1992), Kratzer
(1991), and Hagstrom (1998).
(Tlingit), nani (Japanese), and mokak (Sinhala) have the following characteristic
semantics.
Similarly, the wh-words who (English), aadóo (Tlingit), dare (Japanese), and kauru
(Sinhala) all have the following characteristic semantics.
There are several benefits to this particular treatment of wh-words. First, as orig-
inally shown by Beck (2006) (and as we will see in chapters 3 and 4), this semantics
provides an interesting account of so-called LF- or Focus-Intervention Effects across
languages. Furthermore, it provides a reason why wh-words must be focused in so
many of the world’s languages; if wh-words were not focused, then a semantic crash
would necessarily result.43 It should also be noted that this system, unlike those in
Hagstrom (1998) and Shimoyama (2001), identifies the ‘alternatives’ denoted by wh-
words as focus-semantic values. This eliminates the need for special rules of point-
wise semantic composition specifically for the values of wh-words, and instead
employs the pointwise composition rules that are independently needed for the com-
putation of focus-semantic values. Finally, we will see later that, given plausible an-
cillary hypotheses, this treatment predicts several of the core grammatical properties
of Q.
The semantics in (126) states that the meaning of Qi relative to a variable assignment
g is the value that g assigns to the index i, which is stipulated to be some element
68 THE GRAMMAR OF Q
from the domain of choice functions. To briefly review, a ‘choice function’ is any
function that takes a set as argument and returns a member of that set as its value.
Thus all the functions illustrated in (127) qualify as choice functions.
(127) Illustrative Examples of Choice Functions (Reinhart 1997, Winter 1997)
A choice function takes a set and returns a member of that set.
f({Dave, John, Larry, Phil}) = Larry
g({the Bible, the phonebook, LSLT}) = the Bible
h({Amherst, Boston, Natick, Worcester}) = Worcester
Given that they are variables, higher operators can—and in fact must—bind the
Q-particles. For example, we will see later that the interrogative C head of a wh-
question is semantically an interrogative operator binding the choice function variable
contributed by Q. In addition to this, I assume that an existential operator over
choice functions can be inserted via a rule of existential closure (Reinhart 1992, 1997;
Yatsushiro 2001). I assume here that such existential closure occurs at the level of the
IP, to bind any free variables occurring within the IP (cf. Reinhart 1997, Kratzer
1998).
The final ingredient of our semantics for Q concerns its method of semantic co-
mposition. Q-particles are assumed to semantically compose with their sisters via a
syncategorematic rule specific to Q-particles (Beck 2006). The normal-semantic
value of a Q-particle and its sister is stipulated to be the normal-semantic value of the
Q-particle applied to the focus-semantic value of its sister.
Note that according to this semantics, a Q-particle takes as argument the focus-
semantic value of its sister. Therefore, as alluded to earlier, Q-particles are claimed
to be focus-sensitive operators.
As the derivation in (130) shows, relative to the variable assignment g, the seman-
tic value of the QP Daa sá1 ‘What Q’ is the choice function that g assigns to
the index ‘1’, taking as argument the set of all nonhuman individual concepts.
Of course, given the nature of choice functions, this means that the ultimate
semantic value of this QP is some nonhuman individual concept, and so such a QP
is of type <se>.
The semantic computation becomes a bit more interesting when the Q-particle
is not directly adjacent to the wh-word. Such a structure is witnessed by the
Tlingit phrase Aadóo yaagú sá, ‘Whose boat Q’, which we might assume to have
the structure indicated in (131). The semantic value of (131) is computed in
(132).
(131) Assumed Structure of the Complex QP Aadóo Yaagú Sá, ‘Whose boat Q’
QP n-sem: <se>
As the computation demonstrates, the semantic value of the Tlingit phrase Aadóo
yaagú sá1, ‘Whose boat Q1’, relative to a variable assignment g is the choice
function f that g assigns to the index 1, taking as argument all those individual
70 THE GRAMMAR OF Q
concepts which for some human individual concept x<se> are the concept [λw. the
boat(w) of x(w) in w ]. Therefore, given the nature of choice functions, the ulti-
mate semantic value of this QP is an element from this set, and so such a QP is
type <se>.
Having presented the semantic computations for these two types of QPs, we
can now see how our remaining hypotheses provide a compositional semantics for
Tlingit wh-indefinites. Briefly, the picture is as follows: a Tlingit wh-indefinite sim-
ply arises when the QP containing the wh-word remains inside the IP, and so is bound
by existential closure. For example, the following illustrates the interpretation our
system derives for the Tlingit sentence in (98e), Daa sá xwasiteen, ‘I saw some-
thing’.
An inspection of the tree in (133) reveals that our system assigns to the structure in
(133) the proposition ‘[λw. ∃f. I saw [f ({ x<se> : x ∉ human })] in w ]’.46 At a partic-
ular world w, this proposition is true iff there is some choice function f such that f
applied to the set of nonhuman things yields an entity that the speaker saw in w. Of
course, such a choice function exists if and only if there is a nonhuman entity that the
speaker saw in w. We conclude, therefore, that our semantics correctly assigns to
(133) its observed meaning, a proposition equivalent to ‘there is some nonhuman
thing which the speaker saw’.
Let us next consider a somewhat more complex case, one where the Q-particle
is not directly adjacent to the wh-word. Such a wh-indefinite appears in sentence
(86a): Tléil aadóo yaagú sá xwsateen, ‘I didn’t see anyone’s boat’. The following
illustrates the interpretation that our semantics assigns to (86a).
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 71
The tree in (134) reveals that our system derives as the meaning of (134) the follow-
ing proposition: ‘[λw. ¬∃f . I saw [f ({ h<se> : ∃x<se> ∈ human . h = λw’. the boat(w’)
of x(w’) in w’ })](w) in w]’. To break down this complex formula, let us first note
that at a particular word w, this proposition is true if there is no choice function f such
that the speaker saw the entity which f yields when applied to the set of all individual
concepts of the form [λw. the boat(w) of x(w) in w], where x is human. Of course, this
is equivalent to stating that there is no individual concept h of the form [λw. the
boat(w) of x(w) in w], where x is human, such that the speaker saw h. Finally, note
that this is itself equivalent to simply stating that there is no human x such that the
speaker saw the boat of x in w. We therefore conclude that our semantics correctly
assigns to the structure in (134) its observed meaning, a proposition equivalent to
‘there is no person such that the speaker saw that person’s boat’.
We have thus far seen two cases where our semantics assigns the correct meaning
to Tlingit sentences containing wh-indefinites. In both cases, we find that existential
quantification over choice functions is truth-conditionally equivalent to existential
quantification over the set of entities denoted by the wh-word. Of course, this result is
most interesting in those cases where the Q-particle is not directly adjacent to the wh-
word itself. In such cases, the argument of the choice function variable is not simply
the set contributed by the wh-word, and so the equivalence between existential quan-
tification over that set and over the domain of choice functions is not so intuitively
72 THE GRAMMAR OF Q
obvious. We may rightly wonder, then, whether the result in (134) is but a particular
instance of a more general one. That is, is it ever the case in our system that altering
the position of the Q-particle alters the proposition denoted by the sentence?
The answer to this question is ‘no’. As long as the Q-particle c-commands the
wh-word and is c-commanded by the existential quantifier, the same propositional
function is assigned to the structure, one that is equivalent to existential quantification
over the set denoted by the wh-word. This result will be of some importance to our
discussion in chapter 4, as it will allow us to convert otherwise complex formulas into
more manageable and recognizable ones. Let us therefore see in detail why it is so.
We wish to show that no matter how far the Q-particle is from the wh-word,
existential quantification over the domain of choice functions is still equivalent to
existential quantification over the set of entities denoted by the wh-word. In other
words, we wish to show that any arbitrary structure of the form in (135a), where τ is
the semantic type of the wh-word, is equivalent to the structure in (135b), where the
wh-word is replaced with a (normal) variable of its type, the Q-particle is removed,
and the type-τ variable is existentially bound and restricted to the set denoted by the
wh-word.
First, let us note that our semantics entails that the syntactic structure in (135a)
will be assigned the following as its truth conditions: ∃f [A . . . f [B . . . { x : x ∈
Dτ & P(x) } . . . ] . . . ], where ‘P(x)’ represents the lexical presuppositions of the
wh-word (e.g. ‘human(x)’ for who, and ‘¬human(x)’ for what). Thus the structure
in (135a) will be true iff there is some choice function f ∈ Dcf such that the follow-
ing holds: [A . . . f [B . . . { x : x ∈ Dτ & P(x) } . . . ] . . . ]. Of course, given the
nature of choice functions, this last condition holds if and only if there is some R
within the set constituting [B . . . { x : x ∈ Dτ & P(x) } . . . ] such that [A . . . R . . .
] holds. Note, however, that since R is a member of [B . . . { x : x ∈ Dτ & P(x) } . . .
], our semantics entails that there is some entity a ∈ { x : x ∈ Dτ & P(x) } such that
R = [B . . . a . . . ].47 Thus [A . . . R . . . ] holds iff for some a ∈ { x : x ∈ Dτ & P(x)
}, [A . . . [B . . . a . . . ] . . . ] holds. Finally, this later condition holds iff the structure
in (135b) is true, which was to be proven. This chain of reasoning is laid out in
(136).
The equivalence between (135a) and (135b) can also be seen via reflection upon
the meanings of the QPs themselves and the nature of variable assignments. Recall
that Q-particles are variables over choice functions. Therefore, relative to an arbitrary
variable assignment g, the semantic value of ‘[QP [XP . . . whτ . . . ] Q ]’ will be some
entity A of the normal-semantic type of the XP. Now, consider the structure ‘[XP . . .
yτ . . . ]’, where the wh-word is replaced with a (normal) variable of its type (which
is also assumed to share the presuppositions of the wh-word). Relative to a variable
assignment g, this latter structure is also some entity A′ of the normal-semantic type
of the XP.
Clearly, then, for any variable assignment g such that [[ [QP [XP . . . whτ . . . ] Q ] ]]g
= A, there is some other variable assignment g′ such that [[ [XP . . . yτ . . . ] ]]g′ = A, and
vice versa. Therefore, if there is a variable assignment g such that [[ [A . . . Q1 [B . . .
wh-wordτ . . . ] . . . ] ]]g = T, then there is a variable assignment g′ such that [[ [A . . .
[B . . . yτ . . . ] . . . ] ]]g′ = T, and vice versa. It follows, then, that if ever the structure
in (135a) is true, then the structure in (135b) must also be true, and vice versa.
As mentioned before, the equivalence between (135a) and (135b) will provide
us a powerful tool, as it will permit the simplification of several otherwise complex
formulas that our semantics derives. Of course, a more immediate consequence of
(136) is that the existential interpretation of wh-indefinites in Tlingit may be obtained
from our semantics via existential closure over the choice function variable contrib-
uted by Q. Thus our semantic theory illuminates the essential role played by so-
called Q-particles in the meaning of wh-indefinites in declarative clauses. The reader
may recall, however, that Q-particles also seem to play a crucial role in the semantics
of wh-questions (hence, their very moniker). In a moment, we will see how the
semantics just proposed can be incorporated into a treatment of Tlingit wh-questions.
Before that, however, we will first observe two notable corollaries of our semantics.
As stated, this principle entails that any sentence that cannot be computed to have a
normal-semantic value is ill-formed.
The second crucial assumption is somewhat more complex, though equally as
important.
Let us pause for a moment to consider what the condition in (138) states. First, let us
note that, according to the semantics in (129), repeated in (139), the Q-particle takes
as argument only the focus-semantic value of its sister.
Thus we find that Q differs from “only” in that computing the meaning of a phrase con-
taining Q does not require that one compute the normal-semantic value of Q’s sister. The
principle in (138)—which is crucial for the theory of Beck (2006)—states that, in fact, it
is only the Q-particle which has this peculiar insensitivity to normal-semantic values.
Interestingly, the principles in (137) and (138) alone predict that wh-words
must be c-commanded by Q-particles. Suppose that a wh-word in a given sentence
is not c-commanded by a Q-particle. By assumption, then, either (i) the wh-word is
c-commanded by a focus-sensitive operator OP that is not Q, or (ii) the wh-word is
not c-commanded by any focus-sensitive operator. Let us first consider condition
(i). Since OP is not a Q-particle, principle (138) entails that the semantic computa-
tion for the entire sentence requires one to compute the normal-semantic value of
the sister of OP. However, since OP c-commands the wh-word, it follows that the
sister of OP contains the wh-word. Therefore, computing the normal-semantic
value of the sister of OP requires one to compute the normal-semantic value of
the wh-word, and so the sentence is predicted to be uninterpretable. Now, let us
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 75
In essence, the principle in (141) states that if any phrase X has a normal semantic
value, then the normal-semantic value of any phrase containing X must partly depend
upon X’s normal-semantics. Thus all normal-semantic values must be ‘used’ in the
computation of a phrase’s normal-semantic interpretation.
If we accept the principle in (141), we thereby predict the need for Q-particles to
c-command wh-words. To see this, let us first consider the structure in (142), where
the sister to Q does not contain a wh-word, but does contain a focused DP. As the
semantic derivation in (142) demonstrates, such a structure is, in principle, interpret-
able by our system.
(142) Interpretable Structure Where the Q-Particle Does Not C-Command a Wh-Word
76 THE GRAMMAR OF Q
As illustrated in (142), our semantics requires only that Qs have some focused XP
within their scope; the semantics alone does not entail that the focused XP be a wh-
word. However, with the addition of the principle in (141), structures like that in
(142) may be ruled out. Consider, for example, the structure in (143), which differs
from (142) only in the normal-semantic value of the focused DP.
(143) Interpretable Structure Where the Q-Particle Does Not C-Command a Wh-Word
Although the focused DP in (143) differs in its normal-semantic value from that in
(142), the normal-semantic value of the entire QP remains unchanged. Furthermore,
given our rules for computing focus-semantic values in (119), it follows that if one
attempts to replace ‘Frank’ in (142) with any expression of its type, the interpretation
of the QP will always be the same, namely, f ({ x : x ∈ De}). Thus we find that the
structure in (142) violates the Principle of Full Interpretation in (141). Consequently
the focused phrase within a QP can never be just any arbitrary DP.
On the other hand, if the focused phrase within a QP is a wh-word, no viola-
tion of (141) need result. After all, our assumptions in (124) and (125) state
that wh-words lack normal semantic values. As a result, wh-words vacuously
satisfy the condition in (141). We may conclude, then, that (141) entails that
the focused phrase in a QP must be a wh-word, and so any Q-particle must
c-command a wh-word.
In summary, we have seen that our semantics for wh-words and Q-particles,
when combined with certain independently plausible principles, is able to derive the
obligatory co-occurrence and c-command relation between Q-particles and wh-
words. In as much as these properties are shared by the particles sá, ka, and da, ap-
plying our proposed semantics to each of these particles would thereby account for
the characteristic behavior of each, which lends further credence to their underlying
formal identity.
Similarly, the meaning of the wh-question in (145a) would be the set of propositions
in (145b), again understood to be the set of all its potential answers.
The lambda notation in (146b) may be informally read as “the set of propositions
p such that p is the proposition Dave ate x for some nonhuman thing x.” Intuitively,
this is indeed the set of propositions informally represented in (144b). Similarly,
the notation in (147b) may be read as “the set of propositions p such that p is the
78 THE GRAMMAR OF Q
proposition Dave read x’s book for some human x.” Again, this set is intuitively the
set of potential answers to the question in (145a).
The second syntactic hypothesis concerns the exact position of the QP within the
left periphery. Until now we have simply assumed that such QPs occupy SpecCP. How-
ever, given the assumed existence of ForceQ, we must now ask which specifier position
within the left periphery the QP occupies. Let us assume that the specifier occupied by
QP is located below the Force head. More concretely, adopting the theory of the left
periphery put forth by Rizzi (1997), we might assume that the QP of a wh-question
moves to the Spec of a FocusQ phrase, which is obligatorily selected by the ForceQ
head.
ForceP
ForceQ FocP
QP FocP
The seat of
interrogative FocQ IP
force
Given the English translation of this sentence, we want our semantic theory to assign
as its interpretation the set of propositions represented by the lambda expression in
(150b), which picks out the set of propositions p such that p is of the form ‘your
father is hunting x’, where x is nonhuman.49 Let us, then, determine whether our
semantic system can correctly assign the meaning in (150b) to sentence (150a).
Given our preceding hypotheses, (151) is the structure we assign to sentence
(150a).50
Assuming that the FocQ head here has a trivial semantic value (i.e., [λp. p]), the fol-
lowing derivation computes the meaning that our semantics assigns to (151).
As the derivation in (152) demonstrates, our semantic theory assigns to (151) the
following meaning: ‘[ λp [ ∃f. p = [ λw . [your father](w) is hunting [ f ({x<se> :
x ∉ human}) ](w) in w]]. This lambda notation picks out the set of propositions p
80 THE GRAMMAR OF Q
such that there is some choice function f such that p is the proposition ‘your father is
hunting f({ x<se> : x ∉ human })’. Thus we must now determine whether this set of
propositions is equivalent to the ‘targeted interpretation’ in (150b).
Interestingly, the equivalence between (150b) and (152h) follows from the
general equivalence proven in (135), repeated in (153).
Given this general equivalence, it follows that the formulas in (154a) and (154b) are
equivalent (relative to any value for the free propositional variable).
Given the equivalence of the formulas in (154), it follows that the formulas in (155)
are equivalent.
(155) a. λp [ ∃f. p = [λw. [your father](w) is hunting [f({ x<se> : x ∉ human })](w) in w]]
b. λp [ ∃x<se> ∉ human. p = [λw. [your father](w) is hunting x(w) in w]]
The formula in (155a) is, of course, the meaning computed in (152h) for (151), while
the formula in (155b) is simply the targeted meaning in (150b). It therefore follows
that these meanings are equivalent, and so our semantics correctly assigns the
‘targeted meaning’ in (150b) to the sentence in (150a).
Aside from the argument offered above, we can also perceive at a more intuitive
level the equivalence between (155a) and (155b). Consider any proposition p from
the set represented by (155a). By definition, there is some choice function f such that
p is the proposition ‘your father is hunting x’, where x is the value that f yields when
applied to the set of nonhumans. Given the nature of choice functions, this simply
means that p is the proposition ‘your father is hunting x’, for some nonhuman x.
Consequently, p is also in the set of propositions represented by (155b). Now, let us
consider any proposition p from the set represented by (155b). By definition, there is
some nonhuman x such that p is the proposition ‘your father is hunting x’. Again,
however, by the nature of choice functions, this entails that there is some choice
function f such that p is the proposition ‘your father is hunting x,’ where x is the value
that f yields when applied to the set of nonhumans. Thus p is also in the set repre-
sented by (155a).
We have just seen that our semantics correctly interprets those Tlingit wh-ques-
tions where Q is directly adjacent to the wh-word. Let us now consider a somewhat
more difficult structure, one where Q is not directly adjacent to the wh-word. We will
use as our representative example the wh-question in (67a), Aadóo yaagú sá ysiteen?,
‘Whose boat did you see?’ Given its English translation, we want our semantics to
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 81
Let us now determine whether our semantics can correctly assign the meaning in
(156b) to sentence (156a). First, our syntactic theory entails that (156a) has the struc-
ture in (157).
The derivation in (158) computes the meaning assigned by our semantics to the
structure in (157).
We find, then, that our semantics assigns (157) the following set of propositions as its
meaning: ‘[λp [ ∃f. p = λw. you saw [f ({ h : ∃x ∈ human . h = λw′. the boat(w′) of x(w′)
82 THE GRAMMAR OF Q
in w′ })](w) in w]]’. This rather complex expression denotes the set of propositions p
such that p is the proposition ‘you saw x’, where x is the value obtained by applying
some choice function f to the set of individual concepts of the form ‘[λw. the boat(w)
of x(w) in w]’, where ‘x(w)’ is human. Let us now determine whether this set of
propositions is indeed equivalent to the ‘targeted interpretation’ in (156b).
First, let us observe that the general equivalence in (153) again entails that these
two sets are identical. Given (153), it follows that the formulas in (159) are equivalent.
Again, however, the equivalence of the formulas in (159) entails the equivalence of
the two sets in (160).
(160) a. λp [ ∃f. p = [λw. you saw [ f ({[ λy [ λw′. the boat(w′) of y(w′) in w′ ]]}{ x : x ∈
human }) ](w) in w]]
b. λp [ ∃x<se> ∈ human. p = [λw. you saw [ λy [ λw′. the boat(w′) of y(w′) in w′ ]](x)
(w) in w]]
Finally, Function Application and Pointwise Composition entail that the formulas in
(160) are equal to those in (161).
(161) a. λp [ ∃f. p = [λw. you saw [ f ({ h : ∃x ∈ human . h = λw′. the boat(w′) of x(w′) in
w′ }) ](w) in w]]
b. λp [ ∃x ∈ human. p = [λw. you saw [ λw′. the boat(w′) of x(w′) in w′ ]](w) in w]]
Given that the formulas in (161) are simply the targeted interpretation in (156b) and the
meaning computed in (158), it follows that these two sets are equivalent, and so our
semantics correctly assigns the ‘targeted meaning’ in (156b) to the sentence in (156a).
The equivalence of the formulas in (161) can also be grasped at a more intuitive
level. Let us consider any proposition p from the set in (161a). By definition, this prop-
osition p is of the form ‘[λw. you saw y(w) in w]’, where y is the value obtained by
applying some choice function f to the set of individual concepts of the form ‘[λw. the
boat(w) of x(w) in w]’, where ‘x(w)’ is human. Given the nature of choice functions,
this entails that p is of the form ‘[λw. you saw y(w) in w]’, where y is some individual
concept of the form ‘[λw. the boat(w) of x(w) in w]’, where ‘x(w)’ is human. But this
simply means that p is of the form ‘[λw. you saw [ λw′. the boat(w′) of x(w′) in w′ ](w)
in w], for some human ‘x(w)’. Consequently, p is also a member of the set in (161b).
Now let us consider any proposition p from the set in (161b). By definition, p is
of the form ‘[λw. you saw [ λw′. the boat(w′) of x(w′) in w′ ]](w) in w]’, where ‘x(w)’
is some human. Of course, this simply means that p is of the form ‘[λw. you saw y(w)
in w]’, where y is the individual concept ‘[λw. the boat(w) of x(w) in w]’ and ‘x(w)’
is human. Consequently there is a choice function f such that p is of the form
‘[λw. you saw y(w) in w]’, where y is the value obtained by applying f to the set of
WH-FRONTING AND Q-MOVEMENT IN TLINGIT 83
individual concepts of the form ‘[λw. the boat(w) of x(w) in w]’, where ‘x(w)’ is
human. Thus p is also within the set in (161a), and we have thereby shown the equiv-
alence of these two formulas.
Having established the equivalence between the ‘targeted interpretation’ in
(156b) and the meaning computed in (158), we may conclude that our semantics
assigns the correct interpretation to those Tlingit wh-questions where Q is not di-
rectly adjacent to the wh-word. I therefore conclude that our proposed semantic
system is adequate for all (single) wh-questions in Tlingit.53
3.1 Introduction
84
APPLICATIONS TO WH-IN SITU LANGUAGES 85
This section presents and defends our proposed typology for wh-in situ languages.
Section 3.2.1 introduces the proposed distinction between ‘Q-adjunction’ and
‘Q-projection’ languages, identifying Japanese and Korean as instances of the former
and Sinhala as an instance of the latter. Section 3.2.2 demonstrates that this hypo-
thesis predicts certain features of Sinhala da. Section 3.2.3 then shows how the
hypothesis captures various features of Japanese ka and Korean ka.
QP1 CP
Complementation
XP Q
CQ IP
Agree/
… wh-word… Attract
QP1
Overt Movement
Within this structure, a Q-particle takes its sister as complement, with the result that
a QP node immediately dominates the Q-particle and its sister. Consequently, attrac-
tion of the Q-feature to the projection of the interrogative C head entails that the
entire QP projection is moved.
The reader may recall that this structure is nearly identical to Hagstrom’s (1998)
analysis of Japanese wh-questions, repeated in (2).
CP
CP Q1
IP CQ
XP Agree /
Adjunction Attract
XP Q1
Overt Movement
…wh-word…
86 THE GRAMMAR OF Q
The structures in (1) and (2) differ in only one respect: in (2), the Q does not take
its sister as complement, but rather adjoins to it. Consequently, the node immediately
dominating Q and its sister is not a QP, but is instead of the same type as the sister of
Q. Thus attraction of the Q-feature into the CP entails only that the Q-particle moves,
and the sister of Q is left in situ. Since the sister of Q contains the wh-word, such a
language is predicted to be wh-in situ.
We find, then, that (2) might accurately represent the structure of certain wh-in
situ languages. Interestingly, however, if we accept the existence of the structure in
(1) and the existence of covert movement, then there is also another imaginable
means for deriving wh-in situ structures.
CP
QP1 CP
Complementation
XP Q
CQ IP
Agree/
… wh-word… Attract
QP1
Covert Movement
Like (2), the structure in (3) differs from (1) in just one respect: in (3), the movement
of the QP is covert. Since the QP in (3) is pronounced in its base position, it follows
that the wh-word is also pronounced in its base position. Thus (3) might also repre-
sent the structure of certain wh-in situ languages.
In the following subsections, I will argue that both the structures in (2) and (3)
are attested in the world’s languages. Languages whose wh-questions have the struc-
ture in (2) will be referred to as ‘Q-adjunction languages’, while those whose
wh-questions have the structure in (3) will be labeled ‘Q-projection languages’.1 I
will present evidence that Sinhala is a Q-projection language, while Japanese and
Korean are Q-adjunction languages.
(4) Sinhala Da Cannot Appear at the Right Edge of a Matrix Clause (Kishimoto 2005)
a. Chitra monawa da gatte?
Chitra what Q buy
What did Chitra buy?
b. * Chitra monawa gatta da?
Chitra what buy Q (Kishimoto 2005: 3–4)
Given that this shared pattern was taken as evidence for a shared syntactic category,
it must receive a uniform account in both Tlingit and Sinhala.
When we look back to our account of the parallel Tlingit data, we find that our
analysis crucially relies upon the assumption that Tlingit is a Q-projection language.
Under this assumption, merger of Tlingit sá at the edge of a matrix CP would violate
the QP-Intervention Condition. Such a position for sá would entail the existence of a
QP intervening between a functional head and its complement.
F2P
QP F2
F1P Q
In this context, let us observe that if Tlingit were a Q-adjunction language, nothing
would rule out the appearance of a sá matrix-finally. After all, as the structure in (7)
illustrates, if Tingit sá did not project a QP, then a matrix-final position would not
violate the QP-Intervention Condition.
F2P
F1P F2
F1P Q
88 THE GRAMMAR OF Q
We find, then, that our account of the Tlingit data parallel to (4) ultimately
relies upon the assumption that Tlingit is a Q-projection language. Consequently,
if we wish to extend this account to the parallel Sinhala data under (4), we must
likewise assume that Sinhala is a Q-projection language. Finally, note that the
syntactic and semantic assumptions introduced in chapter 2 entail that, within wh-
questions, a Q-particle must move into the left periphery by LF. Given that the
surface form of a Sinhala wh-question places the Q-particle and the wh-word in
their base positions, we must conclude that the QP of a Sinhala wh-question moves
covertly to CP. We find, then, that when combined with our proposals from chapter
2, the facts in (4) entail that Sinhala wh-questions have the structure in (3).
Further evidence for the analysis in (3) can be found in certain additional con-
straints governing the position of Sinhala da. It appears that like Tlingit sá, Sinhala
da cannot appear (i) between a postposition and its complement, (ii) between a pos-
sessor and its possessed NP, or (iii) between a determiner and its NP complement.
This is stated in the following passage from Kishimoto (2005).
Although Kishimoto (2005) does not illustrate this generalization with negative data,
it is evident from the context that the following data pattern is intended.
We observed earlier that the inability for Tlingit sá and Sinhala da to appear matrix-
finally ultimately follows from their being Q-projection languages. Similarly, we will
90 THE GRAMMAR OF Q
now see that the ability for Q to appear matrix-finally in Japanese and Korean would
follow from their being Q-adjunction languages.
Assuming that Japanese and Korean are Q-adjunction languages, there are actu-
ally two ways in which Q could come to appear matrix-finally. First, Q could be
initially merged in this position. Recall that in Q-adjunction languages, Q does not
project a QP when it merges with its sister. Consequently, the QP-Intervention Con-
dition would not be violated if Q were to take as its sister the phrase immediately
below the interrogative ForceQ head.
ForceQP
FocP ForceQ
FocP Q
ka
… wh-word…
As we will see later in section 3.3, our semantics will interpret the structure in (14)
as a wh-question. Thus (14) represents one possible structure underlying the sen-
tences in (12) and (13).
Interestingly, though, our account does not predict that Q in these languages
must be initially merged as in (14). That is, nothing in our account prevents the merger
of Q lower down, within the IP of the wh-question. What, then, does our theory pre-
dict for this latter type of structure? The answer, of course, is already provided by the
diagram in (2). As we see there, if ever the Q is merged at a lower, clause-internal
position, then that Q subsequently undergoes movement to the periphery of the
clause. In a Q-adjunction language, such movement targets the Q-particle alone,
leaving the wh-word in its base position. Thus the matrix-final position of Q could, in
a Q-adjunction language, be derived via movement of the Q-particle from an under-
lying clause-internal position.
We therefore find that, in the wh-questions of Q-adjunction languages, Q-particles
can (and indeed must) appear at the edge of the matrix clause. Moreover, we saw in
the previous section that Q-particles cannot appear matrix-finally in the wh-questions
of Q-projection languages. Therefore, given the pattern of data in (12) and (13), we
must conclude that Japanese and Korean are Q-adjunction languages, and that their
wh-questions possess the structure in (2).3 In section 3.2.3.2, we will see that the
wh-indefinites of these languages provide further evidence for this analysis.
DP P
DP Q
…wh-word…
As seen here, in a Q-adjunction language, a Q-particle does not project the category
of the phrase minimally dominating it and its sister. Therefore, in a structure of the
form ‘[wh-word . . . Q . . . P]’, there is no QP projection intervening between the P and
the DP it selects for. Consequently, such structures do not violate the QP-Intervention
Condition, and they are predicted to be well formed. This prediction is indeed accu-
rate, as the sentences in (16) and (17) demonstrate.
Furthermore, our account in (2) predicts that Japanese and Korean wh-indefinites
should permit the Q to appear between a wh-possessor and the possessed NP. That is,
in a Q-adjunction language, nothing rules out structures like that in (18).
DP
DP D’
DP Q D NP
POSS
…wh-word… possessum
Condition, then, permits (18). Thus we predict that wh-indefinites in Japanese and
Korean should permit Q to appear between the possessor and the possessum. As the
following sentences demonstrate, this is again an accurate prediction.
Finally, note that (without additional assumptions) our theory predicts that
Q-particles in Q-adjunction languages should be able to intervene between wh-determiners
and their NP complements. After all, nothing stated thus far would rule out structures
like that in (21).
DP
D NP
D Q
wh-word
This prediction, however, is incorrect for Japanese and Korean. Even in Japanese and
Korean, a Q-particle cannot intervene between a D and its NP complement, as the
sentences in (22) and (23) illustrate.
On the other hand, our account is not necessarily inconsistent with the facts in
(22) and (23), as the impossibility of these structures may result from independent
APPLICATIONS TO WH-IN SITU LANGUAGES 93
factors. To build toward one possible explanation, note that the structure in (21) dif-
fers from those in (15) and (18) in that the Q-particle in (21) is adjoined to a head.5
Thus the D-head in (21) is initially merged with Q, rather than with the NP constitut-
ing its internal argument. Suppose, however, that selection for the internal argument
of a head H must be satisfied no later than at the point where H first externally
merges with something.6 Under this assumption, the ill-formedness of (21) would
follow. Since initial merger of D in (21) joins it with Q, and Q does not contain the
phrase selected by D, a selectional violation occurs. Therefore we find that indepen-
dent factors relating to selection may be responsible for the ill-formedness of (21).
In this section we move from the syntax of wh-in situ languages to their semantics.
We will quickly confirm that the semantic system introduced in chapter 2 correctly
interprets the wh-indefinites and wh-questions of the Q-adjunction languages. Such
confirmation will also provide a concrete illustration of how our semantics can be
applied to the structures of these languages.
2 make no reference to whether the Q-particle is adjoined to its sister or not, it fol-
lows that this semantics will also correctly interpret the wh-indefinites of the
Q-adjunction languages.
To illustrate this result, the Q-adjunction structure in (24) is interpreted accord-
ing the semantic system in chapter 2. The reader is invited to confirm that our seman-
tic system can also interpret any of the other examples of wh-indefinites in
Q-adjunction languages found throughout our discussion.
ForceQP <<st>t>
Let us now determine whether our semantics assigns the correct interpretation to
this structure. Given its English translation, we wish to derive as the meaning of (26)
the following set of propositions: those propositions of the form ‘John bought x’
where x is some nonhuman. This set, which is intuitively the set of possible answers
to question (26), may be represented via the lambda notation in (27).
We find, then, that our semantic system assigns the following set of propositions
as the interpretation of the wh-question in (26): those propositions p such that p is the
value some choice function f yields when applied to the set of propositions of the
form ‘John bought x’, where x is nonhuman. As a final step in proving the adequacy
of our semantics, let us determine whether this latter set is equivalent to that repre-
sented in (27). Indeed, such a determination is rather trivial. After all, given that
choice functions simply yield members of their arguments, the set computed in (28)
is clearly just the set of propositions p such that p is ‘John bought x’, where x is non-
human. This latter set is, of course, the one in (27).10
We have thus demonstrated that our semantics assigns the correct interpretation
to the wh-question structure in (26). Again, assuming this structure to be representa-
tive of all wh-question structures in all Q-adjunction languages, we have thereby
shown that our semantics is sufficient to correctly interpret the wh-questions of
Japanese, Korean, and all other Q-adjunction languages.
In chapter 4 we will discuss at length the potential consequences our theory has
for the analysis of LF/Focus-Intervention Effects in wh-fronting languages. In this
APPLICATIONS TO WH-IN SITU LANGUAGES 97
section we will lay the groundwork for that later discussion by introducing the theory
of LF/Focus-Intervention Effects that we will assume here. Our discussion here will
center on the relatively simple case of LF/Focus-Intervention Effects in Q-adjunction
languages such as Korean.
Let us begin by reviewing the basic facts regarding LF/Focus-Intervention Ef-
fects. In many languages, it is not possible for an in situ wh-word to be in the scope
of any of a set of ‘offending operators’. Thus the Korean sentence in (29) is ill-
formed, as the wh-word nuku-lul ‘who-ACC’ is in the scope of the offending opera-
tor–man ‘only’.
(30) Obviation of Intervention Effect via Movement of the Wh-Word (Beck 2006: 3)
[ Nuku-lul Minsu-man po-ss]-ni?
who-ACC Minsu-only see-PST-Q
Who did only Minsu see?
One way of describing the contrast between (29) and (30) is that the none of the
‘offending operators’ can come between, or intervene between, the Q-particle and
the wh-word. If such intervention occurs, the resulting sentence is ill-formed, a state
of affairs referred to as an ‘(LF- or Focus-) Intervention Effect’.
We will assume the analysis of these phenomena developed by Beck (2006).
Under that analysis, the ill-formedness of sentences like (29) follows from the prin-
ciples in (31) and (32), first introduced in chapter 2.
ForceQ FocP
FocP Q
ni
IP Foc
-man IP
ONLY
DP IP
Minsu
I VP
DP V
nuku-lul poss
However, because wh-words are assumed not to have normal-semantic values (see
chapter 2), it follows that a normal-semantic value cannot be computed for sentence
(29), in violation of principle (31). Sentence (29) is therefore uninterpretable, and
thus deviant.
On the other hand, sentences like (30) are predicted to be semantically interpret-
able. The fronting of the wh-word in (30) entails that this sentence may be assigned
the LF in (34).
Force Q FocP
FocP Q
ni
IP Foc
DP 1 IP
Nuku-lul
-man IP
ONLY
DP IP
Minsu
I VP
DP V
t1 poss
In this LF structure, the complement of–man ‘only’ does not contain a wh-word.11
Therefore the interpretation of (30) does not require one to compute the normal-semantic
value of a wh-word. Thus (30) is interpretable by our system, and is not predicted to be
deviant.
APPLICATIONS TO WH-IN SITU LANGUAGES 99
We see, then, that our semantics correctly predicts the contrast between (29)
and (30). Of course, our reasoning here generalizes to many other structural types.
Consequently, our semantics predicts that the type of structure in (35) is that which
triggers an Intervention Effect.
No Q-particle
Again, due to (32), interpreting this type of structure requires that one compute the
normal-semantic value of the sister to the ‘offending operator’. However, the absence
of a Q-particle within its sister entails that one must eventually compute the normal-
semantic value of the wh-word. Since wh-words are assumed not to have normal-
semantic values, the semantic computation therefore crashes, and deviance results.
The following quote nicely states this result of our overall semantic theory: “The
system I have introduced requires a wh-phrase to have as its first c-commanding
operator a Q operator” (Beck 2006: 16).
We have just seen how our adoption of the core hypotheses of Beck (2006) al-
lows us to employ without modification her analysis of the Intervention Effects found
in Q-adjunction languages like Korean and Japanese. In chapter 4 we will see that
when we combine these proposals with our theory of wh-fronting in (1), we can also
capture various features of Intervention Effects found in the wh-fronting languages
of the world.
4
4.1 Introduction
Chapter 2 presented evidence that the structure of Tlingit wh-questions is that in (1).
CP
QP1 CP
Complementation
XP Q
CQ IP
Agree/
… wh-word… Attract
QP1
Overt Movement
In this chapter and the next, I will argue that this ‘Q-based’ analysis should be
extended to the wh-questions of all wh-fronting languages. That is, I will argue that
there is no language whose wh-questions display the structure in (2), where the
wh-word alone is directly attracted into the left periphery.
100
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 101
CP
wh-word 1 CP
Cwh IP
Agree/
Attract
wh-word1
Overt Movement
I begin in section 4.2 with a few brief, initial arguments for the abandonment of
(2). I first raise some general typological and learning-theoretic considerations that
motivate our application of (1) to other wh-fronting languages. Following this, I
briefly discuss the Edo language (Niger–Congo; Nigeria), whose wh-questions have
a surface form that seems well suited to (1). I next consider how the Q-based analysis
in (1) can, in conjunction with the QP-Intervention Condition, predict the ill-formed-
ness of P-stranding and certain left-branch extractions across many wh-fronting lan-
guages. I also discuss the analysis of languages in which these extractions are
allowable. Finally, I briefly compare the account in (1) to three earlier, similar
accounts: Watanabe (1992), Tanaka (1998, 1999), and Sternefeld (2001).
Having presented these initial arguments, section 4.3 begins our longer discussion
of the ways in which (1) might advance our understanding of pied-piping structures.
Here, I focus on the ways in which the syntax and semantics of pied-piping structures
are simplified under the Q-based account. Indeed, as we will see, there is a sense in
which the very existence of ‘pied-piping’ (as defined in chapter 1) is called into ques-
tion. More in-depth discussion of pied-piping will come in chapter 5.
The final section of this chapter examines two widely studied parameters of
variation between the wh-fronting languages: (i) the presence of Superiority Effects,
and (ii) the presence of Intervention Effects. We will see that the Q-based account in
(1), in conjunction with our semantics from chapter 2, accurately predicts that multi-
ple wh-questions should evince Superiority if and only if they do not evince Interven-
tion Effects (Pesetsky 2000). Furthermore, we will see that our Q-based account
captures the fact that ‘pied-pipers’ are universally subject to Intervention Effects
(Sauerland & Heck 2003).
In this section I present three comparatively brief arguments for extending the
Q-based analysis in (1) to all wh-fronting languages. Following this, I discuss the
similarities and differences between the analysis in (1) and those of Watanabe (1992),
Tanaka (1998, 1999), and Sternefeld (2001).
102 THE GRAMMAR OF Q
As our discussion proceeds, two additional parameters will be added to the theory
described above: (i) whether a language permits multiple wh-questions to contain
multiple Qs (section 4.4), and (ii) whether a language requires Q and the wh-word to
undergo Agreement (section 5.2).
Edo obligatorily contain a particle dè, which Baker glosses as ‘Q’. The sentences in
(4) illustrate.
In his study of these structures, Baker (1999) first considers the hypothesis that
dè is some kind of wh-determiner. Interestingly, the evidence he puts forth against
this hypothesis concerns the inability for dè to appear inside of PPs and DPs. As
Baker notes, wh-determiners in English and other languages can generally appear in
such structures, where they pied-pipe the larger phrase in which they are contained.
In Edo, however, the particle dè cannot appear in either of these environments. As the
following datum illustrates, if dè ever intervenes between a possessor and a pos-
sessed NP, the sentence is ill-formed.
(6) Edo Dè Cannot Appear Between Wh-Possessor and Possessed NP (cf. (4c))
* [Èbé [ dè òmwán ] ] nè Úyì dé ?
book Q whose that Uyi buy (Baker 1999)
Similarly, if dè ever appears between a P and its DP complement, the sentence is ill-
formed.
From these facts, Baker concludes that dè should not be analyzed as a wh-determiner
like the English which.
Importantly, our Q-based theory of wh-fronting can provide an account of these
facts, if we assume that Edo dè is a Q-particle. Because Edo is a wh-fronting language,
our analysis entails that it must be a Q-projection language, where the Q-particle takes
its sister as complement and projects a QP. Given our QP-Intervention Condition, we
predict that Q-particles in Edo will be unable to appear between (i) wh-possessors and
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 105
their possessed NPs and (ii) prepositions and their DP complements. Consequently,
if we suppose that dè is the overt realization of Q in Edo, then we predict the data in
(6) and (7).4 Furthermore, let us also observe that Edo wh-questions contain a sepa-
rate, overt realization of the interrogative C head; each of the sentences above con-
tains the element nè, which generally functions as a complementizer in the language.
We must conclude, then, that not only does dè receive an interesting analysis as Q, it
cannot receive an analysis as the interrogative C.
We find, then, that the special behavior of dè in Edo suggests that it is a ‘Q’ in
our sense of the term.5 Thus Edo provides us one other case where the QP structure
in (1) is overtly pronounced in a wh-fronting language.6 Moreover, as we have just
seen, Edo also demonstrates that it is possible for both the Q-particle and the inter-
rogative C to be overtly pronounced in the same language. Thus these structures
provide independent evidence for the distinction between Q-particles and interroga-
tive Cs.
The surface form of Edo wh-questions confirms that the structure in (1) is not
limited to the Tlingit language. This fact, in turn, bolsters the view that (1) is a struc-
ture licensed by UG, and so supports the application of (1) to the wh-questions of
Tlingit. Furthermore, the fact that the structure in (1) is transparently found in two
unrelated languages on separate continents suggests that it is not an ‘aberration’ spe-
cific to a single language family or area, but might instead have a broader distribution
in the languages of the world. This latter conclusion, in turn, renders more plausible
the notion that (1) is, in fact, common to all the world’s wh-fronting languages.
Given the obvious similarity between the facts in (8)–(10) and those in
(11)–(13), we noted in chapter 2 that a unified account for these facts would be opti-
mal. Under such an account, the ill-formedness of the Tlingit sentences in (8)–(10)
would follow as one instance of the more general cross-linguistic pattern seen in
(11)–(13). Unfortunately we found that certain additional facts from Tlingit seem to
prevent such an account. To recall, the predominant view regarding the cross-linguis-
tic pattern in (11)–(13) is that it reflects properties of movement. We saw, however,
that such a ‘movement-based analysis’ cannot be extended to the Tlingit data in
(8)–(10). Rather, a broader examination of Tlingit syntax reveals that the ill-formed-
ness of (8)–(10) follows from constraints governing the position of Q, namely the
QP-Intervention Condition, and not from any constraints on movement itself.
It follows that, under the standard view that (11)–(13) reflect constraints on
movement, a unified account for all the data in (8)–(13) is not possible. This, of
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 107
We therefore find that the parallelism between the Tlingit data in (8)–(10) and the
cross-linguistic patterns in (11)–(13) motivate the view that all wh-fronting structures
receive the Q-based analysis in (1). Under this analysis, the ill-formedness of (11)–
(13) is not explained in terms of the islandhood of the base position of the wh-word.
Rather, the apparent islandhood of those positions is explained in terms of indepen-
dently visible constraints on the placement of Q. Consequently, the empirical motiva-
tion for classifying those positions as syntactic islands is weakened.7 Indeed, within
the Q-based account, it seems best not to view those positions as syntactic islands at
all.8 Furthermore, in rejecting the status of these positions as islands, we must thereby
108 THE GRAMMAR OF Q
also reject any syntactic principles whose goal is to predict the islandhood of these
positions. Therefore I conclude that our Q-based theory of wh-fronting challenges
such principles as the ‘Left Branch Condition (LBC)’ (Ross 1967; Corver 1990,
2007).
Although our Q-based analysis of (11)–(13) is motivated on conceptual
grounds, it might at this stage be criticized for a degree of empirical incomplete-
ness. After all, our discussion has so far ignored a rather well-known fact about
the extractions in (11)–(13), a fact that is moreover one of the key data that the-
ories of (11)–(13) seek to capture: in some languages, these extractions are not
ill-formed. For example, it is well known that English wh-fronting may freely
strand prepositions.
Relatedly, the Slavic languages are also well known to permit extraction of wh-deter-
miners.
This general approach to the data in (17)–(19) can also be adopted within a
Q-based account. In fact, our Q-based account can adopt, without much modifica-
tion, certain already established analyses of (17)–(19).
To see this, let us first consider possessor and determiner extraction in Slavic
languages ((18) and (19)). It has been widely noted that languages permitting posses-
sor and determiner extraction generally permit NPs to appear without overt deter-
miners (Uriagereka 1988; Corver 1992; Bošković 2005a). Many scholars have
therefore sought to connect these two properties, to derive the permissibility of those
extractions from the ability for NPs not to be dominated by DP projections. For ex-
ample, Bošković (2005a) proposes that extraction of possessors and determiners in
these languages is due to their being pure NP adjuncts. That is, instead of the familiar
DP structure, the structure of possessives and determiner-marked nominals in these
languages is as follows.
NP
Possessor NP
N
Possessum
NP
Determiner NP
NP
QP NP
XP Q
N
Wh-Possessor Possessum
Similarly, note that languages possessing the structure in (21) should generally
allow Q to intervene between a determiner and an NP. Since determiners are simply
adjuncts to NP in these languages, the QP-Intervention Condition would allow for
structures like that in (24).
NP
QP NP
Wh-Determiner Q N
in languages like English (see (17))? Are existing perspectives on the nature of
P-stranding consistent with the Q-based approach?
In a sense, they are. Within the literature on P-stranding, it is commonly held
that the possibility of P-stranding in English reflects some special, sui generis prop-
erty of English Ps. For example, Abels (2003) proposes that the general impossibility
of P-stranding follows from the status of P as a ‘phase head’ (Chomsky 2000). Con-
sequently Abels claims that the possibility of P-stranding in English indicates that
English P is, somehow, exceptionally not a phase head (Abels 2003: 233). Similarly,
van Riemsdijk (1978) claims that the general impossibility of P-stranding follows
from the unavailability of SpecPP as a ‘landing site’ for movement. Consequently
the possibility of P-stranding in English must follow from an exceptional ability for
SpecPP in English to be an available landing site.
Throughout our discussion, I will take a similar view regarding the nature of
English P-stranding. As in prior accounts, I will assume that English P-stranding
follows from an exceptional property of English prepositions, one that (unfortu-
nately) plays little role in the broader grammar of English. I will claim, however, that
this special property is that of being a lexical category. First, note that if P in English
is a lexical category, then English should generally allow Q to intervene between P
and its DP complement. Due to the status of P as a lexical head, such structures will
no longer violate the QP-Intervention Condition. Of course, if such structures were
to feed QP-movement, we would thereby derive P-stranding in English. The follow-
ing structure illustrates.
CP
QP1 C’
DP Q CQ IP
…wh-word…
PP Possible PP,
Permitted by
QP P QP-Intervention Condition
Because ‘P’ in English Is a
t1 Lexical Category
account derives the possibility of English P-stranding from English P being a lexical
head. While the stipulative nature of this account could be justly criticized, I submit
that it is no more stipulative than any other currently available approach to the
distribution of P-stranding across languages.
Finally, it is perhaps worth noting, if only in passing, that Huddleston and Pullum
(2002: 597–661) argue that the category P in English is much larger than tradition-
ally conceived, and contains many items that have previously been miscategorized as
adverbs, adjectives, or verbal participles. Consequently, Huddleston and Pullum
(2002: 603) distinguish P in English as a “relatively closed class” (emphasis mine),
since it differs markedly from other functional categories (like determiners) in both
the sheer number of its members, and the relative rapidity with which it gains mem-
bers from the lexical classes N, V, and A. Thus, there are some independent grounds
for distinguishing P in English from clear-cut cases of functional heads, though it
remains to be seen whether these properties also distinguish P in any languages that
disallow P-stranding.
The similarities between the core proposals of Tanaka (1998, 1999) and our own
Q-based theory are rather obvious. Where Tanaka’s work differs from the system
offered here is chiefly in the use to which the analysis is put. Tanaka (1998, 1999) is
primarily concerned with the interpretation and linear order of wh-words in Japanese.
Consequently, there is comparatively little development in this work of the ‘QP/ωP-
analysis’ of overt wh-fronting. Thus, although the overall proposals of our work are
very close, the empirical results obtained are entirely complementary.
Another proposal related to our own is that developed by Watanabe (1992).
Although substantially different in its details, Watanabe’s account shares a certain
logical structure with our own. Indeed, one could view the proposals put forth here
as one development of these common core ideas. In brief, Watanabe (1992) proposes
the following. In both English and Japanese wh-questions, there is overt fronting of
an entity that Watanabe labels ‘Op’ (short for ‘operator’). In Japanese, this ‘Op’ ele-
ment can be moved away from the wh-word, leaving the wh-word in situ. In English,
however, ‘Op’ cannot be detached from the wh-word. Consequently, overt movement
of ‘Op’ in English entails overt movement of the wh-word. Thus the wh-questions of
Japanese and English both involve overt movement of the same formal element, and
their surface differences result from a more basic difference in whether that moved
element can detach from its associated wh-word.
Our Q-based account is thus quite similar to the proposals of Watanabe (1992).
There are, however, several details of implementation that render Watanabe (1992)
rather different from the Q-based analysis advocated here. The chief differences lie
in the theory of wh-fronting languages. Unlike our account in (1), where the Q/Op-
element heads a projection containing the wh-phrase, Watanabe (1992) proposes that
the Q/Op-element is a specifier of the wh-phrase. This is illustrated in (27).
DP
Op/Q D’
D
who
element and the wh-word in a wh-fronting language. Finally, given the assumption
that ‘feature percolation’ is constrained by syntactic islands, it follows that there can
be no pied-piping of islands in any wh-fronting language. This result is summarized in
the following quote: “The prediction of our system is that languages like English
which must move an entire wh-phrase at S-structure will never exhibit large-scale
pied-piping, while languages like Japanese which move pure wh-operators and allow
wh-in-situ are potentially able to employ large-scale pied-piping.” (Watanabe 1992:
65). Although it is indeed true that English does not permit the large-scale pied-piping
of islands (see chapter 5), it is not true that no wh-fronting language permits such pied-
piping. As we have already seen, such pied-piping is readily available in Tlingit.
We find, then, that the assumption in (27), that Q/Op is a specifier of the wh-
phrase, eventually leads the system of Watanabe (1992) to be quite different from
that proposed here. As a result of that assumption, Watanabe (1992) is unable to
adopt our approach to pied-piping structures, and wrongly predicts the nonexistence
of languages like Tlingit. Consequently, the core advantage of our Q-based account
over Watanabe’s is the notion (shared with Tanaka (1998, 1999)) that the wh-phrase
is complement to the Q/Op/ω-head.
Finally, before we leave this discussion, I would like to briefly compare our
Q-based account in (1) to a more general idea that has been ‘in the air’ for some time.
In the syntactic literature on wh-questions, one often encounters a view that can
(vaguely) be stated as follows: the element pronounced as ka in Japanese wh-questions
surfaces in English as the ‘wh-morpheme’ in the wh-word. For example, Watanabe
(1992) suggests that the Op-element is in English overtly pronounced as the /wh-/
component of the wh-word, which accounts for the inability of English Op to ever
move away from the wh-word. This same basic perspective also appears in work by
Cole and Hermon (1998), although it receives a rather different implementation. I
might also add that several individuals have suggested to me that, rather than analyze
English as containing a null version of Tlingit sá, it may be that this ‘Q’ element
receives overt pronunciation as ‘wh-’.
Although the general idea that ‘Q is wh-’ is an interesting one, in the context of
the proposals put forth here, it cannot be made to work. The basic issue concerns the
role played by the Q-particle in the theory of pied-piping structures that we develop
in the following sections. As alluded to in chapter 1, we will assume that the Q-par-
ticle in English plays the same role in pied-piping structures that the Q-particle sá
does in Tlingit. As we will see in section 4.3, this entails that in English pied-piping
structures, the Q-particle sits just above the fronted phrase, separated from the wh-
word by phrasal material. Such a position for Q, detached from the wh-word, is
clearly at odds with the idea that Q is the wh-morpheme.
On the other hand, a proponent of the view that ‘Q is wh-’ might address this
issue by appealing to covert movement of Q. That is, it might be possible within a
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 115
‘single-output architecture’ (Bobaljik 2002) to adopt the following view: (i) the
Q-particle is initially merged in a low position, adjacent to the wh-word; (ii) the
Q-particle then undergoes movement to a higher position, where it triggers the
observed ‘pied-piping’; (iii) the pronunciation of the Q-particle adjacent to the wh-
word (as the wh-morpheme) simply reflects pronunciation of its lowest copy.9 Such
a proposal, however, immediately invites the following question: what is the mech-
anism underlying the hypothesized Q-movement? If the mechanism is Agreement
with some higher head X, then why not simply view the higher head X as the cor-
relate of Q in English? Indeed, chapter 5 presents an analysis rather similar to this,
within which the wh-word of English undergoes syntactic Agreement with the
(higher) Q-particle.
Under our approach, then, there could be a sense in which the wh-morpheme
of English is an instance of Q. Since Q undergoes Agreement with the wh-word in
English, it might be possible to view the ‘wh-’ submorpheme as some reflex of this
Agreement, possibly an uninterpretable instance of the ‘Q-feature’ on the interroga-
tive word. Indeed, Cable (2007) develops this view in greater detail. While we will
not pursue it in this work, it is nevertheless apparent that, in a certain sense, the
Q-based theory proposed here could adopt one implementation of the oft-encountered
notion that ‘English /wh-/ is Japanese ka.’
So far I have presented two general arguments for abandoning the ‘classic’ account
in (2) and adopting the Q-based account in (1) for all wh-fronting languages: (i) doing
so would greatly simplify our typological theory, and thus the hypothesis space
learners must navigate, (ii) doing so is the only way to provide a unified account of
the impossibility of P-stranding and left-branch extractions across languages.
The remainder of this book presents two further, extended arguments in sup-
port of our Q-based account. The overall thrust of these arguments is that apply-
ing (1) to some of the best-studied wh-fronting languages advances understanding
of various phenomena in those languages. The first of these two arguments,
which concerns pied-piping, will begin in this section and will continue into
chapter 5.
I will first argue that our Q-based account can advance understanding of the
syntax of pied-piping structures, and even allows us to eliminate the concept of ‘pied-
piping’ from the theory of grammar. Following this, I will briefly note that our
Q-based account might also advance our theory of the semantics of pied-piping
structures, as it provides an especially simple system for their interpretation.
To recall, the problematic property exhibited by these sentences is that the fronted
phrase is not a projection of the wh-word, but rather properly contains the projections
of the wh-word. Consequently, if we assume, as in (2), that wh-fronting targets fea-
tures of the wh-word itself, then structures such as those in (29) should not exist.
It was as a solution to this basic, empirical problem that the concept of ‘pied-
piping’ was introduced into linguistic theory. Recall that throughout our discussion
we understand the term ‘pied-piping’ to refer to a particular kind of analysis, defined
as follows.
Thus the term ‘pied-piping’ contrasts in meaning with the purely descriptive
label ‘pied-piping structure’, which we use to refer to a particular class of surface
structures.
Following these definitions, there is a clear sense in which one can deny or assert the
existence of pied-piping: to assert/deny the existence of pied-piping is simply to as-
sert/deny that there exist cases of the kind in (30). Furthermore, we find that the clas-
sic analysis in (2) must, in light of the facts in (29), accept that pied-piping exists.
In this context, let us observe that our theory of Tlingit wh-questions has no need
of the concept of ‘pied-piping’ in (30). Moreover, and most importantly, this is despite
the fact that Tlingit clearly possesses pied-piping structures, as defined in (31). That
is, as we saw in chapter 2, and as illustrated below, it is possible in Tlingit for the
fronted phrase of a wh-question to properly contain the maximal projection of the
wh-word.
The structures in (32) clearly satisfy the definition in (31) of a ‘pied-piping structure’,
and their similarity to English pied-piping structures like (29) is rather obvious.
But, in what sense does our theory of Tlingit avoid appeal to (30) in the analysis
of (32)? First, recall that in a Tlingit wh-question, the particle sá must always occur
directly to the right of the fronted phrase, as shown in (33).
Thus the fronted phrase of a Tlingit wh-question never properly contains the QP.
Now, according to our analysis in (1), it is the features of the Q—and not the
wh-word—that trigger movement in Tlingit wh-questions. Under our analysis, then,
the pied-piping structures in (32) are all straightforward cases of simple phrasal
movement. As (34) illustrates for (32b), in each structure, movement simply targets
the projection of the head whose features trigger the movement (namely, Q).
CP
QP1 CP
DP Q CQ IP
sá
Aadóo yaagú ysiteen QP1
Probe/Agree for Q
Attract / Move
Thus none of the structures in (32) are cases where the moved phrase properly contains
the projections of the head triggering the movement. Consequently none of the Tlingit
pied-piping structures in (32) are instances of true ‘pied-piping’, as defined in (30).
118 THE GRAMMAR OF Q
Furthermore, given the generalization illustrated in (33), we find that there are
not any true cases of pied-piping in Tlingit. Following our definition in (30), true
pied-piping in Tlingit wh-questions would have the appearance of the structures in
(33), where the projections of the particle sá (the target of movement) are properly
contained within the fronted phrase. To clarify the point, such hypothetical structures
are illustrated in (35).
CP
DP1 CP
QP DP CQ IP
Thus, under our Q-based account in (1), the pied-piping structures of Tlingit are not
actually cases of pied-piping, and what would be actual cases of pied-piping in the
language are ill-formed. For this reason, the concept of “pied-piping” as defined in (30)
can be eliminated from our theory of Tlingit grammar. By adopting the analysis in (1),
we need not deviate from the null hypothesis that if an operation (in Tlingit) targets the
features of a lexical item, then it applies only to the projections of that item.
Similarly, if we extend the analysis in (1) to all wh-fronting languages, we need
not ever deviate from that null hypothesis. The pied-piping structures of all the best-
studied wh-fronting languages could receive an analysis akin to that shown for English
in (36).
Under this analysis, a pied-piping structure in English is derived exactly like the
pied-piping structures of Tlingit. In such sentences, the (null) Q-particle takes as
sister a phrase properly containing the wh-word, which entails that the fronted phrase
of the wh-question properly contains the wh-word. Thus one need not view sen-
tences like (29) as cases where movement targets ‘more than’ the phrase whose fea-
tures trigger it. Under such a Q-based theory, then, one need not ever accept that
pied-piping (as defined in (30)) truly exists in English.10 If this same basic result
can be maintained for all wh-fronting languages, then one need not ever admit the
existence of pied-piping in human language.
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 119
[wh]DP
Movement /
Percolation of
Wh-feature
From [wh]DP D’
SpecDP
wh-possessor POSS possessum
120 THE GRAMMAR OF Q
However, as we saw from data like (38), English does not generally permit movement
from SpecDP.13
Furthermore, it was noted earlier that many other wh-fronting languages share with
English the pattern in (38): they permit pied-piping by possessors, but do not gener-
ally permit extraction of possessors. Consequently the notion that pied-piping struc-
tures are generated by feature-movement, as in (37), is in conflict with the tendency
against movement from SpecDP.
It therefore appears that the putative operation of feature percolation cannot be
reduced either to Agreement or to movement. In the absence of other proposals, we
must conclude that any theory appealing to ‘feature percolation’ necessarily invokes
an additional, primitive syntactic operation. Most worrying of all, however, is that
such a primitive feature percolation mechanism appears to have little purpose outside
of deriving pied-piping structures. Given these considerations, theories of pied-
piping that appeal to ‘feature percolation’ do seem rather suspicious.
What, then, of those theories that adopt (2) but do not appeal to feature percola-
tion? Under these approaches, the theory of movement is weakened so as to permit
structures where the moved phrase does not itself bear the feature triggering the
movement (Ross 1967; Heck 2008).14 Thus, according to such analyses, sentence
(29a) is permissible because it satisfies the (weak) condition requiring that the moved
phrase contain the wh-word somewhere inside it. Unfortunately, a pervasive problem
for this form of analysis is the potential for overgeneration. That is, it is not generally
the case that any phrase containing a wh-word may be fronted in an English
wh-question, as the ill-formedness of (39b) illustrates.15
Of course, if appeal is made to feature percolation, then the observed limits on pied-
piping may be encoded into the percolation mechanism itself, by placing limits on
how ‘far’ feature percolation may carry a feature from its lexically associated head.
However, without this sort of mechanism, it is difficult to identify the source of
anomaly in sentences like (39b), especially since sentences like (39a) show that the
embedded wh-words are, in principle, accessible to the matrix interrogative C.
In summary, then, we find that the classic analysis in (2) is not only saddled with
the additional concept of ‘pied-piping’, but also requires special mechanisms to gen-
erate pied-piping structures. To put the matter more acutely, under the classic picture
in (2), the pied-piping structures in (29) represent a deviation from the expected
grammatical pattern. That is, on its own, (2) predicts that structures like (29) should
not exist, as it can only account for such structures via the addition of special mech-
anisms. This is in stark contrast to the Q-based analysis in (1). Under that analysis,
pied-piping structures merely represent cases where the sister of Q is not the
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 121
wh-phrase itself, but rather a phrase containing the wh-phrase. Importantly, the possi-
bility of such configurations requires no special assumptions or additions to our basic
theory. Indeed, our theory would require special assumptions to rule out pied-piping
structures.
Consequently our Q-based analysis permits us to dispense with the special con-
cept of ‘pied-piping’ in (30), as pied-piping structures can be generated via normal
phrasal movement of the QP. Indeed, within our Q-based account, the only distinc-
tive property of pied-piping structures is that the Q is not directly adjacent to the
wh-phrase. Such structures are, again, immediately predicted by our account; they
have no interesting syntactic or semantic properties, and do not stand out as a special
taxonomic class. Thus, in a very real sense, our Q-based account entails that there is
no such thing as pied-piping.
In addition, we will see that our Q-based theory makes a range of further predictions
regarding pied-piping structures. In this chapter, we will see that our analysis predicts
that ‘pied-pipers’ are universally subject to Intervention Effects. Moreover, chapter 5
argues that our account may capture several other properties of pied-piping structures,
including the rather severe limits placed on pied-piping by languages like English.
Thus we find that extending the Q-based analysis to all wh-fronting languages—
and thereby doing away with the concept of ‘pied-piping’ in (30)—would seem to
yield a simpler overall theory of phrasal movement.
(Hamblin 1958; Groenendijk & Stokhoff 1982, 1984). What unites all these accounts is
the notion that the wh-word is of a simple nonquantificational semantic type. That is, as
noted by Cable (2007), the special mechanisms developed by von Stechow (1996) and
Sharvit (1998) are required because they assume (following much earlier work) that wh-
words are of a higher quantificational type. Under such an assumption, a wh-question can
only be interpreted if the wh-word has scope over the entire IP of the clause. In a pied-
piping structure, however, the wh-word is buried within the fronted phrase, and thus it
does not at its surface position have the proper scope. Consequently, special mechanisms
are required, either to alter the position of the wh-word prior to interpretation (von
Stechow 1996), or to permit it to take exceptional scope from its surface position (Sharvit
1998).
Of course, if we assume that wh-words are not of a quantificational type, then
they are free from the need to take scope over the IP. Consequently, they can be inter-
preted in their surface position, deep down within the fronted phrase. Thus a seman-
tics that assumes wh-phrases to be of type e, such as our Q-based account, can provide
a rather straightforward semantics for pied-piping structures, one in which their
meaning directly follows from the rules required for simple wh-questions without
pied-piping.
In this section we will see how our Q-based account in (1) can capture the cross-
linguistic relationship between Intervention Effects and Superiority Effects in the
multiple wh-questions of wh-fronting languages. Inasmuch as the account offered
here seems like a promising direction, we find further support for the notion that the
structure in (1) is common to all wh-fronting languages.
We will begin by briefly previewing the phenomena of interest. We saw in chap-
ter 3 that in many languages, an in situ wh-word cannot appear in the scope of certain
‘offending operators’. Although we originally discussed such ‘(LF/Focus) Interven-
tion Effects’ in the context of wh-in situ languages, they can also be observed in
some wh-fronting languages. For example, the well-formed German multiple wh-
question in (40a) becomes ill-formed if the in situ wh-word is placed in the scope of
the offending operator niemanden ‘nobody’, as in (40b).
What, then, accounts for this interesting contrast between English and German?
As noted by prior scholars (Beck 1996; Pesetsky 2000), our theory of Intervention
Effects should be able to derive it from some broader, more fundamental difference
between the two languages.
In the following discussion we will see how our Q-based account might accom-
plish this. Moreover, we will see that our account ties this contrast in Intervention
Effects to another well-known contrast between English and German: Superiority
Effects. As is well known, multiple wh-questions in English typically require that
only the structurally highest wh-word undergo overt wh-fronting; lower wh-words
must stay in situ.18
Adapting the terminology of Pesetsky (2000), we may say that English wh-questions
exhibit ‘Superiority Effects’. This is in contrast to the multiple wh-questions of
German, as seen in (43).
Unlike the English pattern in (42), German multiple wh-questions permit overt front-
ing of the structurally lower wh-word. Thus it may be said that German does not
exhibit Superiority Effects.19
Taken together, then, we find that English multiple wh-questions exhibit Superi-
ority Effects but fail to exhibit Intervention Effects, while those of German exhibit
Intervention Effects but fail to exhibit Superiority Effects. Following earlier work
(Beck 1996; Pesetsky 2000), we will take up the view that this is no accident.20 We
will see that our Q-based account is able to tie the presence of Superiority Effects in
multiple wh-questions with the absence of Intervention Effects, and vice versa.
In addition, our Q-based account of (40)–(43) will be found to make the following
prediction regarding pied-piping structures.
124 THE GRAMMAR OF Q
That is, our account predicts that internal to the fronted phrase of a wh-question, a
wh-word is always subject to Intervention Effects. This generalization has
been independently shown by Sauerland and Heck (2003) to hold for German
wh-questions. The data in (45) illustrate.
We will also see that this generalization holds for pied-piping in English, as shown
by data like (46).
It is important to observe here that the Intervention Effects associated with ‘pied-
piping’ wh-words are found in both German and English. This is, of course, quite
unlike the Intervention Effects associated with in situ wh-words. We will see that our
Q-based account of (45) and (46) indeed predicts this contrast between these two
types of ‘intervention environments’.
In the following subsection, we begin to develop our account of these facts. We
start with the multiple wh-questions of English.
According to the semantics in (47), this ForceQ head contributes exactly one existen-
tial quantifier to the meaning of the wh-question. As we have seen, this is indeed
correct for simple, single wh-questions. For multiple wh-questions, however, this
will be insufficient.
To see this, let us consider the meaning of a multiple wh-question like ‘Who
saw what?’ Intuitively, this question is answered by propositions like Dave saw
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 125
the table, Frank saw the picture, etc. That is, an answer must name a particular
person x and a particular thing y such that x saw y. Consequently, the meaning of
this wh-question is the set of propositions p such that there is some human x and
some nonhuman y such that p is the proposition ‘x saw y’. These facts are summa-
rized in (48).
Thus ForceQ2 differs from ForceQ in that it contributes two existential quantifiers to
the meaning of the question. We will see that such a semantics allows our system to
correctly interpret sentences like (48a).
However, before we extol the virtues of this ‘ForceQ2’, let us first acknowledge
one serious criticism against it. Given the semantics in (49), this head contributes
only two existential quantifiers. Consequently it will not be enough to interpret,
for example, ternary multiple wh-questions, which contain three wh-words.
As illustrated in (50), the meanings of such questions contain three existential
quantifiers.
Although this is a serious issue, I will put off discussion of ternary wh-questions until
the next subsection. We will see then that a minor change to our account will not only
allow it to interpret such questions, but will also predict a certain curious property
they exhibit.
126 THE GRAMMAR OF Q
Let us, then, examine how (49) permits our system to interpret multiple wh-
questions. We assume that (binary) multiple wh-questions are headed by ForceQ2, as
illustrated in (51).21
As the reader is invited to confirm, the lexical entry for ForceQ2 in (49) entails that our
semantic system assigns as the meaning of (48a) the set of propositions indicated
above. This is the set of propositions p of the form ‘x saw y’, where (i) x is the value
of some choice function f over the set of humans, and (ii) y is the value of some
choice function h over the set of nonhumans. Given the nature of choice functions,
this is simply the set of propositions p of the form ‘x saw y’, where x is some human
and y is some nonhuman. Therefore our semantics correctly assigns to (48a) the
meaning in (48c).22
We see, then, that the addition of (49) allows our system to correctly interpret
the binary wh-questions of English. Interestingly, we will also see that it predicts
the crucial data in (41) and (42). How so? Ultimately the data in (41) and (42) fol-
low from a more abstract prediction of our system: in an English binary wh-
question, there are multiple Q-particles, one for each wh-word. This latter prediction
follows straightforwardly from our semantics in (49). Since the ForceQ2 head
existentially quantifies over two choice-function variables, its complement must
contain two Q-particles; otherwise, the structure will violate a general ban on
vacuous quantification.
So how does the presence of multiple Q-particles in English binary wh-questions
capture the data in (41) and (42)? Let us first consider the Superiority Effects in (42).
The following diagram summarizes the core of our analysis.
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 127
Superiority-Satisfying Structure:
[ [QP Who Q ]1 [ t1 bought [QP what Q ] ]
Superiority-Violating Structure
* [ [QP What Q ]1 [ did [QP who Q ] buy t1 ]
Attract Closest Satisfied
Under our Q-based account, the fronting of a wh-word is ultimately fronting of the
QP projection dominating that wh-word. Such Q-fronting is, of course, assumed to
follow general syntactic principles like ‘Attract-Closest’ (Richards 1997) or the
‘Minimal Link Condition (MLC)’ (Chomsky 1995). Consequently, if there are mul-
tiple QPs within a sentence, Attract-Closest/MLC will require that the structurally
highest QP undergo Q-movement prior to any lower QPs. Thus only the structurally
highest QP will undergo overt movement in English, while any lower QPs must un-
dergo subsequent, covert movement. Given that each wh-word in the sentence is
dominated by a QP, it follows that the structurally highest QP will dominate the
structurally highest wh-word. Therefore only the structurally highest wh-word will
appear overtly fronted in an English binary wh-question.
Let us now consider the Intervention Effect data in (41). First, note that our the-
ory assumes that all the QPs in a wh-question move into the left periphery by LF (see
(51) and note 21). Consequently, by the time an English binary wh-question is inter-
preted, all the overtly in situ wh-words have been moved above any ‘offending oper-
ators’ that might c-command them on the surface. Thus, by the time they are
interpreted, sentences like those in (41) do not contain any wh-words within the
scope of any ‘offending operators’, and so they are predicted to be well-formed.
Suppose, though, that we were to give up the assumption that in situ QPs move
covertly. Interestingly, our account would still predict the data in (41). To see this, let
us assume that QPs that are in situ at PF remain in situ at LF. The LF structure of a
sentence like (41a) would then be as follows.
(52) The LF Structure of (41a), Assuming That In Situ QP Does Not Covertly Move
[ [QP Who Q ]1 [ didn’t t1 read [QP what Q ] ] ]
While the wh-word in (52) is indeed within the scope of the ‘offending operator’
(didn’t), recall from chapter 3 that Intervention Effects are only triggered when sen-
tences contain structures of the type in (53).
no Q-particle
128 THE GRAMMAR OF Q
That is, Intervention Effects result when the first focus-sensitive operator c-
commanding a wh-word is not the Q-particle. Such a structure clearly does not occur
in (52). Because the in situ wh-word is paired with its own Q-particle, the first focus-
sensitive operator c-commanding it need not be the ‘offending operator’. Conse-
quently these structures are predicted not to trigger Intervention Effects. We find,
then, that our system’s prediction of the data in (41) is rather robust; it rests entirely
upon the semantics assigned to ForceQ2, and persists through minor alterations in the
surrounding syntactic theory.
In summary, we have seen that the addition of ForceQ2 to the lexicon of English has
the following consequences: (i) it provides a compositional semantics for binary wh-
questions; (ii) it predicts that English binary wh-questions should exhibit Superiority
Effects (42); and (iii) it predicts that the in situ wh-words of binary wh-questions should
be immune to Intervention Effects (41). Thus it is possible in our Q-based account to
derive the English data in (41) and (42) from a single core property of the language.
It appears, then, that languages may differ in how many Q-particles multiple wh-
questions are permitted to have. Given the central role that this property plays in our
analysis of English, the variation seen in (54) and (55) invites the following hypothesis
regarding German. Perhaps the variation between English and German seen in (40)–(43)
is ultimately that observed between Tlingit and Navajo in (54) and (55). That is, perhaps
German, like Navajo, does not permit multiple wh-questions to have multiple Qs.
As we will see, such an analysis of German does indeed predict the data in (40)
and (43). Before we come to this, however, let us first determine how our Q-based
theory is to analyze languages where multiple Qs are not possible. Recall that the
English ForceQ2 head requires a wh-question to contain multiple instances of Q. It
follows, of course, that this ForceQ2 head must be absent from those languages where
multiple wh-questions cannot have multiple Qs. For these latter languages, then, let
us assume that their lexicons contain a separate Force head, one that is, in turn,
absent from languages like English and Tlingit. This Force head, which we will label
‘ForceQ+’ is assumed to have the following semantics.
Thus ForceQ+ shares with ForceQ2 the property of introducing multiple existential
quantifiers into the meaning of the wh-question. Unlike ForceQ2, however, the mean-
ing of ForceQ+ already contains the choice function variable bound by one of those
quantifiers. Consequently this ForceQ+ head can bind only one Q-particle. Therefore,
even though the complement of ForceQ+ may have multiple wh-words, only one of
those wh-words can be dominated by a QP.23
Let us then examine the semantics that (56) yields for German multiple wh-
questions. We assume that the multiple wh-questions of German are headed by
‘ForceQ+’, as illustrated in (57).
The reader is invited to confirm that our semantics in (56) yields as the meaning of
(43a) the set of propositions indicated in (57). This is the set of propositions p such
that there is some choice function h that yields p when given the following set of
propositions: those propositions q of the form ‘x bought y’, where y is some nonhu-
man and x is the value of some choice function f over the set of humans. Given the
nature of choice functions, this is simply the set of all those propositions p of the
form ‘x bought y’, where x is some human and y is some nonhuman. Thus the formula
derived in (57) is equivalent to that in (58), which quite obviously represents the set
of possible answers to (43a). Thus we find that our semantics in (56) and (57) cor-
rectly interprets German multiple wh-questions.
We find, then, that the addition of the ForceQ+ head in (56) allows our system to
correctly interpret German multiple wh-questions.24 We will now see that it also pre-
dicts the data in (40) and (43). Let us begin by considering the Superiority data in
(43). The following diagram sketches the central ideas.
Recall, again, that under our Q-based account, the fronting of a wh-word is ultimately
fronting of a QP projection dominating it. It follows that if a wh-word is not domi-
nated by a QP, then it will not undergo any fronting. More concretely, if a wh-word is
not dominated by a QP, then its position will fail to contain a ‘goal’ for probing by
interrogative C. Now, via our addition of (56), German multiple wh-questions only
contain a single QP. Under the simplest assumptions, this single QP is free to domi-
nate any wh-word in the sentence. Therefore, while this QP could dominate the struc-
turally highest wh-word (59a), it could also dominate any of the lower wh-words
(59b). Clearly, if the QP dominates the highest wh-word, then that wh-word will ap-
pear overtly fronted, deriving the ‘Superiority-satisfying’ order (59a). However, if the
QP dominates any of the lower wh-words, then it will be those wh-words that appear
overtly fronted in the sentence, yielding the ‘Superiority-violating’ order (59b).
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 131
In summary, we have seen that the addition of ForceQ+ (rather than ForceQ2) to
the lexicon of German has the following consequences: (i) it provides a composi-
tional semantics for German multiple wh-questions; (ii) it predicts that German mul-
tiple wh-questions should not exhibit Superiority Effects; and (iii) it predicts that the
in situ wh-words of German multiple wh-questions should be subject to Intervention
Effects. Thus it is possible in our Q-based account to derive the data in (40) and (43)
from a single core property of the language.
Furthermore, when we compare our analyses of German and English, an interesting
picture emerges of the cross-linguistic variation observed in (40)–(43). Under this account,
the core, underlying difference between English and German lies in their inventory
of Force heads.25 English possesses ForceQ2 and not ForceQ+, which entails that its
wh-questions exhibit Superiority Effects, but not Intervention Effects. German, however,
possesses ForceQ+ and not ForceQ2, which entails that its wh-questions exhibit Intervention
Effects, but not Superiority Effects. We see, then, that our Q-based theory can provide a
unified account of the data in (40)–(43), one that derives these two well-known differences
between German and English from a single point of variation in their lexicons.
To begin, our analysis of the data in (40)–(43) easily generalizes beyond simply
English and German. Indeed, our Q-based account of (40)–(43) makes the following
general prediction.
(65) Nonbinary Multiple Wh-Questions in English Can Violate Superiority (Kayne 1983)
a. Who1 gave what2 to whom3?
b. What2 did who1 give to whom3?
c. To whom3 did who1 give what2?
(66) Nonbinary Multiple Wh-Questions Are Subject to Intervention Effects (Pesetsky 2000)
a. Who1 didn’t give what2 to whom3?
b. * What2 didn’t who1 give to whom3?
c. * To whom3 didn’t who1 give what2?
wh-questions. As we have seen, such an analysis would largely capture the excep-
tional properties of English nonbinary wh-questions in (65) and (66).
The similarity of the facts in (67) to those in (40) have lead most linguists to consider
them instances of the same general phenomenon, that is, Intervention Effects.
However, the theory of Intervention Effects we assume here offers no obvious
explanation for the facts in (67), a property that it inherits from Beck (2006).30 After all,
Intervention Effects are assumed to result from the uninterpretability of the structure in
(53), and there is no obvious instance of (53) in (67b). Therefore, unlike earlier
theories of Intervention Effects (Beck 1996; Pesetsky 2000), ours does not straight
forwardly extend to the data in (67).
On the other hand, our inability to predict the data in (67) may actually be a
virtue of our account. That is, there is in fact some indication that the data in (67)
represent a distinct phenomenon from that seen in (40). To ease our discussion here,
I will refer to the phenomenon in (67) as ‘S(eparation)-Intervention Effects’ and the
phenomenon in (40) as ‘I(n situ)-Intervention Effects’.
Examination of German reveals that the set of ‘offending operators’ triggering
S-Intervention Effects is different from those that trigger I-Intervention Effects. First,
as originally reported by Pesetsky (2000), S-Intervention Effects in German can be
triggered by wh-words.
(68) In Situ Wh-Words Are Interveners for the German Separation Structures
“Guenther Grewendorf (personal communication) points out that wh-in-situ
appears to generate an intervention effect in German separation constructions. . .”
(Pesetsky 2000: 117)
136 THE GRAMMAR OF Q
(69) In Situ Wh-Words Are Not Interveners for Other In Situ Wh-Words
Wer hat wem was gegeben?
who-NOM has whom-DAT what given
Who gave what to who? (Kai von Fintel, personal communication)
Thus, in a pied-piping structure, phrasal material intervenes between the focused wh-
word and the focus-sensitive Q-particle that it ‘associates’ with. Consequently, if the
wh-word in such structures were to appear within the scope of a focus-sensitive op-
erator within the pied-piped phrase, then a structure of the type in (53) would result.
The following diagram illustrates.
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 137
Given our theory of Intervention Effects, we predict that structures of the type in (71)
will be uninterpretable, and thus ill-formed. Therefore we find that by combining our
theory of pied-piping structures with our theory of Intervention Effects, we thereby
make the following, general prediction.
But is the prediction in (72) true? Sauerland and Heck (2003) have independently
shown that it is indeed borne out for German. They show that, for every ‘offending
operator’ triggering an Intervention Effect with in situ wh-words, placement of that
operator in configurations like (71) creates an ill-formed structure. For example, we can
see in (73) that it is generally possible for a ‘pied-piping’ wh-word in German to be
c-commanded by a determiner in the pied-piped constituent (73a). If that determiner,
however, is the offending operator kein ‘no’, the sentence becomes ill-formed (73b).
(73) Intervention Effects in German Pied-Piping Structures (Sauerland & Heck 2003)
a. Fritz möchte wissen [ ein wie schnelles Motorrad ] du fahren darfst.
Fritz wants to.know a how fast motorbike you drive may
Fritz would like to know how fast a motorbike you are allowed to drive.
b. * Fritz möchte wissen [ kein wie schnelles Motorrad ] du fahren darfst.
Fritz wants to.know no how fast motorbike you drive may
Sauerland and Heck (2003) go on to show that parallel data hold for each of the
‘offending operators’ of German.
We can also detect evidence for (72) in the pied-piping structures of English.
Here, however, the evidence is somewhat more subtle than the striking contrasts
observed above for German. This is largely because independent features of English
render (72) somewhat difficult to test. The overall issue that, unlike pied-piping in
German, the only way that one can pack an offending operator into a pied-piping
structure of English is if the fronted phrase undergoes so-called Massive Pied-Piping
(Heck 2004, 2008). The phenomenon of ‘Massive Pied-Piping’ will receive a proper
introduction and discussion in chapter 5. For our purposes here, however, all we need
know is that such pied-piping, illustrated by structures like (74), is already indepen-
dently ‘marginal’ and ‘unnatural’ in English.
138 THE GRAMMAR OF Q
Although sentences like (74) are already marginal in English, they become dis-
tinctly worse if the ‘massively pied-piping’ wh-word is in the scope of an ‘offending
operator’.
Importantly, varying the grammatical function of the pied-piped phrase does nothing
to lessen the contrasts in (75). Rather, the contrasts are made even shaper.
(78) No Intervention Effect if the Wh-Word Is Not Within the Scope of the Offending
Operator
a. [ Which picture [ only of presidents ] ] does Jim own?
b. [ Which picture [ only of presidents ] ] hangs behind Jim’s desk?
Similarly, we should note that the ill-formedness of (75b,c)–(77b,c) is not due to any
‘pragmatic unnaturalness’ of the questions themselves. As the sentences below dem-
onstrate, such questions are well-formed and natural, just so long as there is no pied-
piping of the offending operator.
APPLICATIONS TO OTHER WH-FRONTING LANGUAGES, PIED-PIPING , AND INTERVENTION EFFECTS 139
(80) The Well-Formedness of (75) and (77) Without Pied-Piping of the Operator
a. [ Which president ]1 does Jim own [ no picture of t1 ] ? (cf. (75b))
b. [ Which president ]1 does Jim own [ only PICTURES of t1 ] ? (cf. (75c))
c. [NP Pictures of which president ] ] don’t hang in Jim’s office? (cf. (77b))
We find, then, that the pied-piping structures of English provide further evidence
for the generalization in (72). These facts also make several additional points con-
cerning the nature of Intervention Effects. First, recall that English does not (gener-
ally) exhibit Intervention Effects with its in situ wh-words. Thus, despite the
ill-formedness of (75b,c), the following are well formed.
However, such a local position for Q is not possible in (75b,c). Given our Q-based
theory of wh-fronting, the Q-particle must take as complement the fronted phrase of
the wh-question. Thus, in (75b,c), the Q must be located outside the scope of the
offending operator, yielding a configuration of the type in (71). Thus our Q-based
account correctly predicts the contrast between (75b,c) and (81a,b).
This is, to my knowledge, a unique prediction of our Q-based account. For
example, these data cannot be captured under the syntactic theory of Intervention
Effects put forth by Beck (1996). In brief, Beck (1996) proposes that Intervention
Effects follow from a syntactic constraint banning covert movement over the offend-
ing operators. In order for such an account to capture the contrast between (75b,c)
and (81a,b), it must be assumed that (i) pied-piping wh-words in English are covertly
extracted from within the pied-piped phrase, and (ii) in situ wh-words in English
undergo no covert movement. Such a difference between pied-pipers and in situ
wh-words, however, seems hard to independently motivate.
More generally, the contrast between (75b,c) and (81a,b) demonstrates that In-
tervention Effects do not simply result from a wh-word being inside the scope of an
offending operator. After all, in both (75b,c) and (81a,b), the wh-word is inside the
scope of such an operator. Rather, the real issue is whether a Q-particle is inside the
scope of the offending operator, intervening between it and the wh-word. In ill-
formed sentences like (75b,c), the Q-particle cannot be in the scope of the operator,
and so the sentences are uninterpretable. However, sentences like those in (81) can be
140 THE GRAMMAR OF Q
parsed as having a Q-particle in the scope of the offending operator, and so the
sentences are well formed.
Finally, note the following related property of our theory of Intervention Effects
in pied-piping structures. Our prediction of the generalization in (72) in no way rests
upon any assumptions regarding multiple wh-questions. That is, our derivation of
(72) in no way appeals to whether a given language possesses ForceQ2 or not. Thus
our theory predicts that (72) will hold across all languages, regardless of whether or
not their in situ wh-words are subject to Intervention Effects. As we have seen, this is
indeed a correct prediction, as (72) holds for both English and German. We find,
then, that our proposed account of (72) rightly (and uniquely) predicts this differ-
ence between the Intervention Effects triggered in pied-piping structures and those
associated with in situ wh-words.
5
5.1 Introduction
In the previous chapter we began our argument that the Q-based structure in (1)
extends beyond the Tlingit language, and in fact underlies the wh-questions of all
wh-fronting languages.
CP
QP1 CP
Complementation
XP Q
CQ IP
Agree/
… wh-word… Attract
QP1
Overt Movement
In this chapter we continue our argument, focusing primarily on the rather stringent
restrictions on pied-piping found in languages like English. Such restricted pied-
piping I will label ‘limited pied-piping’. In languages exhibiting limited pied-piping,
certain structures are not permitted to dominate the wh-word within the fronted
phrase of a wh-question. I will put forth a ‘Q-based’ theory of limited pied-piping,
141
142 THE GRAMMAR OF Q
one in which the restrictions on pied-piping structures ultimately follow from a re-
quirement that Q undergo Agreement with the wh-word. Moreover, we will see that
this account lays the groundwork for treatments of two related phenomena: secondary
wh-movement (Heck 2008) and massive pied-piping (Heck 2008).
Section 5.2 presents our treatment of limited pied-piping. It begins with an intro-
duction to the phenomenon, as well as the questions it raises for our Q-based account.
Following this, I provide some background to the leading ideas employed by our
analysis, particularly the work of Kratzer and Shimoyama (2002) and Pesetsky and
Torrego (2007). With this background in place, I then introduce the central claim of
our analysis: languages exhibiting limited pied-piping are those where the Q-particle
must Agree with the wh-word inside its complement. I show how this generalization
can capture the core aspects of limited pied-piping, as well as a number of other, re-
lated phenomena noted in the literature on pied-piping.
Having shown how our account derives the core properties of limited pied-piping,
I turn in section 5.3 to some further predictions made by our account. I first argue that
our Q-based account is able to derive the ‘Generalization on Recursive Pied-Piping’
of Heck (2008). I then discuss the predictions our account makes regarding the pied-
piping of main predicates, as well as the relationship between pied-piping and left
peripherality. Following this, I argue that our Q-based account correctly predicts
that pied-piping should (in principle) be able to occur in free variation with extrac-
tion of the wh-word. Finally, section 5.3.5 demonstrates how our Q-based account
can capture the fact that limited pied-piping is subject to the Coordinate Structure
Constraint.
In section 5.4, we momentarily turn away from pied-piping per se to consider
a related phenomenon. In the pied-piping structures of many languages, a ‘pied-
piping’ wh-word must sometimes undergo movement within the pied-piped phrase.
This phenomenon is dubbed ‘secondary wh-movement’ by Heck (2008). I present
a Q-based analysis of secondary wh-movement under which the wh-fronting occurs
to place the wh-word in a position where it is accessible for Agreement with
Q. I then argue that secondary wh-fronting in Chol and Tzotzil (Mayan) may pro-
vide interesting support for our Q-based account.
The final topic of this chapter is so-called massive pied-piping. Following
Heck (2008), the term ‘massive pied-piping’ describes cases where the stringent
constraints of limited pied-piping languages appear somewhat weakened. I argue
that a particular Q-based account of massive pied-piping can capture certain prop-
erties that it seems to exhibit. However, I also note that the properties of most impor-
tance to Heck (2008) do not obviously follow under our proposed account.
Consequently, massive pied-piping remains an important outstanding puzzle for our
Q-based theory.
In chapter 4 I argued that our Q-based analysis in (1) seems to provide an attractive
theory of pied-piping structures. Under this theory, pied-piping structures are nothing
more than cases where the Q takes as its sister a phrase strictly containing the wh-word
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 143
associating with it. Thus even complex pied-piping structures like (2a) can be ana-
lyzed as in (2b), as simply cases of normal phrasal movement of the QP.
This general theory of pied-piping structures is, of course, based on their surface
form in Tlingit. To recall, Tlingit pied-piping structures transparently possess the
structure in (2b), as the Q-particle sá always appears directly to the right of the
fronted phrase in a Tlingit wh-question.
We find, then, that our Q-based theory would view all pied-piping structures as
homologous to the Tlingit phenomenon in (3). However, when we compare the pied-
piping structures of languages like English to their putative correlates in Tlingit, we
find that there are some striking differences between them. Generally speaking, the
differences lie in the ‘size’ of what can be ‘pied-piped’. Tlingit permits the wh-word
in a pied-piping structure to be dominated by structures that English and other well-
studied languages never allow.
We have already encountered one rather prominent example of this disparity. Recall
that the wh-word of a Tlingit wh-question can be contained inside an island within the
fronted phrase (4). I will refer to such structures as “pied-piping past islands”.
Curiously, while Tlingit allows pied-piping past islands, the most commonly studied
wh-fronting languages do not. As illustrated in (5), the English correlates of (4)
are ill-formed.
144 THE GRAMMAR OF Q
Furthermore, as documented by Heck (2004, 2008), such structures are similarly ill-
formed in many other wh-fronting languages.
But it is not merely pied-piping of islands that distinguishes the pied-piping
structures of Tlingit. To facilitate our discussion here, let us adopt the following,
more general terminology:
With this terminology in place, note that several authors have offered the generaliza-
tion that English and other wh-fronting languages do not permit pied-piping past
lexical categories (Cowper 1987; Webelhuth 1992; Grimshaw 2000).1 That is, in the
most commonly studied wh-fronting languages, no wh-operator can be dominated by
a lexical category within the fronted phrase of the wh-question. The ill-formed
English structures in (7) illustrate.2, 3, 4
Similarly, I will use the term “limited pied-piping language” to refer to languages
where all pied-piping structures are instances of limited pied-piping.
Now, in order for our Q-based theory to be applied to the limited pied-piping
languages, some account must be offered for why those languages do not permit
pied-piping past islands or lexical categories. In the remainder of this section, I put
forth such an account. I begin, in the following subsection, with some critical back-
ground.
(9) The Independence of Valuation and Interpretability in Pesetsky and Torrego (2007)
Within Pesetsky and Torrego’s (2007) system, there are two principles that drive
syntactic valuation. The first is the requirement that every feature must possess a value
by LF. Because of this principle, any unvalued feature F[ ] must probe for a valued
instance of itself F[val], at which point the usual mechanics of long-distance Agree
apply (Chomsky 2000). The second is the requirement that all uninterpretable features
uF must, by LF, be matched to some interpretable instance iF. For further details regard-
ing this theory of feature valuation, I refer the reader to Pesetsky and Torrego (2007).
With these ideas in place, let us now incorporate the morphosyntactic hypotheses of
Kratzer and Shimoyama (2002) into our broader system. First, we assume that the wh-
words of some languages (e.g., German, English) bear uninterpretable, valued Q, while
the wh-words of other languages (e.g., Japanese, Tlingit) do not bear any instance of Q.
In languages where wh-words bear uQ[+], this uninterpretable Q-feature must, by LF,
be ‘checked’ against some interpretable instance iQ. Given that the only head assumed
to carry iQ is the Q-particle itself, the Q-particle in languages like German and English
must Agree with the wh-word. In order for this Agreement to take place, however, we
must assume that the Q-particles of such languages initially bear unvalued instances
of Q.9 That is, in languages where the wh-words bear uQ[+], the Q-particle must in
turn bear iQ[ ]. Of course, in languages where wh-words do not bear uQ[+], we can
assume that the Q-particle simply bears an interpretable, valued instance of Q:
Assuming the initial valuations in (10) and (11), we predict the necessity of Q/wh-
Agreement in languages like German and English, and the absence of such Agreement
from languages like Japanese and Tlingit. First, in languages where the Q-particle is lexi-
cally assigned iQ[ ], the lack of a value for iQ entails that the particle probe for a valued
instance of the feature. Following Chomsky’s (2000) algorithm for probing, the first ele-
ment bearing Q[val] that the Q-particle probes is the wh-word that it c-commands. There-
fore the Q-particle will undergo Agreement with that wh-word, as shown in (12):
QP
XP Q
YP X iQ[+]
… was uQ[+] …
Q/Wh-Agreement
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 147
This Q/wh-Agreement has the following results: (i) the unvalued instance of [Q] on
the Q-particle receives a value, and (ii) the uninterpretable instance of [Q] on the
wh-word is matched to an interpretable instance. Consequently, both the Agreement-
driving principles of Pesetsky and Torrego (2007) are satisfied, and the structure is
well-formed.
In languages where the Q-particle is lexically assigned iQ[+], however, the
presence of a value for iQ entails that the Q-particle will not act as a probe. Further-
more, since the wh-words of such languages are assumed not to bear any instance of
the Q-feature, nothing will require them to undergo Agreement with the Q-particle.
Consequently, in such languages, there is no Q/wh-Agreement.
XP
XP Q
YP X ka iQ[+]
… dare …
That is, I propose that the constraints governing pied-piping in languages like
English would follow from a single requirement that the Q-particle and the wh-word
Agree.
Let us first consider the condition against pied-piping past islands. Recall our as-
sumption from chapter 2 that probing and Agreement cannot apply across islands.
148 THE GRAMMAR OF Q
Under this general assumption, Q/wh-Agreement would, of course, not be able to ap-
ply across islands. Therefore, if we assume that limited pied-piping languages require
Q/wh-Agreement, we correctly predict that such languages will not permit pied-piping
past islands. As illustrated in (15), the domination of the wh-word by an island within
the sister of Q would prevent Agreement between the Q-particle and the wh-word.
DP Q iQ[ ]
D NP
A ISLAND
NP CP
XXXXXX
More concretely, in languages like German and English—where the Q-particle bears
interpretable but unvalued Q (iQ[ ])—an island between the Q-particle and the wh-
word prevents the Q-particle from receiving a value for its Q-feature by LF. As a
result, such structures induce a crash at the LF interface.
On the other hand, in languages where the Q-particle lexically bears valued
Q (iQ[+]), an island between the Q-particle and the wh-word does not affect the
LF interpretability of the structure.
DP Q iQ[+]
sá
D NP
CP NP
ideas regarding the structure of lexical projections. Let us adopt the following view,
taken from recent work in Distributed Morphology.
(17) The Fine Structure of Lexical Categories (Embick & Marantz 2008)
Every lexical projection (VP, NP, AP) is complement to a phase head (little-v,
little-n, little-a).11
Diagram of the Lexical Projections
nP vP aP
Spell-Out Spell-Out
n NP Domain v VP Domain a AP
Spell-Out
…. …. Domain …
Under the assumption that each of the ‘little categorial heads’ shares with little-v
the property of being a phase head (Chomsky 2000), it would follow that Q/wh-
Agreement cannot cross into lexical projections. As illustrated in (18), any material
inside a lexical projection would occupy a separate spell-out domain from material
outside the lexical projection. However, under the original formulation of the Phase
Impenetrability Condition (PIC), syntactic operations such as Agree cannot apply to
heads in separate spell-out domains (Chomsky 2000).12 Under these assumptions,
then, no material inside a lexical projection can Agree with any head outside that
lexical projection. Consequently, we predict that Agreement between a Q-particle
and a wh-word buried within a lexical projection should be impossible.
QP
… n/v/aP… Q
XXXXXXX
Agreement Impossible, Due to the PIC
Thus, in languages like English and German, where the Q-particle bears unvalued
‘iQ[ ]’, it should be impossible for a lexical category to intervene between Q and the wh-
word. Consequently, pied-piping past lexical categories should be ill-formed in these
languages. However, for languages like Tlingit and Japanese, where the Q-particle bears
valued ‘iQ[+]’, no problem arises if the wh-word and Q-particle are separated by a lexi-
cal projection, and so pied-piping past lexical categories should be possible there.
150 THE GRAMMAR OF Q
(19) Complements of Lexical Heads Cannot Pied-Pipe (in Limited Pied-Piping Languages)
a. I wonder [ [DP whose [NP pictures ] ] John bought ]?
b. * I wonder [ [NP pictures of whom ] John bought ]?
c. * I wonder [ [AP proud of whom ] John was ]?
d. * I wonder [ [VP eaten what ] John has ]?
Now, one can certainly imagine other explanations for the impossibility of
(19b,c). Indeed, facts such as these have lead some to propose the stronger general-
ization that no complements of any head can serve as pied-pipers (Kayne 1994;
Koopman 2000; Koopman & Szabolcsi 2000). However, this stronger generalization
is challenged by the ubiquity of pied-piping by complements of P. That is, contrary
to the stronger generalization, it is possible in all the best-studied wh-fronting lan-
guages for wh-words to pied-pipe a PP from the complement of P, as briefly illus-
trated in (20).
To allow for these rather common structures, proponents of the stronger general-
ization must introduce special assumptions whereby the prima facie complements
of P are (at the relevant level of syntax) actually specifiers of P. In contrast, the
weaker generalization against pied-piping past lexical categories actually predicts
the ubiquity of PP pied-piping. Because P may be regarded as a functional category
152 THE GRAMMAR OF Q
(in most languages), the weaker generalization correctly predicts that wh-words
may generally be dominated by PPs within the fronted phrase.
Thus a ban on pied-piping past lexical categories correctly predicts that comple-
ments of lexical heads cannot ‘pied-pipe’. In addition, it also predicts that modifiers
of lexical projections cannot pied-pipe (Webelhuth 1992):
(21) Modifiers of Lexical Heads Cannot Pied-Pipe (in Limited Pied-Piping Languages)
a. * [QP[DP The [NP party where ] ] Q ] will John enjoy?
b. * [QP[VP Go where ] Q ] will you?
c. * [QP[DP A [NP [DegP how big ] party ] ] Q ] will you throw?
Again, since the wh-words in each of these structures are contained within a lexical
projection inside the fronted phrase, our Q-based account predicts their ill-formedness.14
A third prediction of our ‘weaker generalization’ concerns pied-piping by pos-
sessors. A pervasive phenomenon across limited pied-piping languages is the inabil-
ity for postnominal possessors to pied-pipe (Heck 2008: 89–94). Sentence (19b)
demonstrates this for English, and (22) illustrates this for German. Note that although
German generally permits possessors to follow the possessed noun (22a), this cannot
occur when the possessor is a pied-piping wh-word (22c). Rather, in such construc-
tions, the pied-piping possessor must appear prenominally (22b).
As discussed by Heck (2004, 2008, 2009), similar data are observed in many other
European languages, as well as the Mayan language, Tzotzil.15 Now, recall that in
these head-initial languages, it is commonly held that postnominal possessors are
complements of the possessed N. Consequently, if they were ever to pied-pipe the
possessive phrase, they would be pied-piping past a lexical category. Therefore a ban
on pied-piping past lexical categories would correctly predict the general impossibil-
ity of such structures.16
Our ‘weaker generalization’ can also capture a related, though subtly different,
phenomenon in Hungarian. Szabolcsi (1994) argues that possessors in Hungarian
can occupy either of two positions inside the DP: one internal to the NP and one
external to it. Importantly, however, both these positions are prenominal; Hungarian
does not permit postnominal possessors. Nevertheless, the position a possessor oc-
cupies can in part be determined by its case: possessors internal to NP bear nomina-
tive, while those external to NP bear dative. Interestingly, only dative-marked
possessors can pied-pipe.
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 153
Given the evidence that the nominative possessor in (23b) is NP-internal, the
ill-formedness of (23b) would follow from our Q-based account. Furthermore, our
account provides a unified treatment of both (23) and (22), despite the fact that
(23) superficially concerns the case of the pied-piper, while (22) concerns its hierar-
chical position. To put the matter more acutely, both the wh-possessors in (23) are
assumed to occupy specifier positions within the fronted phrase. Consequently, any
attempt to capture the contrast in (22) via the ‘stronger generalization’ that pied-
pipers must be specifiers would not straightforwardly capture the parallel data in
(23). Thus the behavior of Hungarian wh-possessors supports our ‘weaker general-
ization’ against the notion that pied-piping is only possible from specifier positions
(Horvath 2007a: sect. 2.2).
A parallel argument can be made by certain intriguingly similar data from
English. As reported by Horvath (2007a), Culicover (1999) observes the following:
As in (23), the contrast between the sentences in (24) seems to lie in the case of the
‘pied-piper’. In the ill-formed (24a), the gerundive subject bears accusative case,
while in the well-formed (24b), it bears genitive. As noted by Horvath (2007a), our
account of the Hungarian data in (23) could easily be extended to these English data.
Let us assume that accusative-marked gerundive subjects in English occupy the NP-
internal position of Hungarian nominative possessors. Furthermore, let us assume
that genitive-marked gerundive subjects occupy the NP-external position of Hungarian
dative possessors (i.e., SpecDP). Under these plausible assumptions, the data in (24)
follow from our condition against pied-piping past lexical categories. Finally, we
should again make the point that both the wh-words in (24) are left-peripheral spec-
ifiers, and so these data would not follow as easily from a constraint against pied-
piping from the complement position.
We have seen, then, that an inability to pied-pipe past lexical categories predicts
an interesting variety of pied-piping phenomena. We may reasonably conclude, then,
that our Q-based theory of limited pied-piping is correct in its prediction of such a
constraint.
The reader may recall, however, that our Q-based account actually yields an
even stronger generalization. That is, via our assumption in (17), the impossibility of
pied-piping past lexical categories follows from a more general prediction that pied-
piping across separate phases should be impossible. Following the logic presented
in (18), in a Q/wh-Agreement language, the wh-word cannot be in a separate phase
from the root of the pied-piped phrase. If it were, then Agreement between the
wh-word and the Q would be impossible, and the derivation would crash.
154 THE GRAMMAR OF Q
This more general prediction might receive some support from an oft-noted
property of CP pied-piping in Basque and Ancash Quechua. While both Basque and
Ancash Quechua permit pied-piping of subordinate clauses, it has often been noted
that such pied-piped clauses must possess a particular form. In both these languages,
a subordinate clause S can only be pied-piped if the wh-word is fronted into the left
periphery of S (Heck 2008: 106–107). Thus, as we see in (25), neither Basque nor
Ancash Quechua permits subordinate clauses to be pied-piped by wh-words internal
to the subordinate IPs.
In light of the pattern in (25), it is interesting to note that such CP pied-piping cannot
occur in English if the wh-word is within the subordinate IP. For example, if the wh-
word is not a subject, it must appear fronted in the pied-piped clause.
Thus the best candidate in English for CP pied-piping appears to also follow the
pattern seen in (25): subordinate clauses can only be pied-piped if the wh-word has
been moved into their left periphery.
If we assume that Basque, Ancash Quechua, and English are all Q/wh-Agreement
languages, then the data in (25)–(28) would follow from our Q-based theory of pied-
piping. Since C heads are undoubtedly phase heads, each of the ill-formed structures
in (25a(ii)), (25b(ii)), (27b), and (28b) would require Q/wh-Agreement to apply across
separate spell-out domains, contrary to the PIC.
QP
CP Q
Spell-Out
C IP Domain
… wh-word ….
XXXXX
Agreement Impossible, by Virtue of the PIC
Thus, in order for a Q-particle to Agree with a wh-word contained within a subordi-
nate CP, that wh-word must undergo movement to the specifier of the CP. As shown
in (30), such movement would place the wh-word and the Agreeing Q-particle within
the same phase.18
QP
CP Q
Wh-word1 CP
Spell-Out
C IP Domain
… t1 ….
Q/Wh-Agreement Possible
156 THE GRAMMAR OF Q
We find, then, that the special properties of pied-piped CPs in languages like
Basque and Ancash Quechua lend support to our broader prediction that pied-piping
past phases should be impossible in Q/wh-Agreement languages. As we have seen,
when combined with the hypothesis in (17), this broader prediction entails that pied-
piping past lexical categories should likewise be impossible in these languages. This
latter prediction, in turn, appears to capture a variety of phenomena that have been
observed in the literature on pied-piping.
Thus far, our discussion has centered on the two core properties of limited pied-pip-
ing languages: their inability to pied-pipe past islands, and their inability to pied-pipe
past lexical categories. In this section I will discuss some further predictions of our
Q-based theory. In some cases these predictions are fully general, and are expected
to hold for all wh-fronting languages. In other cases we will see that our system
makes different predictions for Q/wh-Agreement and non-Agreement languages.
As discussed by Heck (2008), the possibility of such ‘recursion’ can also be seen
in PP pied-piping. It is most easily observed in languages where pied-piping of PPs
is obligatory, such as German. In German, the complement of P can pied-pipe the PP
in which it is contained (33a). Moreover, if a PP containing a wh-complement is
itself the complement of a larger PP, then the larger PP may also be pied-piped by the
wh-word (33b). Such embedding can be further iterated, and there again seems to be
no principled limit on the depth.
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 157
Finally, just as we would expect from the general statement in (31), the possibility
of recursive pied-piping is not sensitive to the exact position of the pied-piping
phrase. That is, possessor DPs can also pied-pipe larger PP complements, as we see
in (34).
Let us now confirm that our Q-based theory can predict the generalization in
(31). First, recall that within our theory of pied-piping structures, the informal notion
that ‘A can pied-pipe B’ amounts more precisely to the claim that ‘A can be domi-
nated by B in the fronted phrase of a pied-piping structure’. With this in mind, we
restate the generalization in (31) so that it reads as in (35).
Non-Agreement languages can easily be shown to satisfy (35). First, note that
the only time a non-Agreement language disallows X from being dominated by Y in
a pied-piping structure is when Y is not a permissible sister to Q. Let us, then, assume
that B can be dominated by C in a pied-piping structure, and that B contains A. This
is diagramed in the figure in (36).
QP
C Q
A
…wh-word…
158 THE GRAMMAR OF Q
QP
B Q
A
… wh-word…
Agreement Possible:
Thus, No ‘Barriers’ to Q/Wh-Agreement Within B
QP
C Q
B
… wh-word…
Agreement Possible:
Thus, No ‘Barriers’ to Q/Wh-Agreement Within C
QP
C Q
A
… wh-word…
We find, then, that our Q-based account predicts that across languages, pied-
piping should obey the generalization in (31)/(35), even in those languages where
pied-piping is otherwise subject to rather stringent locality conditions.
Let us adopt the picture in (43) regarding the syntax of copulas and auxiliary verbs.
IP
DPj IP
Subject
I F1P
F1 …
FnP
Fn VP
V VP/AP/DP
Copula,
Auxiliary tj Main Predicate
That is, as illustrated in (43), we assume that copulas and auxiliary verbs are initially
merged as complements to the lowest functional projection in the ‘functional spine’
of the clause. Further, we assume that the main predicate of the clause is merged as a
complement to the copula/auxiliary. Finally, we assume that both the copula and the
auxiliary verb are truly verbal in nature. That is, they are V heads, and thus instances
of a lexical category.
With these assumptions in place, it is apparent that the ill-formedness of
(40a) follows from the QP-Intervention Condition in (42). Since the copula is
pied-piped in (40a), it follows that its projection must be sister to the Q. Conse-
quently, a QP intervenes between the projection of the copula and the functional
head (Fn) which selects for it. As illustrated in (44), such a configuration would
violate (42).20
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 161
CP
QP1 CP
VP Q CQ IP
V DP DP IP
Is Dave
whose doctor I …
FnP
Impossible FnP,
Ruled out by (42) Fn QP
t1
Importantly, however, (42) would not rule out the structure in (41a). Since the copula
is a lexical category, it should be possible for a QP to intervene between it and the
main predicate it selects. Furthermore, unlike (41b–d), the wh-word is not dominated
by a lexical category within the fronted phrase, and so Q/wh-Agreement is possible.
This reasoning is illustrated in (45).
(45) Pied-Piping of a Main Predicate in English
CP
QP1 CP
DP Q CQ IP
DP DP DP …
Whose Dave
doctor FnP
Fn VP
No violation of (42) V QP
is
t1
It is clear from the logic of the preceding discussion that our Q-based theory of
pied-piping predicts the following generalization(s).
The argument surrounding (44) makes clear the prediction in (46a). The generaliza-
tion in (46b) follows from the natural assumption that in the absence of any copula
or auxiliary, the main predicate is the complement of the lowest functional head Fn.
That is, as illustrated in (47b), sentences like (47a) also violate the QP-Intervention
Condition.
QP1 CP
VP Q CQ IP
V DP DP IP
Eats what Dave
I …
Fn P
Impossible F nP,
Ruled out by (42) Fn QP
t1
Thus far we have seen that the predictions in (46) are true for the Q/wh-Agreement
language English. Moreover, it is apparent from the work of Heck (2008) that (46) is also
true of many other Q/wh-Agreement languages. Indeed, throughout Heck’s extensive
study of pied-piping structures across the world, there is (to my knowledge) no counter-
example to the generalizations in (46). However, many of the languages that Heck (2008)
examines might be classifiable as Q/wh-Agreement languages. Is there clear evidence
that (46) does indeed also hold for non-Agreement languages like Tlingit?
In this context, it is useful to again note the ill-formedness of the Tlingit sen-
tences in (48).
Given our Q-based analysis of Tlingit wh-questions, each of the structures in (48) con-
tain pied-piping past a main predicate in a clause containing no copula or auxiliary. As
already noted in chapter 2, our QP-Intervention Condition predicts the ill-formedness
of these structures.
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 163
Thus we find that Tlingit supports the prediction in (46b) for non-Agreement
languages. Support for (46a) comes from the ill-formedness of structures like (49b).
As shown in (49a), Tlingit permits pied-piping of main predicates when they are
complement to its copular verb si-tee, ‘to be’. In such sentences, however, the copular
verb can never itself be pied-piped (49b). Tlingit, then, supports the prediction that
(46) holds of even the non-Agreement languages.21
In addition to the generalization in (46), our account of the data in (41) entails
the general prediction in (50).
Again, (50) follows from the fact that pied-piping of a main predicate P will not vio-
late the QP-Intervention Condition when P is complement to a lexical head, such as
an auxiliary or copula (45).
Is the generalization in (50) accurate? We have already seen that the Tlingit data
in (49) support the prediction for non-Agreement languages, and the English data in
(41) support the prediction for Q/wh-Agreement languages. Furthermore, the exten-
sive study of Heck (2008) again lends support to this prediction; throughout Heck’s
study, there are no (indisputable) counterexamples to (50).22
Thus far in our discussion, we have limited our attention (and our generaliza-
tions) to main predicates and auxiliaries. The reason for this is that, as discussed
below, our Q-based account makes rather different predictions for subordinate
clauses.
The diagram in (52) illustrates how our account makes the prediction in (51).
164 THE GRAMMAR OF Q
CP1
QPj …
CP2 Q LP
L QP
… Wh-Word …
tj
As illustrated above, let us assume that (a) an entire subordinate clause CP2 is
pied-piped, and (b) CP2 is selected by a lexical head L. Since CP2 is pied-piped, it
must be sister to the Q. Thus a QP intervenes between the projection of CP2 and
the head L selecting for CP2. By assumption, however, this head L is a lexical
head, and so the entire structure satisfies the QP-Intervention Condition. There-
fore, if the language in question is a non-Agreement language (like Tlingit), our
principles predict that (52) should be well-formed. The accuracy of this prediction
is confirmed by the acceptability in Tlingit of CP pied-piping sentences like the
following.
Thus, while our account correctly predicts that pied-piping the highest verbal cate-
gory of a main clause should not be possible ((48)/(49)), it also correctly predicts that
it should (in principle) be possible to pied-pipe the highest verbal category of a
subordinate clause (53).
Of course, if the language in question requires Q/wh-Agreement, then there are
further constraints on pied-piping of subordinate predicates and auxiliaries. That is, as
stated in (51c), such pied-piping can only take place if the wh-word is not dominated by
any lexical projections or phase boundaries in the pied-piped clause. This, of course,
follows from the general assumptions regarding Agreement that were presented in sec-
tion 5.2. As noted in that section, support for the prediction in (51c) might be found in
the constraints on clausal pied-piping in languages like Basque and Ancash Quechua.
b. Ancash Quechua:
(i) [CP Imata1 [IP wawa t1 mikuchun ] ]2-taj Maria t2 munan?
what child eat Q Maria want
What does Maria want the child to eat?
(ii) * [CP [IP wawa imata mikuchun ] ]2-taj Maria t2 munan?
child what eat Q Maria want
Interestingly, such ‘analytic uniformity’ does not hold for most other theories of
pied-piping (Sells 1985; Cowper 1987; Kayne 1994; Koopman 2000; Grimshaw
2000; Koopman & Szabolcsi 2000). Under most other accounts, the analysis of struc-
tures like (55b) requires mechanisms not needed for the analysis of (55a). For exam-
ple, Kayne (1994) and Koopman (2000) hold that only specifiers can truly pied-pipe.
Consequently structures like (55b) can only be analyzed via appeal to covert move-
ment operations. On the other hand, Grimshaw (2000) proposes that complements of
P can pied-pipe because PP is an ‘extended projection’ of D, and so can inherit D’s
wh-feature. However, since her system assumes that a phrase can never be an extended
166 THE GRAMMAR OF Q
projection of its specifier, it follows that some other mechanism (i.e., Spec-Head
Agreement) must be responsible for pied-piping structures like (55a).
While such distinct treatments of (55a) and (55b) might seem inelegant at first,
they do hold one potential advantage over the uniform treatment of our Q-based ac-
count. Indeed, it is largely by design that these other accounts treat pied-piping by
CompPP as a distinct phenomenon from pied-piping by specifiers. Such accounts
generally seek to predict that, except for CompPP, all pied-piping must take place
from the specifier position. Importantly, this prediction is in fact true, at least for the
best-studied languages. As is clear from the cross-linguistic studies of Heck (2008)
and Horvath (2007a), in these languages it is generally the case that only P permits
pied-piping from its complement. In all cases but pied-piping of PP, a pied-piper
must be a specifier. Given this tendency, which Heck (2008, 2009) dubs the “Edge
Generalization”, we must therefore question whether our uniform treatment of pied-
piping is in fact accurate. Does our Q-based account fail to capture an important
property of pied-piping structures?
In fact, it does not. Rather, there is, under our Q-based account, a straightfor-
ward explanation for this pattern: P happens to be the only functional (non-lexical)
category that takes interrogative words as complements. To see this more clearly, let
us first consider the class of functional categories, which we might reasonably as-
sume to be the following: C, I, D, Deg, P.23 Now, consider the class of wh-words,
which we might reasonably assume to be represented by the following: who, what,
which, where, why, how. Placing these two sets side by side, we observe that the
only member from the former that can take as complement a member of the latter is
the category P. It follows, then, that if a wh-word ever occupies a complement posi-
tion, and is not complement to P, then it must be complement to some lexical head.
Consequently, in the best studied wh-fronting languages, such a wh-word will not be
able to pied-pipe.
In short, the reason why pied-pipers tend to be specifiers (in the Q/wh-Agreement
languages) is not that there is some special importance of the specifier position per se.
Rather, this tendency simply follows from the twin facts that (i) these languages only
permit pied-piping past functional categories, and (ii) there is only one functional
category, P, that takes wh-phrases as complements. Thus, for the Q/wh-Agreement
languages, if the wh-word of a pied-piping structure is ever contained within a phrase
other than PP, it must be within the specifier of that phrase. For this reason, in the
great majority of pied-piping structures (in the Q/wh-Agreement languages), the
wh-word will be a left-peripheral specifier.
Thus, while our Q-based theory does predict that, in the best-studied languages, pied-
pipers will almost always be specifiers, it does not derive this tendency from a general ban
on pied-piping from complement position. Rather, it simply follows from the independent
fact that P is the only functional (non-lexical) head to take wh-complements.
To begin, our Q-based theory predicts that pied-piping can, in principle, be op-
tional. That is, it predicts that pied-piping of a phrase XP by a wh-word W can, in
principle, be in free variation with subextraction of W from within XP. To see this, let
us first note that nothing within our Q-based theory requires that Q be ‘as close as
possible’ to the wh-word it associates with. Suppose that a wh-word is contained
within a structure XP, which is in turn contained within a larger structure YP.
YP QP
YP Q
QP
XP Q XP
…wh-word… …wh-word…
Moreover, let us suppose that neither XP nor YP contain any barriers to Q/wh-Agree-
ment. Finally, let us suppose that placement of Q as sister to either XP or YP would
not violate the QP-Intervention Condition. Under these assumptions, our Q-based
theory licenses both the structures in (56). That is, placement of Q as sister to XP is
predicted to be as well formed as placement of Q as sister to the larger structure YP.
Given our Q-based theory of wh-fronting in (1), the general well-formedness
of both structures in (56) entails that pied-piping of a larger phrase YP and sub-
extraction from within YP should be freely alternating options. That is, under our
Q-based account, the mere ability to extract an XP from within some constituent
YP does not alone entail that pied-piping of YP should not be possible. Pied-
piping and subextraction are predicted to, in principle, occur in free variation
with one another.
This predicted optionality of pied-piping seems to be a property of many attested
cases of pied-piping. For example, it is a property of the pied-piping of subordinate
CPs, as the data in (57)–(59) illustrate.
This predicted optionality is also a key empirical difference between the Q-based
approach advocated here and the theory of pied-piping developed by Heck (2004,
2008, 2009). A core prediction of Heck’s account is the following generalization.
(64) DP Splits in German (and Other Germanic Languages) (Heck 2008: 153–154)
a. Fritz fragt [ was1 du [DP t1 für Leute ] einlädst ].
Fritz asks what you for people invite
Fritz asks what kind of people you invite.
b. Fritz fragt [ [DP was für Leute ] du einlädst ].
Fritz asks what for people you invite
Fritz asks what kind of people you invite.
(65) DP-Splits in French (and Other Romance Languages) (Heck 2008: 155)
a. Je sais [ combien1 Marie a décidé d’engager [DP t1 de personnes ].
I know how.many Mary has decided to.employ of persons
I know how many people Mary has decided to employ.
b. Je sais [DP combien de personnes ] Marie a décidé d’engager.
I know how.many of persons Mary has decided to.employ
How many people has Mary decided to employ?
(66) DP-Splits in Russian (and Other Slavic Languages) (Heck 2004: 187)
a. Ja sprosil [ čju1 ty cital [DP t1 knigu ] ]
I asked whose you read book
I asked whose book you read.
b. Ja sprosil [ [DP čju knigu ] ty cital ]
I asked whose book you read
I asked whose book you read.
(69) DegP-Splits in German (and Other Germanic Languages) (Heck 2004: 188)
a. Ich frage mich [ [ wieviel Grade ]1 der Ofen [DegP t1 zu heiss ] war ].
I ask myself how.many degrees the oven too hot was
I wonder how many degrees too hot the oven was.
170 THE GRAMMAR OF Q
b. Ich frage mich [ [DegP wieviel Grade zu heiss ] der Ofen war ].
I ask myself how.many degrees too hot the oven was
I wonder how many degrees to hot the over was.
In each of the sentence pairs in (61)–(69), we find that subextraction of the wh-
word in the (a)-sentence exists side by side with pied-piping of a larger structure in
the (b)-sentence. Therefore these data appear to challenge the ‘Repair Generaliza-
tion’ in (60). We might, then, conclude that these data support our Q-based account
over the system developed by Heck (2004, 2008, 2009).
Such a conclusion, however, would be premature. First, it should be noted that
Heck (2004, 2008) is well aware of these data, and suggests several programmatic ways
that they might be approached within his framework (Heck 2008: 273–294). Further-
more, many of these data actually challenge our own Q-based theory of pied-piping.
To see this, let us begin by noting that our Q-based account only predicts that it
should, in principle, be possible for pied-piping and subextraction to be in free variation.
Our account does not predict that pied-piping will always be optional. In fact, there are
many circumstances where our Q-based account predicts that subextraction will block
pied-piping. This is especially so in the Q/wh-Agreement languages. Suppose, for
example, that we are examining a Q/wh-Agreement language, and we observe that
extraction of a wh-phrase from within YP is possible.
CP
QP1 …
XP Q YP
… wh-word…
t1
Given the QP-Intervention Condition, it follows that the sister of Q, XP, must be se-
lected by a lexical head. Consequently, QP is complement to a lexical projection.
CP
QP1 …
XP Q YP
… wh-word…
LP
L QP
t1
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 171
(72) Pied-Piping of YP
CP
QP 1 …
YP Q QP
LP t1
L XP
… wh-word …
XXX
Q/Wh-Agreement Blocked by LP
Under our analysis in chapter 4, P-stranding is possible in English and (by extension)
Icelandic because P in these languages is actually a lexical category. However, since
both English and Icelandic are presumably Q/wh-Agreement languages, our theory
from section 5.2 predicts that pied-piping past P should be ungrammatical. Clearly,
however, such structures seem to be possible in these languages.
172 THE GRAMMAR OF Q
(75) DP-Splits in Russian (and Other Slavic Languages) (Heck 2004: 187)
a. Ja sprosil [ čju1 ty cital [ t1 knigu ] ]
I asked whose you read book
I asked whose book you read.
b. Ja sprosil [ [ čju knigu ] ty cital ]
I asked whose book you read
I asked whose book you read.
NP
Possessor NP
N
Possessum
However, if wh-possessors in Russian are always contained within NPs, then the
possibility of possessor pied-piping in (75b) is unexpected. Such structures would
require pied-piping past a lexical category (NP), and so should not be possible in a
Q/wh-Agreement language like Russian.
There are two imaginable ways in which our Q-based account can approach
these challenges. The first is to posit that the problematic pied-piping structures are
instances of so-called massive pied-piping (Heck 2008). The phenomenon of ‘mas-
sive pied-piping’ will be treated in more detail in section 5.5. In brief, ‘massive pied-
piping’ refers to the ability for Q/wh-Agreement languages to tolerate pied-piping
past lexical categories in nonembedded clauses. For example, pied-piping past NP is
noticeably more tolerable in matrix questions (77a) than it is in embedded questions
(77b).
Of course, such an analysis would clearly not work for the pied-piping structures
in (73b) and (75b), which do appear in embedded clauses. However, it might be the
right analysis for PP pied-piping in English (74a). As noted by Heck (2008) inter
alia, pied-piping of PPs in English is noticeably worse in embedded clauses.
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 173
Consequently, Heck (2008: 124) proposes that cases like (74a) in English are likely
to be instances of ‘massive pied-piping’.25 Since massive pied-piping by definition
permits pied-piping past lexical categories (in Q/wh-Agreement languages), we find
that the data in (74a) are not necessarily inconsistent with our view that P in English
is a lexical category.
But what of the data in (73b) and (75b)? Another approach to pursue might be to
posit structural ambiguity. For the Russian data in (75), we might propose that Russian
possessive nominals can have either the structure in (76) or a classic ‘DP-structure’ as
in English. When the possessor is a mere adjunct to NP, extraction of the possessor
occurs, as in (75a). However, when the possessor occupies SpecDP, pied-piping of the
entire possessive phrase occurs, as in (75b).26 A similar ‘ambiguity approach’ could be
applied to the Icelandic data in (73). Let us suppose that in Icelandic, prepositions
have an ambiguous status and can be categorized as either functional or lexical heads.
When P enters the derivation as a lexical head, P-stranding as in (73a) is derived.
However, when P enters the derivation as a functional head, P-stranding is prevented,
and PP pied-piping is required (73b).
Of course, these proposals are all merely programmatic suggestions, and it is far
from clear whether they can ultimately be made to work. We find, then, that while our
Q-based account predicts that pied-piping should sometimes be optional, our account
actually has some difficulty capturing many attested examples of optional pied-piping.
Given these considerations, I conclude that it is far from clear which of the two
accounts under consideration—Heck’s (2008) account or our own—fare best with
respect to the optionality of pied-piping. While the data in (61)–(69) challenge Heck’s
(2008) prediction that ‘true’ optionality should never occur, our own account does
not straightforwardly capture those data either.27
As one final note, let us observe that our Q-based account is not entirely incon-
sistent with the ‘Repair Generalization’ in (60). That is, while our account does not
on its own predict (60), certain assumptions could be added in order to derive it. As
noted by Heck (2008: 52–53), one could imagine a Q-based account in which some
principle requires that Q be merged as close as possible to the wh-word.28 Clearly,
within such an account, the two structures in (56) would not be allowed to coexist,
and so pied-piping would never be in free variation with subextraction. Thus I leave
to future research the question of whether Heck’s ‘Repair Generalization’ should be
captured under a Q-based account, and if not, how best to capture the substantial
evidence that Heck marshals in support of it.
(80) a. (?) [ [ John’s books ] and [ whose paintings ] ] did you sell?
b. ?? [ [ whose paintings ] and [ John’s books ] ] did you sell?
Interestingly, however, if both conjuncts contain wh-words, then the two wh-words
together can pied-pipe the entire coordinate structure.
As noted by many, these facts bear a striking resemblance to the Coordinate Structure
Constraint (CSC), which prevents movement from targeting a single constituent
within a coordinate structure.
Naturally, then, previous studies have sought to reduce the facts in (79)–(81) to the
CSC. Under an account that assumes ‘feature percolation’, such reduction is straight-
forward: simply assume that ‘feature percolation’ is subject to the CSC (Heck 2008:
310).30 In a system like ours, however, which does not employ feature percolation,
how are these facts to be captured?
One possibility might be to posit that the CSC is ultimately about Agreement
rather than movement.31 Consider the following constraint.
First, note that, under the assumption that all movement is a reflex of Agreement,
the condition in (83) is enough to capture the data in (82). In brief, (82a) is impos-
sible because it would require the matrix interrogative C head to bear an Agree
relation only with an element of the second conjunct, and not also the first. However,
(82b) is possible because in this structure the interrogative C head bears an Agree
relation with elements within both conjuncts.
Now let us see how the condition in (83) might account for the data in (79)–(81).
Since English is a Q/wh-Agreement language, Agreement must hold between the Q and
the wh-word it associates with. Moreover, under our Q-based account, the Q-particle
associated with a wh-word always dominates the fronted phrase of the wh-question. As
illustrated in (84), it therefore follows that (79) and (80) will violate (83).
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 175
QP
ConjP Q
XXXXX
AGREEMENT-BASED CSC (83) VIOLATED
Agreement between Q and member of DP2, but
No Agreement between Q and a member of DP1
According to this analysis, the problem with the structures in (79) and (80) is that the
Q-particle sitting above the pied-piped coordination only Agrees into one of the two
conjuncts. It follows, then, that if a wh-word is added to the other conjunct of these
structures, then they should be well-formed.
QP
ConjP Q
Thus we find that, when combined with the statement of the CSC in (83), our analysis
accurately predicts the contrast between (79)/(80) and (81). Furthermore, this ac-
count successfully derives these data from the same condition responsible for the
similar, classic CSC facts in (82).
Finally, let us consider what our Q-based account predicts for non-Agreement
languages like Japanese. Clearly, since our account ties the impossibility of config-
urations like (84) to the necessity of Agreement between the Q and the wh-word, it
follows that such configurations should be licit in non-Agreement languages. It has
indeed been noted that wh-questions in Japanese seem to violate the CSC (Cheung
176 THE GRAMMAR OF Q
Assuming that the Q-particle ka in (86) is initially merged as an adjunct to the entire
coordinate structure, we find that our analysis of (79)–(81) provides a straightfor-
ward account of these facts.
Con jP
Con jP Q
ka
DP 1 to DP2
niku nani
Under this analysis, the Agreement between the interrogative C head and the
Q-particle ka does not violate the CSC, because the Q-particle is not contained within
a conjunct of the coordinate structure. Moreover, no violation of the CSC is incurred
by the presence of only a single wh-word within the coordinate structure, because
there is no Agreement between Q and the wh-word in Japanese.
We find, then, that when combined with an Agreement-based statement of the
CSC (83), our Q-based theory of pied-piping can account for the interesting contrast
between the English data in (79)–(82) and the Japanese data in (86).
This section examines cases where a wh-word undergoes movement within the front-
ed phrase of a pied-piping structure, a phenomenon dubbed ‘secondary wh-fronting’
by Heck (2008). I first present the proposed Q-based theory of secondary wh-front-
ing, illustrating the account with examples from English. Following this, I argue that
possessor pied-piping and preposition inversion in the Mayan languages Chol and
Tzotzil provide interesting evidence in support of our Q-based account.
seen that these locality restrictions often prevent a wh-word W occupying a posi-
tion A inside a larger phrase B from pied-piping B. Interestingly, many
languages possess a clever ‘technique’ for nevertheless allowing W to pied-pipe
B. In these languages, W may be moved from its base position A to a higher po-
sition inside B, from which it can pied-pipe B without violating the locality re-
strictions on pied-piping. For example, the ill-formed structures in (88a) and
(89a) become well-formed if the pied-piper is located at the left edge of the front-
ed phrase.
Following Heck (2004, 2008), I will use the term ‘secondary wh-fronting’ to
describe cases where a wh-word has undergone movement inside a pied-piping
structure.
In this section we will develop a Q-based account of this phenomenon, one
in which the ‘engine’ driving secondary wh-fronting is the need for the wh-word
to be in a position where it is visible for Agreement with the Q-particle. For ex-
ample, we already possess within our account the means for deriving the pattern
in (88). We have already seen in section 5.2.2 that the ill-formedness of (88a)
follows from the impossibility of Q/wh-Agreement when the wh-word is com-
plement to NP (90a). Moreover, we also saw that the well-formedness of (88b)
follows from the possibility of Q/wh-Agreement when the wh-word occupies
SpecDP (90b).
QP
DP Q
D NP
N PP
Pictures
of who XXXXX
No Q/Wh-Agreem ent
Across a Lexical Projection
178 THE GRAMMAR OF Q
DP Q
DP1 DP
Whose D NP
N t1
pictures
Thus, in this sense, the need for Q in English to Agree with a wh-word ‘drives’ the
secondary wh-fronting of possessors.
In a similar way, the obligatory ‘DegP inversion’ exemplified in (89) could result
from the need to position the DegP so that it may Agree with its c-commanding
Q-particle. The ill-formedness of sentences like (89a) would follow from the wh-
word being buried within the modified NP, as illustrated in (91a). However, as shown
in (91b), the secondary fronting of the wh-word in (89b) would put the wh-word in a
position where it is visible for Agreement with Q.
DP Q
D NP
a
DegP NP
Deg AP boat
how
big
XXXXXXXXXX
No Q/Wh-Agreement Across a Lexical Projection
DP Q
DegP1 DP
Deg AP D NP
How a
big t1 NP
boat
It seems, then, that our Q-based theory could view ‘secondary wh-fronting’ as a
means for facilitating Agreement between the Q-particle and the wh-word it
c-commands.
Note, however, that there is a certain lacuna in this proposed account. While the
account would predict that wh-words internal to lexical projections must move in
order to pied-pipe, we have not yet accounted for the possibility of such movement.
Now, in certain cases this gap is innocuous, since the phrase-internal movement is an
independent, freely available option in the language. For example, the general ability
for possessors in English to move to SpecDP accounts for the possibility of the sec-
ondary wh-fronting in (88b) and (90b). In other cases, however, this gap is a bit more
problematic, particularly when the phrase-internal movement undergone by the pied-
piper is not a freely available option. In fact, in some cases of secondary wh-fronting,
the phrase-internal movement of the pied-piper can only occur with wh-operators.
The secondary wh-fronting in (89b) and (91b) is a potential example of this. Although
such ‘DegP inversion’ is obligatory when the Deg head is a wh-operator, it is impos-
sible in most other cases.33
Our account of the data in (89b) is therefore somewhat incomplete, in that it lacks
any explanation of how the ‘rescuing’ movement in (91b) is triggered.
Let us, then, begin to fill this lacuna by developing a general approach to those
cases where the movement seen in secondary wh-fronting is only allowable for wh-
operators. We will take English DegP inversion as our paradigmatic example. Since
the fronting of the DegP in English is tied to it being a wh-operator, it would be nat-
ural to view this movement as triggered by the wh-feature. Let us, then, suppose that
English possesses a special indefinite article awh, which bears an unvalued instance of
the feature Wh.
Given our theory of feature valuation under Agree (section 5.2.1), when the deter-
miner awh is merged to its NP complement, it must Agree with something that
bears a valued instance of Wh. Assuming that wh-words all bear valued instances
of Wh, this Agreement can take place when the complement of awh contains the
Deg head how. Finally, as a reflex of the Agreement between awh and the DegP
headed by how, the DegP moves into the specifier of the DP. The following struc-
ture illustrates.
180 THE GRAMMAR OF Q
DP
DegP 1 DP
Deg AP D NP
How big awh
Agree
t1 NP
boat
Clearly, from the logic of this proposal, it follows that any DP-internal fronting of
DegP can only occur if the DegP is headed by a wh-operator.34 Thus our account
correctly predicts the impossibility of (92b).
Our discussion here nicely illuminates a rather subtle property of our Q-based
theory of secondary wh-fronting, one that warrants some spotlighting. Under the
account proposed above, the proximal cause of DegP inversion is an Agreement
relation holding between the fronted DegP and a special D-head bearing unvalued
Wh. Nevertheless, the necessity of DegP inversion in English ‘degree questions’
is due to the impossibility of Q/wh-Agreement without it. We find, then, that
while our theory does propose that secondary wh-fronting is a consequence of
obligatory Q/wh-Agreement, it is only an indirect consequence. That is, second-
ary wh-fronting is not directly triggered by Q/wh-Agreement under our account,
but is rather the structural precondition for such Agreement. This property also,
of course, holds for those cases of secondary wh-fronting where the phrase-inter-
nal fronting is a free, independently available option in the language. Even though
the phrase-internal fronting in these other cases is not triggered by the wh-fea-
ture, it is locally triggered by some feature of the wh-word. For example, in the
case of possessor fronting in English, we might suppose that the fronting of the
possessor to SpecDP is due to phi-Agreement between the possessive D-head and
the possessor.
In the following two subsections we will present some evidence in support of
this Q-based approach to secondary wh-fronting. Before we come to these argu-
ments, however, let us briefly note an interesting consequence of our proposals con-
cerning DegP inversion. Recall that our analysis derives the possibility of DegP
inversion from the existence of a special indefinite article awh, which is lexically
specified as bearing unvalued Wh. Our account therefore provides an explanation of
the impossibility of DegP inversion within other indefinite DPs. Note the following
contrast between (95a) and (95b,c).
The impossibility of (95b,c) would follow from the absence of putative somewh and
threewh in English. Interestingly, it is unclear how these facts could follow from any
semantic or syntactic difference between the normal indefinite article a and the indef-
inite determiners some and three. Thus it seems correct to tie the possibility of DegP
inversion to an idiosyncratic, unpredictable lexical property of the D-head rather than
view it as a productively driven reflex of pied-piping (cf. Heck 2004, 2008).
Crucially, as pointed out by both Coon (2009) and Aissen (1996), in examples
like (98a), there can be no pied-piping by the wh-possessor strictly within the larger
pied-piped possessive DP. That is, movement of the wh-word maxki, ‘who’, to the
left periphery of the possessive DP cannot pied-pipe the embedded NP headed by
iyotyoty, ‘house’. Such pied-piping would produce the ‘roll-up’ structure in (99),
which is judged in both languages to be ill-formed.
(99) No Pied-Piping of Embedded Possessive DP by Embedded Possessor
* [ [ Maxki1 [ iyotyoty t1 ] ]2 ijol t2 ] tyi puli?
who house roof burned
DP
DP
D NP
N DP
ijol
DP
D NP
N DP
iyotyoty maxki
XXXXXXXXXXXXX
IMPOSSIBLE PIED-PIPING
As first observed by Coon (2009), the contrast between (98a) and (99) is problematic
for accounts that treat both secondary wh-fronting and the fronting of pied-piped phrases
as instances of ‘wh-movement’. After all, it is clear from sentences like (97a) and (98a)
that wh-fronting in these languages generally allows wh-possessors to pied-pipe posses-
sive DPs. Why, then, can such pied-piping not occur when wh-fronting occurs internal to
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 183
a larger, pied-piped DP? As Coon notes, there is no straightforward answer within the
prevailing theories of pied-piping, which can only rule out the roll-up structure in (99) by
external stipulation.36 On the other hand, Coon (2009) argues that the Q-based theory of
wh-fronting advocated here predicts the impossibility of (99).
Abstracting away from Coon’s own particular implementation of the Q-based theory
for Chol, let us assume that the secondary wh-fronting found with Mayan wh-possessors
is, like English DegP inversion, an instance of ‘real’ wh-movement, triggered by unval-
ued Wh[ ] on a higher D-head. Following our general theory of secondary wh-fronting,
the appearance of this Wh-bearing D-head is required in order to put the wh-possessor in
a position where it can Agree with the Q-particle sitting above the possessive DP. As we
see in (100), if the Wh-D does not appear within a possessive DP, then the wh-possessor
will remain in a position where Q/wh-Agreement cannot occur.
QP
DP Q
D NP
N DP
Iyotyoty maxki
XXXXX
No Q/Wh-Agreement Across a Lexical Projection
On the other hand, the presence of Wh-D within a possessive DP entails that the
wh-possessor must front to SpecDP, which puts the wh-word in a position where
Q/wh-Agreement is possible.
QP
DP Q
DP 1 DP
Maxki
D Wh[ ] NP
N t1
iyotyoty
Under this analysis, structures like (98a), with ‘long-distance’ possessor fronting, are
derived via the insertion of multiple instances of DWh[ ]. Such long-distance fronting
of the wh-possessor is likewise required in order put it in a position where no lexical
projections intervene between it and the Q-particle.
DP Q
DP 1 DP
Maxki
D Wh[ ] NP
N DP
ijol
t1 DP
Successive- DWh[ ] NP
Cyclic Wh-
Movement N t1
iyotyoty
Importantly, however, our Q-based account is unable to derive roll-up structures like
(99). Recall that under our Q-based theory, there are no real instances of pied-piping.
Whenever a phrase larger than the projection of the wh-word is fronted, such fronting is
ipso facto not triggered by the wh-feature of the wh-word, but rather by the Q-features of
a c-commanding Q-particle. Thus the only conceivable means by which the roll-up struc-
ture in (99) could be created is by QP-movement internal to the pied-piped DP.
(103) Roll-Up Structures Like (99) Within the Q-Based Account
QP3
DP Q3
QP 2 DP
DP Q2 D NP
DP 1 DP N t2
Maxki ijol
iyotyoty t1
Wh-Movement to Spec DP
QP-Movement to Spec DP
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 185
QP1
ZP
QP2 ZP
YP Q2 t2
W
wh-word
Pied-piped YP Internal
to pied-piped XP
As with the particular structure in (103), the problem with (104) is semantic. If W
pied-pipes XP, then XP must be sister to a Q-particle Q1 which takes as argument the
focus-alternatives that are ultimately contributed by W. Moreover, if W also pied-
pipes YP, then YP must also be sister to a Q-particle Q2 which takes as argument the
focus-alternatives that are contributed by W. However, given that Q2 is located inside
XP, it follows that the focus-alternative contributed by W will not be able to ‘project’
up to XP, and thus Q1 will not receive an argument of the appropriate semantic type.
A semantic crash results, and the structure is predicted to be deviant.
We find, then, that our Q-based theory rightly predicts the impossibility of the
‘roll-up’ structure in (99). More generally, it predicts that secondary wh-fronting—
wh-fronting of a pied-piper internal to the pied-piped constituent—should never
itself exhibit pied-piping. Importantly, these predictions are inextricably linked to the
186 THE GRAMMAR OF Q
distinction our theory makes between the type of movement seen in secondary wh-
fronting (real wh-movement) and the type undergone by the pied-piped phrase
(Q-movement).
Moreover, this P-inversion can only target the wh-possessor of the possessive DP.
The entire possessive DP can never itself undergo movement to the left of the prepo-
sition.
Besides being an intrinsically interesting pattern, the Mayan data in (105) and
(106) raise the following challenge to theories of secondary wh-fronting. Although
this pattern is fully general for both Chol and Tzotzil, it is notably absent from other,
more familiar wh-fronting languages. For example, although Russian permits both
possessor extraction (66a) and PP pied-piping (20b), it does not require, or even al-
low, the kind of P-inversion seen in (105a).
Thus a full account of the Mayan P-inversion seen in (105a) should also predict the
absence of such secondary wh-fronting in languages like Russian.
Interestingly, our Q-based theory of secondary wh-fronting provides an analysis
of the Mayan pattern in (105) that successfully ties its P-inversion to an indepen-
dently visible property of these languages, one that, moreover, distinguishes them
from Russian and most other wh-fronting languages. A striking property of both
Chol and Tzotzil is that neither permits P to take full DPs as complements. Rather, as
described by Jessica Coon (personal communication), Ps in these languages must
take bare NP complements.40
As shown in (108) and (109), the Chol preposition tyi, ‘to’, can only take as comple-
ment bare NPs like otyoty, ‘house’, and ja’, ‘river’. If tyi, ‘to’, takes as complement
an unambiguous instance of a Chol DP, such as ili otyoty, ‘this house’, or ixä ja’, ‘that
river’, then the resulting structure is ill-formed.
This inability to take full DP complements distinguishes the Ps of Chol and
Tzotzil from the Ps of Russian and most other languages. Structures like those in
(110) demonstrate that the prepositions of Russian can take full DPs as comple-
ments.41
As we will see, our Q-based account can derive from this basic difference between
Mayan and Russian prepositions the contrast seen between (105) and (107). To see
this, let us first examine how our Q-based theory would analyze the P-inversion of
Chol and Tzotzil.
To begin, let us consider the potential consequences of the fact that Chol and Tzotzil
P cannot take full DP complements. From this fact, it follows that possessive nominals that
188 THE GRAMMAR OF Q
are complement to P must be bare NPs, and cannot have the DP functional projections of other
possessive nominals.
PP
P NP
tyi
N DP
possessum
possessor
PP Q
DP1 PP
Maxki PWh[ ] NP
tyi
N t1
iyotyoty
Secondary Wh-Fronting
We find, then, that our Q-based account can derive the necessity of P-inversion
with Chol/Tzotzil wh-possessors from the inability for P in these languages to take
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 189
full DP complements. Furthermore, our system correctly predicts that this P-inver-
sion will never target the entire possessive NP itself (106). Structures like (106)
would necessarily be instances of ‘roll-up pied-piping’, as in (103), which we have
shown to be ruled out by our account. Because the phrase moved to SpecPP in (106)
is not a projection of the wh-word, it follows that such fronting is not wh-movement,
but rather Q-movement. Thus such a structure would necessarily involve the interpo-
sition of a Q between the wh-possessor and the Q external to the pied-piped PP.
Consequently, (106) is predicted to be uninterpretable.
Let us now examine the predictions our account makes for PP pied-piping in
Russian. As demonstrated by (110), it is possible for Ps in Russian to take full DPs
as complements. Thus, as the structure in (113) illustrates, it should be possible for
wh-possessors in Russian to pied-pipe a larger PP without having to move to the
Spec of PP.42
(113) P-Inversion Is Not Necessary for PP-Pied-Piping by Wh-Possessors in Russian
QP
PP Q
P DP
v
DP NP
ey mashyne
In the previous section we examined cases where the locality constraints on pied-
piping in Q/wh-Agreement languages are obviated by movement of the wh-word
within the pied-piped phrase. Interestingly, many Q/wh-Agreement languages pos-
sess a further means for obviating their stricter locality constraints on pied-piping. As
has been observed numerous times in the literature, there appear to be environments
where these stricter locality constraints are relaxed. For example, it has often been
observed that the constraint against pied-piping from CompNP is weaker in matrix
questions than in subordinate questions (Ross 1979; Sells 1985; Webelhuth 1992;
Kayne 1994; Heck 2008).
Following Heck (2004, 2008), I use the term “massive pied-piping” to refer to struc-
tures like (114a) and (115a), where the stricter constraints of the limited pied-piping
languages can (marginally) be violated. In this last section we will explore how this
phenomenon can be approached within our Q-based theory.
As first articulated by Heck (2004, 2008), any theory of massive pied-piping
should accomplish two highly important goals. First, it should offer a generalization
characterizing the environments where massive pied-piping can occur. Secondly, it
should explain why massive pied-piping can occur in precisely those environments.
Although numerous authors had observed the basic phenomenon of massive pied-
piping, it was not until Heck (2004, 2008) that a general theory meeting these two
basic criteria was developed. Indeed, Heck (2004, 2008) is the first work to both (i)
observe that the weakening of the constraints on pied-piping in matrix clauses is
identical to the weakening of those constraints in nonrestrictive relatives, and
(ii) provide a unified account of these contrasts.
A major insight of Heck’s (2004, 2008) theory of massive pied-piping is his
characterization of the environments where it can take place. This characterization is
summarized in (116).
That the generalization in (116) predicts the contrast in (114) between matrix and
subordinate questions is rather obvious. Somewhat less obvious, however, is that it
also predicts the contrast in (115) between restrictive and nonrestrictive relative
clauses. In order to derive this contrast from (116), Heck (2008) assumes an analysis
of nonrestrictive relatives where they are ‘parentheticals’, and thus not truly subcon-
stituents of the larger sentence where they are interposed. Under such an analysis,
nonrestrictive relatives are therefore CPs that are not dominated by any other phrasal
nodes, and so classify as being nonsubordinated CPs. Thus (116) predicts that they
should also permit massive pied-piping.
Given its ability to unify the phenomena in (114) and (115), let us therefore adopt
the generalization in (116) as the proper characterization of the environments permit-
ting massive pied-piping. Of course, we must now ask why this generalization should
hold. Why should the stricter locality constraints in Q/wh-Agreement languages appear
to be (marginally) weaker in nonsubordinated CPs?
Although I am unable to offer an entirely principled account, I would like to
observe here that several properties of massive pied-piping can be seen to follow
from the stipulation in (117).43
That is, I claim that ‘massive pied-piping’ occurs when the Q-particle of a Q/
wh-Agreement language is (marginally) permitted not to undergo Agreement with
any wh-word. Under this view, the stipulation in (117) trivially predicts that massive
pied-piping will be restricted to nonsubordinated CPs (116).
Importantly, however, (117) can also capture several other properties of massive
pied-piping. First, we derive the fact that the constraint against pied-piping from
CompNP is weakened in nonsubordinated CPs ((114) and (115)). Recall that the ill-
formedness of (114b) and (115b) is due to (i) the requirement that Q Agree with a
wh-word within its domain, and (ii) the inability for Q to Agree with a wh-word
buried inside a lexical projection. Given the stipulation in (117), however, it follows
that Q/wh-Agreement can (marginally) fail to take place in matrix questions and
nonrestrictive relatives.
As illustrated in (118), (117) allows the use of valued iQ[+] in matrix clauses. Con-
sequently, Agreement between Q and the wh-word in the pied-piped phrase need not
occur. Therefore, the fact that the wh-word is buried within a lexical projection in
(118) does not impact the well-formedness of the clause.
We find, then, that (117) predicts the contrasts seen in (114) and (115). We also
predict that in such nonembedded environments, it should be possible for Q and the
192 THE GRAMMAR OF Q
We have just seen that (117) predicts that pied-piping past NP should be
marginally possible in nonsubordinated CPs. Generally speaking, however, our
reasoning demonstrates that any constraint on pied-piping that derives from con-
ditions on Q/wh-Agreement should be weaker in nonsubordinated CPs. Thus we
should find that for all lexical categories LP, it is marginally possible to pied-pipe
past LP in such environments. The following show that this prediction is borne
out for VPs.
Furthermore, data like the following show that our prediction is also borne out for
APs.
Finally, recall that the possibility of P-stranding in English led us to propose in chap-
ter 4 that P in English is a lexical head. Consequently, the stipulation in (117) predicts
that pied-piping of PP in English should be noticeably better in nonsubordinated
clauses than in embedded questions and restrictive relatives. As the following data
demonstrate, this prediction is borne out.
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 193
The data in (114)–(125) support our prediction that the prohibition against pied-
piping past lexical categories will be weakened in the nonsubordinated CPs of limited
pied-piping languages. Furthermore, recall from section 5.3.5 that the sensitivity of
limited pied-piping to the CSC derives from the fact that Q/wh-Agreement is governed by
the CSC. It follows, then, that (117) predicts that pied-piping in nonsubordinated clauses
will marginally permit violations of the CSC. This prediction appears to be borne out.
Thus far we have seen a variety of data demonstrating that, in the limited pied-
piping languages, pied-piping in nonsubordinated clauses is marginally less constrained
than pied-piping in subordinate clauses. One might naturally ask, then, whether pied-
piping in nonsubordinate clauses is entirely unconstrained in these languages. That is,
one might conclude from the data above that all constraints on pied-piping are weak-
ened in nonsubordinated contexts. Such a conclusion, however, would be incorrect. As
the following data illustrate, even massive pied-piping is still subject to the constraint
that the highest verbal category of a main clause cannot be pied-piped (section 5.3.2).
As the data indicate, there is no sense in which the sentences in (128) and (129) are
even marginally acceptable in colloquial English. This is in interesting contrast to
sentences like the following.
194 THE GRAMMAR OF Q
that they are a general property of massive pied-piping, then they stand as a difficult
challenge to the Q-based account offered here.
It has long been noted that English does not permit massive pied-piping past
specifiers (Nanni & Stillings 1978; Ishihara 1984; Sells 1985; Heck 2008). Other-
wise acceptable massive pied-piping structures in English are ill-formed if a specifier
c-commands the wh-word within the fronted phrase.45
Similar to this is the observation that wh-words in a specifier position cannot mas-
sively pied-pipe (Nanni & Stillings 1978; Kayne 1983; Sells 1985; Heck 2008). That
is, if a wh-word is contained within a specifier inside a phrase XP, then XP cannot be
massively pied-piped by that wh-word.46
Finally, we might note a third generalization, similar to that in (133). Data like that
in (134) suggest that English does not permit massive pied-piping by adjuncts. That
is, if a wh-word is contained within an adjunct inside a phrase XP, then XP cannot be
massively pied-piped by that wh-word.
The first thing to note about the contrasts in (132)–(134) is that none of them
currently follow from our Q-based theory of massive pied-piping. Nothing within
our Q-based theory would rule out structures like the following, where the Q-particle
undergoes no Q/wh-Agreement with the wh-word.
Interestingly, while our Q-based account does not yet predict these facts, we might
nevertheless discern a wider generalization from which they would follow. Note
that if we abandon the distinction between specifiers and adjuncts (Kayne 1994),
all three generalizations may be derived from the following, more general state-
ment.
196 THE GRAMMAR OF Q
QP
XP Q
… wh-word…
QP
XP Q
…wh-word…
CONSTRAINTS ON PIED-PIPING AND SECONDARY WH-FRONTING 197
English does not permit massive pied-piping past adjuncts. It is worth noting that the
relative clause island in (140) is an adjunct to NP that it modifies. Thus the impossi-
bility for wh-words to massively pied-pipe relative clause islands in English could be
seen as a particular case of the pattern in (134). Finally, since (134) is itself a conse-
quence of (136), we find that the pattern in (140) might ultimately follow from that
latter stipulation.
In summary, the generalizations regarding massive pied-piping in (132)–(134),
(139), and (140) do not to follow from any independently known principles of our
Q-based theory. Nevertheless, we have seen that each of them would follow from the
broader generalization in (136). Unfortunately, however, the broader condition in
(136) remains a bald stipulation, and is irreducible to anything independently known
about Q-particles. Moreover, (136) seems to be rather parochial to English, given that
pied-piping in non-Agreement languages like Tlingit is rather obviously not subject
to it. For the moment, then, we find that our Q-based theory has no special insight to
offer into the question of why English massive pied-piping is subject to the conditions
above. Consequently, I consider the facts in (132)–(140) to be a difficult empirical
challenge for our Q-based theory of massive pied-piping.
6
Conclusion
6.1 Introduction
CP
QP1 CP
Complementation
XP Q
CQ IP
Agree/
… wh-word… Attract
QP1
Overt Movement
199
200 THE GRAMMAR OF Q
Of course, there are many open problems and challenges that our Q-based account
faces, several of which were encountered in chapter 5. The solution to these problems
is far from trivial, and quite a few seem to directly undermine the strength of our
proposals. However, I hope that the preceding chapters have shown that the analysis
in (1) is a viable and interesting hypothesis, one that has rather far-reaching conse-
quences for the general theory of movement.
In this final chapter we will briefly explore two additional ways in which the
project begun with this study could be continued further. We begin in section 6.2,
with an examination of the consequences our account has for the analysis of other
types of A-bar movements. Following this, we briefly examine the way in which our
Q-based theory might inform the study of free relatives.
The central theoretical claim of this book has been that various phenomena surround-
ing wh-fronting directly result from the properties of Q-particles rather than from
properties of the movement relation itself. Three prominent examples of such phe-
nomena are (i) the inability to strand adpositions, (ii) the inability to extract posses-
sors and determiners, and most notably, (iii) the existence of pied-piping structures.
One of our central goals throughout has been to argue that these three phenomena do
not, contrary to perception, reflect general properties of movement per se.
It is important to note, however, that wh-fronting is not the only movement con-
struction to exhibit these three properties. Indeed, it seems that any form of A-bar
movement can be seen to exhibit them. For example, in English, we find that focus-
movement is also able to ‘pied-pipe’.
In the underlined portion of sentence (2), the possessive phrase Dave’s book un-
dergoes focus-movement. Importantly, however, the preceding context makes clear
that only the subconstituent Dave bears focus in this sentence. If we assume that the
focus-movement in (2) is triggered by the focus feature, then (2) is a structure where
the fronted phrase properly contains the projection bearing the features triggering the
fronting. Thus sentence (2) seems to show that focus-movement allows ‘pied-piping’.
Similarly, from the ill-formedness of sentences like (3), we find that focus-movement
is unable to extract possessors.
In the noun phrase in (4), the relative clause has been formed by fronting the entire
possessive whose father. Again, if the fronting seen in a relative clause targets the
features of the relative operator, then structures like (4) demonstrate that relativiza-
tion permits pied-piping. Similarly, the ill-formedness of structures like (5) show that
relativization in English is unable to extract possessors.
Similarly, because English relativization evinces the properties seen in (4) and (5),
we will assume that such constructions are formed via movement of a QRELP, which
dominates the relative operator.
The man [CP [QP [DP [ whose ] father ] QREL ]1 [IP I met t1 ] ]
Within the broader contours of our Q-based account, the analyses in (6) and (7)
would predict that both focus-movement and relativization should exhibit the three
properties of (i) inability to strand adpositions, (ii) inability to extract possessors and
determiners, and (iii) possibility of pied-piping structures.
The analyses in (6) and (7) illustrate a more general approach to A-bar move-
ment that our Q-based theory entails. Although the arguments throughout this book
center on wh-fronting, our Q-based theory should be more broadly understood as a
general theory of all those movement constructions that exhibit, for example, pied-
piping. Of course, we will have to leave to future study the full development of
Q-based accounts of these other movement constructions. Nevertheless, it should be
apparent that our Q-based theory of wh-fronting need not be an embarrassing step
backwards from the results gained by Ross (1967). Rather, it invites a new under-
standing of (some of) those generalizations, one where they are not properties of the
movement relation per se, but rather of the preconditions for movement in many of
the most well-studied movement constructions.
Although the analyses sketched above demonstrate how our Q-based theory
might handle the facts in (2)–(5), we should nevertheless ask if there is any indepen-
dent evidence for these Q-based analyses. Furthermore, what role do the putative Qs
play in the semantics of these movement constructions? What, exactly, is the QFOC
‘doing’ in a focus-movement construction?
The Q-based analysis of focus-movement in (6) receives some independent sup-
port from recent work by Horvath (2000, 2007b).1 This work is principally concerned
with the question of whether the feature ‘Focus’ is present in the syntactic representa-
tion of a sentence. Importantly, however, the main empirical claim of Horvath (2000,
2007b) is that so-called focus-movement does not actually target the features of the
focused phrase inside the fronted constituent. Rather, such movement is triggered by
CONCLUSION 203
As we see above, the preceding question indicates that the capitalized phrases in (8)
are understood as bearing focus. Moreover, the ill-formedness of (8b) demonstrates
that such focused phrases cannot occupy postverbal positions in Hungarian. Rather,
they must undergo focus-movement to a preverbal focus position, as in (8a). Thus
Hungarian seems to be a language where focused phrases must obligatorily undergo
focus-movement.
Because of the ubiquity of focus-movement in Hungarian, it has often been cited
by linguists as evidence that movement can be triggered by the feature ‘Focus’. How-
ever, as Horvath (2000, 2007b) argues at length, Hungarian ‘focus-movement’ does not
seem to be triggered by the feature ‘Focus’ per se. That is, contrary to the widespread
impression resulting from facts like (8), not all focused phrases in Hungarian can un-
dergo so-called focus-movement. For example, a focused DP that associates with the
focus-sensitive operator még. . . is ‘even’ cannot undergo the movement seen in (8).
(9) Foci Associating With ‘Even’ Cannot Undergo Focus-Movement (Horvath 2007b)
a. Mari elkésett még [ AZ ESKŰVŐJERŐL ] is.
Mary.NOM she.was.late yet the her.wedding.from also
Mary was even late to HER WEDDING.
b. * Mari még [ AZ ESKŰVŐJERŐL ] is késett el
Mary.NOM yet the her.wedding.from also she.was.late
As we see in (9), even though the phrase ‘to her wedding’ is focused in (9a), it need
not undergo focus-movement to the preverbal focus position. Indeed, as we see in
(9b), such a focused phrase cannot undergo such movement, despite the fact that it
bears the feature ‘Focus’.
Further evidence that so-called focus movement does not directly target focus
comes from wh-questions. Contrary to the common understanding of the facts in
(8), it is not always the case that the focused answer to a wh-question in Hungarian
must undergo focus-movement. Indeed, if the wh-question is given a so-called
204 THE GRAMMAR OF Q
‘mention-some’ reading, where the answer to the wh-question need not be exhaustive,
then the focused phrase in the answer cannot undergo the movement seen in (8).
Agreement/Movement
According to this analysis, the movement in (8a) actually targets the features of the
EI-OP, which c-commands the focused phrase. Because of the semantics of EI-OP,
the fronted phrase in (8a) is understood to exhaustively identify the true answers to
the question. Of course, just as with its overt cousin czak ‘only’, a focused phrase in
Hungarian need not necessarily be c-commanded by EI-OP. Therefore Horvath’s
analysis in (11) predicts that some focused phrases in Hungarian need not undergo
‘focus-movement’, and that such unmoved foci will be interpreted nonexhaustively.
Thus the analysis in (11) correctly predicts that so-called focus-movement corre-
lates not with focus per se, but with an exhaustive interpretation of the focused
phrase.
Finally, what about the prima facie appearance that focus-movement is triggered
by ‘Focus’? According to Horvath (2000, 2007b), this is ultimately due to the special
relationship between EI-OP and Focus. Because EI-OP is simply a null version of
csak ‘only’, it is a focus-sensitive operator. Therefore it must always c-command
some focused phrase, and so movement of the EI-OP will always entail movement of
a focused phrase. Therefore, because of the phonological invisibility of the EI-OP,
linguists have incorrectly concluded that the movement triggered by its features are
triggered by ‘Focus’ instead.
It is interesting to note the strong similarity between the analysis in (11) and
our own Q-based theory of focus-movement in (6). Indeed, if we were to simply
identify our QFOC with Horvath’s EI-OP, the two analyses would be the same. More
acutely, if we assume that QFOC possesses the exhaustive semantics of only (and the
EI-OP), then our analysis in (6) could account for the Hungarian facts in precisely
the way that (11) does. In this sense, then, we find that the evidence discussed by
Horvath (2000, 2007b) provides interesting support for our Q-based analysis in (6).
Furthermore, Hovath’s (2000, 2007b) analysis potentially answers the question of
what role the putative ‘QFOC’ plays in the semantics of the focus-movement con-
struction.3
More generally, we have seen in this section that it is possible and productive to
view other A-bar movements as also being instances of the QP-movement postulated
for wh-fronting in (1). Under such analyses, the A-bar movement in question is ulti-
mately phrasal movement of a Q-particle. Importantly, such analyses allow us to
provide a uniform account of the facts in (2)–(5), one where they receive the same
analysis as the parallel facts for wh-fronting discussed throughout this work. Thus
our Q-based theory need not commit us to the patently false claim that these prop-
erties will hold only for wh-fronting.
206 THE GRAMMAR OF Q
Another area where our Q-based account might yield results is the theory of so-called
free relatives.
A fundamental question in the theory of free relatives concerns the apparent dis-
connect between their internal syntactic form and their external syntactic distribution.
Across languages, free relatives tend to have the internal appearance of subordinate
questions rather than of adnominal relatives (Rooryck 1994; Jacobson 1995; Bury &
Neeleman 1999; van Riemsdijk 2007; but cf. Citko 2009). For example, in English,
the wh-word what can function as the wh-operator in a free relative (12a) and in a
subordinate interrogative (12b), but not in a normal adnominal relative clause (12c).
Other languages provide more striking examples of the alignment between free rela-
tives and subordinate interrogatives. Indeed, one rather clear example comes from
Tlingit. The following sentences illustrate the free relatives, subordinate questions,
and adnominal relatives of Tlingit.
As discussed in detail by Cable (2006a), there are numerous ways in which Tlingit
free relatives (13) pattern with subordinate interrogatives (14) rather than adnominal
relatives (15). The most obvious concerns the presence of a wh-operator. As we see
in (13), free relatives in Tlingit are formed via movement of an overt wh-operator.
Of course, such an overt wh-operator is also a crucial component of Tlingit subordi-
nate interrogatives (14). Importantly, however, overt wh-operators do not appear in
Tlingit adnominal relatives like (15). Rather, adnominal relatives in Tlingit are
formed via null relative operators, and do not contain the overt wh-operators of
either free relatives or subordinate questions. Furthermore, it should be stated that
CONCLUSION 207
the verb in a Tlingit free relative bears morphology that is otherwise characteristic
of verbs in subordinate interrogatives, and it lacks the morphology characteristic of
verbs in adnominal relative clauses. I refer the reader to Cable (2006a) for more
details.
We find, then, that free relatives across languages have the internal syntactic
form of subordinate questions. Crucially, however, free relatives do not have the
external syntactic distribution of subordinate questions. Generally speaking, the
external distribution of a given free relative will match that of the phrase fronted
into its left periphery (Bresnan & Grimshaw 1978; Groos & van Riemsdijk 1981).
For example, when the fronted phrase of the free relative is a DP, the free relative
will be able to appear in only those positions where a DP is permitted, as illustrated
in (16).
In the sentences in (16), the fronted phrase in the free relative is the DP whatever city.
Consequently the entire free relative can follow the verb visit, which selects for a DP,
but not the verb go, which requires an adverbial. On the other hand, if the fronted
phrase of the free relative is a PP, then it will have the opposite distribution.
In (17), the fronted phrase in the free relative is the PP to whatever city. Consequently
this free relative cannot be complement to the verb visit (17a), which selects for a DP,
but it can appear with the verb go, which can be modified by an adverbial PP (17b).
In summary, free relatives exhibit the following puzzling combination of prop-
erties: they possess the internal syntactic form of wh-questions, but they have the
external syntactic distribution of whatever phrase is fronted into their left periphery.
Understanding this combination of properties is the most central problem in the the-
ory of free relatives. How can something with the internal appearance of a subordi-
nate question have the external distribution of whatever happens to be in its left
periphery?
There are, of course, a wide variety of answers to this fundamental question
(Bresnan & Grimshaw 1978; Groos & van Riemsdijk 1981; Rooryck 1994; Jacob-
son 1995; Bury & Neeleman 1999; van Riemsdijk 2007; Citko 2009). Nevertheless,
our Q-based theory of wh-questions in (1) might provide a novel perspective on this
issue. First, recall that our QP-Intervention Condition from chapter 2 entails that
QPs are (in certain cases) ‘transparent’ for selection. That is, a core property of QP
projections is that they may intervene between heads and the phrases that those
heads select for. Now, let us entertain the hypothesis that, just as QPs are ‘transpar-
ent’ for selection, any phrasal node bearing the feature ‘Q’ will likewise be transpar-
ent. That is, let us suppose that any phrase XP bearing the feature ‘Q’ will be able
208 THE GRAMMAR OF Q
to intervene between a head H and the phrases selected by H.4 This idea is illustrated
in (18).
HP
H XPQ
YP XPQ
XQ ZP
Let us now consider what the picture in (18) would predict regarding subordi-
nate interrogative CPs. Following our analysis of wh-questions in (1), all such CPs
would bear the ‘Q’ feature. Thus the hypothesis in (18) entails that subordinate ques-
tions will be transparent for selection.
HP
H CPQ
YP CPQ
CQ ZP
VP
V CPQ
visited
QP CPQ
DP you went to
VP
V CPQ
went
QP CPQ
PP you went
In summary, the hypothesis in (18), combined with our Q-based theory of wh-
questions, predicts that subordinate questions should be able to exhibit the external
syntactic distribution of the phrases occupying their left periphery. Of course, the
very existence of free relatives could be taken as evidence that this prediction is
borne out. That is, the hypothesis in (18) could provide a new analysis of free rela-
tives, one that offers a unique answer to the ‘fundamental question’ regarding these
structures. Under this analysis, free relatives have the internal syntactic form of inter-
rogative CPs because they are interrogative CPs (Rooryck 1994; Jacobson 1995; van
Riemsdijk 2007). However, because such CPs bear instances of the feature ‘Q’, they
can exhibit the external syntactic distribution of their left-peripheral phrases. Thanks
to the special property of Q-particles illustrated in (18), the specifiers of Q-bearing
CPs are able to satisfy the selectional requirements of heads external to the CP. Con-
sequently, interrogative CPs will appear to ‘match’ the external behavior of their
left-peripheral specifiers.
Of course, any serious attempt to work out the analysis sketched here will have
to provide some explanation of the principle stipulated in (18). Assuming that this
can be done, we find that our Q-based theory in (1) might provide a syntactic theory
for free relatives, one where they may be analyzed simply as interrogative CPs, and
where their peculiar external distribution follows entirely from independently moti-
vated principles regarding Q.
This page intentionally left blank
NOTES
Chapter 1
1. As a side-note here, the reader should observe that the presence of the Q-par-
ticle sá can sometimes be obscured by its forming a portmanteau with the ‘focus
particles’ áwé, áyá, áyú, and áhé, the two surfacing together as sáwé, sáyá, sáyú, and
sáhé. This pattern can be seen, for example, in sentences like (1b).
2. Note that, following the definition in (7), I do not include under the rubric of
“pied-piping” all instances of phrasal movement. That is, I accept as uncontroversial
the existence of a mechanism of feature projection, which places the features of a
head onto the projections of that head. What is at issue is any mechanism that places
the features of a head onto nodes outside the projections of that head. This is a signif-
icant distinction, because feature projection is arguably indispensable, while the lat-
ter sort of devices are of little utility outside of deriving pied-piping structures.
3. Of course, such a language would exhibit one property that would saliently
distinguish it from the best-studied wh-fronting languages. Given the possibility of
structures like (10) in Tlingit, our hypothetical language would appear to allow
the pied-piping of islands, a pattern not permitted in the best-studied wh-fronting
languages. This variation will be discussed in chapter 5, where I propose that it
follows from a rather superficial difference in the morphosyntax of the languages’
wh-words.
4. On the other hand, one cannot dispute that there are extraction types besides
wh-fronting that are unable to extract from PPs or left branches. Indeed, this is the
ultimate reason why the inability to extract from those two environments is com-
monly thought to be a property of movement in general, and not simply some idio-
syncratic restriction on wh-fronting. One might worry, then, whether the account in
(15) does not incorrectly predict that extractions from PP and left branches should be
211
212 NOTES TO PAGES 11–22
well-formed for all other types of movement constructions, like focus-movement and
relativization. If so, then our account in (15) would seem to be a step backwards,
abandoning the immensely successful program of cross-constructional generaliza-
tions begun by Ross (1967).
Discussion of this important issue appears in chapter 6. There we will see that
the Q-based theory sketched in (15) can capture the cross-constructional ill-formed-
ness of such extractions by assuming that all the constructions in question also
involve some subvariant of the Q-movement illustrated in (15). That is, besides
the Q-particle found in wh-questions, there also exist separate, featurally distinct
instances of the category ‘Q’ in focus-movement constructions, relative clauses, etc.
As we will see later, this idea receives some independent support from recent work
on focus-movement by Horvath (2000, 2007b), who argues that so-called focus-
movement is actually movement of a (null) focus-sensitive operator sitting just above
the fronted phrase.
Chapter 2
1. Since the transliteration conventions for Sinhala can vary between authors, I
will briefly explain the conventions I follow here. Certain authors variably transliter-
ate the Sinhala character representing the vowel /a/, transliterating it as “ə” in con-
texts where the vowel is reduced, and as “a” in all other contexts. Thus certain authors
transliterate the Sinhala Q-particle as “də”. Throughout this book I follow the simpler
convention of uniformly transliterating this character as “a”; thus I transliterate the
Sinhala Q-particle as “da”. Aside from this, I make no changes in the transliterations
of the authors whose data I cite.
2. Throughout this book I follow Thompson (1996) in his use of the term ‘Na-
Dene’ to mean a language phylum containing Tlingit, Eyak, and Athabaskan. In other
work, this phylum is sometimes referred to as ‘Athabaskan-Eyak-Tlingit’ (Leer
2000), since the term ‘Na-Dene’ originally denoted a hypothetical (and likely erro-
neous) grouping that includes the neighboring language Haida.
3. Indeed, in some texts, (S)VO order slightly outnumbers (S)OV order. This is
reported by Dryer (1985) and also conforms to my own experience.
4. While the data in (5) were elicited, Cable (2007) also provides textually
attested examples of each of the six logically possible word orders.
5. Keri Edwards (personal communication) reports that some speakers find the
SVO and VSO orders in (5b,e) to be highly awkward, characterizing them as ‘back-
wards’ and only said in moments of great excitement. A generalization that would cover
this pattern might be that for such speakers, if the object follows the verb (V > O), then
the subject must follow the object (O > S).
6. The reader may note that the verbal form in (5b) differs from that in (5a). This
is due to a morphophonological rule that deletes third obviative object agreement when
the verb is directly preceded by an NP marked by the optional ergative postposition.
The effect of this rule appears in many of the examples throughout our discussion.
7. It is not known to me whether there is any difference in meaning or grammar
between aa and aadoo. Their origins are also unknown to me, as well as which—if
any—is the older form.
NOTES TO PAGES 22–29 213
1991). This idiom can also be interpreted as “X likes Y”, which may in fact be the
original meaning. This idiom has undergone a certain amount of grammaticalization
and phonetic reduction, but its original structure can be seen in sentences like the
following:
17. Sentence (25b) can reportedly be interpreted to mean “Our teacher told us
what we read.” Thus the asterisk here is intended only to represent that the sentence
cannot be interpreted as a matrix wh-question.
18. Throughout this book we will see that the syntactic label ‘Q’ must be under-
stood as an arbitrary name for a particular class of objects, ones that do not necessar-
ily have any inherent connection to ‘questions’ per se.
19. It should be noted, however, that there are a number of particles besides da
and ka with which wh-indefinites in Sinhala and Japanese may appear. This is not so
for Tlingit.
20. In highly colloquial Japanese, it is possible to drop ka in matrix wh-ques-
tions like (33a) (Lasnik & Saito 1992; Yoshida & Yoshida 1996). However, there are
stringent conditions governing this ‘particle drop’, and under one current account,
such sentences contain an unpronounced ka (Ko 2005).
21. One salient difference, however, between Tlingit sá and the other two parti-
cles is that sá can only appear in sentences containing wh-words. The particle sá sim-
ply has no use outside of its obligatory co-occurrence with wh-words. This is unlike
Japanese ka and Sinhala da, which can function both as markers of polar questions and
as disjunctive operators (Hagstrom 1998).
In this context, however, note that polar (yes/no) questions in Tlingit are formed
via insertion of the particle gé, as illustrated by the following sentence:
Thus, in Tlingit, wh-questions and yes/no questions are formed via two distinct parti-
cles. I suspend judgment here as to whether the particle gé should also be regarded as
an instance of Q. Nevertheless, given the distinction between gé and sá in Tlingit, I
assume that the use of da/ka in Sinhala/Japanese polar questions reflects the existence
of a separate, homophonous ‘yes/no’ particle. Thus the apparent difference noted
above might be only apparent, as the actual, underlying correlates of sá in Japanese
and Sinhala likewise appear only in wh-questions.
22. Note that sentences like (34f) demonstrate that subordinate CPs in Tlingit
may be pied-piped.
23. Japanese speakers tend not to allow the particle ka to be detached from the
wh-word when it functions as a wh-indefinite (cf. Yatsushiro 2001). In wh-questions,
NOTES TO PAGES 32–42 215
however, the particle ka must appear detached, where it c-commands the wh-word
from a clause-peripheral position. This ‘detachment’ of ka in wh-questions will be
analyzed in chapter 3. As for the condition that ka in wh-indefinites be as close to the
wh-word as possible, I currently have no principled account of this fact (which does
seem limited to the wh-indefinites of Japanese).
24. Because Japanese ka obligatorily appears at the end of the interrogative
clause, it cannot be easily determined whether this property also holds of the Japanese
Q-particle. However, Hagstrom (1998: 40) argues that the behavior of the emphasis
marker ittai in Japanese provides indirect evidence that it does. Thus, accepting the
evidence put forth by Hagstrom (1998), we find that this pattern also unites Japanese
ka with Sinhala da and Tlingit sá.
25. Like many languages, Tlingit possesses only a small, closed class of adjec-
tives, and so most nominal modification is accomplished with relative clauses. Thus
questions regarding the degree to which some NP possesses a given property (e.g.,
“How ADJ a NP”) must in Tlingit be asked using a structure in which the wh-word
is buried within a relative clause.
26. This property clearly does not hold of the Japanese particle ka.
27. Hagstrom (1998) and Kishimoto (2005) describe some limited cases where
Sinhala da may appear at the right edge of the matrix clause, although they disagree
about what characterizes these cases. We should also note here that Sinhala da can
generally appear matrix-finally in yes/no questions.
Recall from note 21, however, that I assume that the particles required in yes/no
questions are distinct (though possibly homophonous to) the Q-particles appearing in
wh-questions and wh-indefinites. Thus the behavior of da in Sinhala yes/no ques-
tions does not bear on the identity between Tlingit sá and the particle da in Sinhala
wh-questions. In this context, it is also worth noting that the Tlingit yes/no particle
gé can appear clause-finally.
28. Note that the Q-particle in (51) is not part of the functional projection of the
wh-word itself. As we saw in (37), the sister of Q may contain lexical heads selecting
for the wh-head. Thus the analysis in (51) must be distinguished from the competing
claim that wh-in situ involves pure ‘feature-movement’ of [+wh] (Chomsky 1995).
29. Again, we should note that, just as in (51), the Q-particle in (53) is not part
of the functional projection of the wh-word (see (34)). Thus the proposal in (53) must
be distinguished from the less interesting claim that the wh-feature of a wh-word
heads its own projection within the functional projection of the wh-word.
30. Note that the contrast between (62) and (63) also effectively refutes the analysis
in (56).
31. One might offer the following objection to the idea that no syntactic relation
may cross an island. Given that wh-words in Tlingit obligatorily co-occur with
Q-particles, there is presumably some syntactic relation between them. Therefore the
216 NOTES TO PAGES 42–59
well-formedness of (62) indicates that at least this relation may cross an island. How-
ever, we will see in section 2.7 that there is not necessarily any syntactic relation
between Q and the wh-word, since their obligatory co-occurrence independently fol-
lows from their semantics.
32. Similarly, it is the contrast between sentences (38a) and (38b) that most
strongly motivates the Hagstrom/Kishimoto analysis of Sinhala wh-questions in
(51). In both cases, the fact that only the position of Q affects the well-formedness of
the question indicates that only Q bears a relation to the matrix interrogative C.
33. Sentences (83) and (84) reflect the well-formedness judgments provided by
Mary Anderson of Atlin to Roby Littlefield. Mary Anderson speaks the Interior Tlin-
git subdialect of Northern Tlingit.
34. Sentence (90) reflects the judgments provided by Mary Anderson of Atlin to
Roby Littlefield.
35. I believe, however, that some speakers greatly prefer the NPI reading of the wh-
indefinite, to the point that the NPI reading is essentially obligatory in the environments
that license it.
36. Keri Edwards (personal communication) reports that some speakers find the
translation of (100e) incorrect, and can only interpret it to mean “if any of Dave’s
in-laws win the lottery . . .”. Thus, for such speakers, it may indeed be that
wh-indefinites cannot scope out of adjunct islands. However, it may also be that such
speakers share the tendency, observed in note 35, to interpret wh-indefinites as NPIs
in those environments that allow such an interpretation, including the antecedent of
a conditional. Although the strong preference to interpret wh-indefinites as NPIs
should not be taken lightly, the data in (98)–(100) show that such interpretations are
(at least for some speakers) not obligatory.
37. Sandra Chung (personal communication) has suggested that the con-
straints governing the placement of Tlingit sá may be prosodic in nature. This par-
ticle is a clitic in Tlingit, and in other well-known languages, prosodic constraints
on clitics have a similar effect on their distribution. Most famously, in the South
Slavic languages, it has been proposed that prosodic constraints account for the
inability for second-position clitics to intervene between (a) possessors and pos-
sessed NPs, (b) determiners and their complements, and (c) prepositions and their
complements.
Although this is a highly intriguing suggestion that should be independently
pursued, I will put it aside here. As we will see, the account proposed here will be
able to capture facts about Tlingit that I believe such a prosodic account would be
unable to (e.g., the facts regarding the clause-final positioning of sá).
38. In this context, I use the term ‘intervene’ as follows: a QP intervenes between
F and XP if either (i) QP dominates XP but does not dominate F, or (ii) if QP domi-
nates F but does not dominate XP.
39. Although introduced here as a special stipulation, we will see later in this
section how this condition might follow from a more general hypothesis concerning
selection.
40. The reader may recall that this constraint also governs the Q-particle da in
Sinhala, a wh-in situ language. In the next chapter, we will see how our QP-based
theory of wh-questions can treat this fact.
NOTES TO PAGES 60–89 217
41. Note that there is a further problem with the structure in (115): the DP
projection appears to be exocentric, in as much as it does not immediately dominate
a D-head.
42. The meaning stated in (122) is only offered as an approximation of the seman-
tics of ‘only’. Although satisfactory for our purposes, the meaning in (122) is actually
too strong, as it would incorrectly predict that the truth of “only XP” rules out all the
logical entailments of XP.
43. Of course, this account renders problematic those languages where it seems
that wh-operators need not be obligatorily focused, such as English. As the reader
will later see, difficult questions will also arise regarding the tendency for wh-indef-
inites to be unfocused in languages such as German.
44. Properly speaking, although Hagstrom (1998) considers this proposal, he
ultimately rejects it, opting for a theory in which the Q-particles are operators while
the traces of the Q-particles are variables over choice functions.
45. Throughout this discussion I adopt the label cf as a means for abbreviating
the logical type of the choice function. Furthermore, I assume a cross-categorical
definition for choice functions of the kind used by Winter (1997).
46. Note, however, that this tree abstracts away from certain complexities con-
cerning the interpretation of extensional closure in an intensional system of the kind
we adopt here.
47. The fact that there is some a ∈ { x : x ∈ Dτ & P(x) } such that R = [B . . . a
. . . ] follows ultimately from the fact that wh-words in our system only ever seman-
tically compose with their sisters via pointwise semantic composition.
48. Recall that we observed earlier, under note 31, that this result undermines
one possible objection to our arguments for the analysis in (1)/(53).
49. I ignore here the complexities surrounding the indexical your in these sen-
tences.
50. The reader may note that (151) also assumes the Heim and Kratzer (1998)
theory concerning the interpretation of structures with movement. According to this
theory, movement of a phrase XP to a position Y results in the insertion of a lambda
operator directly below Y, binding the trace of the XP.
51. The abbreviation ‘FA’ here stands for the rule of ‘Function Application’.
52. The abbreviation ‘LC’ here stands for the rule of ‘Lambda Conversion’.
53. Multiple wh-questions will be treated in chapter 4.
Chapter 3
1. The term ‘Q-projection language’ will also be used to refer to all languages
where the Q-particle takes its sister as complement, including wh-fronting languages
like Tlingit.
2. On the other hand, it may be that a uniform account of these facts would not
be desirable. Kishimoto (2005) adopts a movement-based account of (9)–(11), assum-
ing the Q-adjunction analysis in (2). Recall that such a movement-based account was
rejected for Tlingit, largely because this pattern is also observed in Tlingit wh-indefi-
nites. Interestingly, Hideki Kishimoto (personal communication) reports that the pat-
tern in (9)–(11) is not observed for Sinhala wh-indefinites. When appearing with
218 NOTES TO PAGES 89–103
Sinhala wh-indefinites, the particle da can come between functional heads and the
phrases those heads select. Such facts indeed lend credence to a movement-based
account of (9)–(11), and so challenge the analysis proposed here.
3. It should be acknowledged that many have argued that wh-words in Japanese
(and Korean) undergo full phrasal movement in wh-questions. Such arguments chal-
lenge our analysis in (2). Hagstrom (1998), which is the original source for the analysis
in (2), includes a substantial discussion of the alleged evidence for wh-movement in
these languages, and shows how it might nevertheless be captured by the account in
(2). I refer the reader to Hagstrom (1998) for a thorough discussion of these important
issues.
4. The reader may note that the Q in (17) is separated from the wh-word by a
‘linking’ morpheme eyn. This ‘linker’ is obligatory in Korean wh-indefinites. Unlike
Japanese ka, Korean ka cannot directly combine with wh-indefinites. As the reader will
observe, there are other linking morphemes besides eyn, and the choice of morpheme
depends upon the syntactic position of the wh-indefinite.
5. Another possibility, mentioned by native speakers of both languages, is that
Q-particles in Japanese/Korean can only cliticize onto nominal categories, and the
words I identify above as D-heads (dono/enu) are actually adjectives.
6. After all, within a Bare Phrase Structure system, some principles must entail
that the phrase initially merging with a head H must be the internal argument of H,
rather than its external argument. Presumably these principles could also entail that
the only thing that may undergo initial merger with a head is its internal argument.
7. Note that (24) is agnostic as to the structural position (and nature) of the case
markers in Japanese, which can be assumed to be semantically vacuous.
8. The stipulation in (25) is not absolutely required for our system to correctly inter-
pret wh-questions with pure Q-movement. See Cable (2008) for an analysis where (25) is
dispensed with.
9. Example numbers here refer to those from chapter 2, where the rules in
question were introduced.
10. Also note that the equivalence of (27) and (28) follows from the general
equivalence proved in (135) in chapter 2.
11. Although the complement of–man in (34) does contain the trace of the fronted
wh-word, it might be assumed that such traces possess normal-semantic values as indi-
vidual variables.
Chapter 4
1. Of course, given that relative clause islands in Tlingit may be ‘pied-piped’
(chapter 2), it is apparent that the class of pied-piping structures are wider in Tlingit
than in languages like English, where structurally parallel sentences are not possible.
However, we will see in chapter 5 that this difference may be due to an independent
morphosyntactic difference between the wh-words of English and Tlingit.
2. In addition, Cable (2007) discusses several Tupí languages of Central and
South America, arguing that they are also wh-fronting languages with overt QPs. In
this context, one should also note the analysis of Finnish questions developed by Hol-
mberg (2008). Given that the behavior of Q-particles in Finnish questions is similar to
NOTES TO PAGES 103–105 219
However, as noted previously, Baker (1999) provides strong arguments against the
analysis of dè as a wh-D.
The following is a possible account of these facts that retains the parallelism
between Tlingit sá and Edo dè. Suppose that Edo dè differs from Tlingit sá in that
it need not take only focus-semantic values as its argument. That is, suppose that
Edo dè is interpreted as a choice function taking as argument the normal-semantic
value of its sister. Consequently dè could combine directly with set-denoting ex-
pressions like NPs. Under this view, sentences like (i) could be interpreted as
follows:
Note that such a meaning indeed reflects the English translation provided to (i).
A second possibility to consider is that Edo might simply possess a phonologi-
cally empty wh-determiner. Thus it may indeed be correct to assign Edo dè the exact
same meaning as Tlingit sá, with sentences like (i) receiving the following analysis:
From facts like this, Abels (2003) concludes that the ill-formedness of P-stranding
cannot be due to the general islandhood of CompPP.
8. On the other hand, it is indisputable that wh-fronting is not the only type of
extraction for which the positions in (11)–(13) seem to be islands. Indeed, this is a
central reason why the positions in (11)–(13) are thought to be general extraction
islands. For example, the inability to extract possessors holds not only for English
wh-fronting, but also for focus-movement, relativization, etc.
(i) I’ve read John’s book, but [ DAVE’s book ] I haven’t read.
One might rightly worry, then, whether our Q-based account is not too parochial, and
fails to derive the possibility of pied-piping structures in A-bar constructions beyond
wh-fronting.
This issue will receive extended discussion in chapter 6. There we will see how
our Q-based account can be extended to other types of A-bar movement that are
observed to permit pied-piping structures. In general, the idea we will pursue is that
NOTES TO PAGES 118–123 221
mentioned in note 8: the movements in question are all subvariants of the Q-move-
ment seen in wh-questions.
11. Certain earlier work, however, does assume theories of agreement where
such ‘feature transfer’ could take place (e.g., Grimshaw 2000). Furthermore, Pesetsky
and Torrego (2001) put forth an Agreement-based theory of pied-piping that may
circumvent this problem.
12. Heck (2004) also notes that such an account faces the conceptual problem
that it countenances ‘feature movement’, in the sense of Chomsky (1995). In most
current work, the possibility of ‘feature movement’ has been rejected in favor of
mechanisms like long-distance Agree.
13. Of course, under our Q-based account, the facts in (38) do not follow from
constraints on movement per se. However, anyone adopting our Q-based account of
(38) would ipso facto be rejecting the ‘feature movement’ analysis of pied-piping in
(37), and vice versa.
14. To be precise, Heck (2004, 2008) proposes a hybrid theory, where the labor
of deriving pied-piping structures is divided between a limited mechanism of feature
percolation (identified as ‘feature movement’) and a limited degree of ‘nonlocality’
between the moved phrase and the feature inside it promoting the movement. Such a
theory is able to avoid the problems faced by accounts that appeal to only one of
these two general forms of analysis.
15. Note that this is also a prima facie problem for our Q-based theory of pied-
piping structures, one that will receive extensive attention in chapter 5.
16. The semantic issues discussed here are treated in greater depth by Cable
(2007). They are also discussed by Sternefeld (2001).
17. As we will discuss later, there are some environments in English where wh-
words are subject to Intervention Effects. For our current purposes, however, we will
temporarily put such cases aside.
18. As we will discuss later, there are some environments in English where
multiple wh-questions seem not to be subject to Superiority. For our current pur-
poses, however, we will again temporarily put such cases aside.
19. I should note here, however, that the empirical status of Superiority Ef-
fects in German remains controversial. For example, as first argued by Fanselow
(1991, 1997), Superiority Effects do seem to surface in German if the structurally
lower wh-word undergoes long-distance movement from a subordinate clause.
That is, sentences like (i(b)) are decidedly ill-formed in comparison to sentences
like (i(a)), a fact that is quite reminiscent of the English facts in (42).
However, Pesetsky (2000: 76–83) argues that structures like (i(b)) are ill-formed for
reasons independent of Superiority.
222 NOTES TO PAGES 123–132
On the other hand, Featherston (2005) reports that German speakers do voice
a slight preference for (43a) over (43b). I assume here that such preferences indi-
cate that Superiority-violating structures like (43b) are slightly more difficult to
process than those satisfying Superiority (43a), as independently argued by Sag
et al. (2006). However, I assume such differences in processing ease are a dis-
tinct phenomenon from the categorical grammaticality judgments in (42), which
indicate that Superiority-violating structures are not generated by English
grammar.
20. While Pesetsky (2000) makes an explicit connection between Intervention
and Superiority Effects, this claim is not made explicitly by Beck (1996). Neverthe-
less, Beck’s (1996) analysis of the contrast between (40) and (41) does implicitly
relate this contrast to that in (42) and (43). Note that Beck (1996) ties the presence of
Intervention Effects in a language to the presence of (overt) A-scrambling. Interest-
ingly, the presence of A-scrambling is claimed by Fanselow (1991, 1997) to entail an
absence of Superiority Effects. Thus, taken together, Beck (1996) and Fanselow
(1991, 1997) share the prediction that Intervention Effects and Superiority Effects in
multiple wh-questions should be mutually exclusive.
21. Note that the structure in (51) assumes, as is commonly done, that all the
wh-words of an English multiple wh-question move into the left periphery by LF.
22. Note that the equivalence between (48c) and the formula derived in (51)
also follows from the more general equivalence proven in chapter 2 under (135) and
(136).
23. Another important difference between (56) and (49) is that ‘ForceQ+’ in (56)
combines with the focus-semantic value of its complement. Thus, unlike ‘ForceQ2’ in
(49), this ‘ForceQ+’ is itself a focus-sensitive operator.
24. Moreover, the reader may note that the ForceQ+ head is sufficient to interpret
wh-questions containing any number of wh-words. Thus, unlike English ForceQ2,
German ForceQ+ is sufficient to interpret ternary wh-questions, and generally any
wh-question containing more than two wh-words.
25. In this way, the account we offer here is akin to that of Pesetsky (2000),
which similarly derives the variation in (40)–(43) from a hypothesized difference in
the range of interrogative C heads that German and English possess.
26. For a more complete explanation of the term ‘D-linked’, I refer the reader to
Pesetsky (1982, 1987, 2000). Note that Pesetsky (1982, 1987, 2000) assumes that
‘D-linkedness’ is a property of wh-words, while I assume here that it is a property of
wh-questions. Some evidence in favor of the latter notion is the fact, first observed by
Comorovski (1996), that violations of Superiority like those in (62a) are again im-
possible if one of the two wh-phrases is replaced with a simple wh-word.
These facts cohere best with an account similar to the one offered here, where
D-linkedness is a global property of the wh-question, and are more difficult to cap-
ture if ‘D-linking’ is assumed to be a local property of wh-words.
NOTES TO PAGES 132–144 223
27. Note that we do not state that ForceQ-Dlink appears in all D-linked wh-ques-
tions, only that ForceQ-Dlink cannot appear in a question unless the question is D-linked.
This distinction is crucial for predicting the data in (63). Since, as described above,
the presence of ForceQ-Dlink is the only means by which a (binary) wh-question in
English can violate Superiority, that head must be present in sentences like (63a).
Consequently, in such sentences, in situ wh-words will be subject to Intervention
Effects. However, for D-linked wh-questions where Superiority is respected (63b),
nothing in the structure of these sentences entails that they contain ForceQ-Dlink. Thus
such questions could instead contain ForceQ2, and we correctly predict that their in
situ wh-words are not subject to Intervention Effects.
If our system required instead that all D-linked wh-questions contain the head
ForceQ-Dlink, then the difference between (63a) and (63b) could not be captured.
28. As observed in note 24, ForceQ+ is sufficient to correctly interpret multiple
wh-questions containing any number of wh-words.
29. Note, however, that it would fail to fully predict the data in (66). Crucially it is
only when nonbinary wh-questions violate Superiority that their in situ wh-words are
subject to Intervention Effects. The account proposed here, however, wrongly predicts
that the in situ wh-words of any nonbinary multiple wh-question will be subject to
Intervention Effects, and so it wrongly predicts that (66a) should be uninterpretable.
30. To be precise, Beck (2006) does put forth an account of the data in (67). Beck
(2006) proposes that in such separation structures, the wh-word is interpreted as part
of the IP-internal DP remnant. Such a lower interpretation of the wh-word would
indeed allow our account to derive these data as a subcase of the phenomenon in (40).
Unfortunately this analysis would rely upon a special, sui generis stipulation regard-
ing the interpretation of separation structures.
31. We should also note that our theory of I-Intervention Effects correctly pre-
dicts the well-formedness of structures like (69).
32. The judgments reported in (75)–(82) summarize those collected from five
native speakers of English. Only one of these speakers failed to perceive a contrast
between (75a, b, and c) or between (77a, b, and c).
Chapter 5
1. Properly speaking, it is only Grimshaw (2000) that explicitly states this gen-
eralization. Webelhuth (1992) states that there is no pied-piping past theta-assigners,
while Cowper (1987) states that there is no pied-piping past categories that can be
lexically specified as ‘[+Wh]’. However, in the context of their respective theories,
these latter two generalizations are equivalent to there being no pied-piping past N,
V, or A.
2. One might legitimately doubt whether this is the correct generalization to
draw from the handful of data in (7). I return to this issue in section 5.2.3.
3. Interestingly, the ill-formedness of subordinate questions like (7b) weakens if
they are used instead as matrix questions (i) or appositive relative clauses (ii).
In section 5.5 we will explore this phenomenon in more depth. For the moment,
however, I will follow Heck (2008, 2009) in temporarily putting aside such cases
of ‘massive pied-piping’. That is, I follow Heck (2008, 2009) in his assessment
that the pied-piping in (7b–d) is indeed ill-formed in English, and that there is
something special about matrix environments that (marginally) improves their ac-
ceptability.
4. Recall, however, our claim in chapter 4 that prepositions in English are actu-
ally lexical heads. Under this assumption, the generalization that English disallows
pied-piping past lexical categories would predict that English should disallow pied-
piping of PPs.
Although this might strike many readers as a blatantly false prediction, it has
often been noted that PP pied-piping is actually a rather marginal structure of English
(or, more accurately, those registers of English that permit P-stranding). The skepti-
cal reader is asked to suspend judgment until our discussion of ‘massive pied-piping’
in section 5.5. There we will see that our overall system actually makes the more
nuanced (and accurate) prediction that pied-piping of PPs in (colloquial) English is
noticeably better in matrix clauses than in subordinate clauses.
5. It is difficult to find more direct evidence that Tlingit permits pied-piping past
lexical categories. It is impossible to construct direct correlates to the English sen-
tences in (7b–d). First, Tlingit has only a very small number of (putative) ‘adjec-
tives’, none of which appear with complements. Furthermore, nominal arguments of
Ns in Tlingit do not appear to ever remain within the NP projection. Finally, Tlingit
has no process of VP-fronting.
On the other hand, clausal arguments of Ns like neek ‘rumor, news’ do seem to
stay inside the NP projection. Thus one could examine whether Tlingit permits pied-
piping past neek by a wh-word within its CP complement. If this is indeed the case,
it would provide additional evidence that Tlingit permits pied-piping past lexical
categories. As of yet, I have not been able to conduct this test with speakers.
6. Furthermore, Kratzer and Shimoyama (2002) propose that the presence of
[Q] on a language’s wh-word is tied to the phonological form of the wh-word. That
is, they claim that a language’s wh-words bear [Q] iff they all share some mor-
phophonological feature in common (e.g., the ‘wh-’ submorpheme of English wh-
words). Cable (2007) likewise adopts this proposal, and thereby attempts to predict
from the form of the wh-words whether a language is a ‘limited pied-piping
language’.
In this work, however, I will not adopt this assumption. That is, I will simply
stipulate for a given language whether its wh-words bear [Q]. I leave to future work
the question of whether this property of the wh-word can be predicted by indepen-
dent means.
7. The original formal implementation in Kratzer and Shimoyama (2002) is
couched in terms of “greedy” feature movement, and so is not compatible with the
syntactic system that I assume here.
8. This system for feature valuation is also based on earlier work by Brody (1997).
9. It may at first blush seem suspicious for a Q-particle to initially bear an
unvalued Q-feature. However, our proposals regarding Q are rather parallel to the
NOTES TO PAGES 146–152 225
proposals regarding tense in Pesetsky and Torrego (2007), where the tense node of
the clause initially bears unvalued tense, and receives its tense value only under
Agreement with the uninterpretable (but valued) tense feature of the verb.
10. It is interesting to note that, although Japanese is not a wh-fronting language,
the hypothesis in (14) predicts that, given its putative lack of Q/wh-Agreement (13),
it will permit wh-words to be dominated in the sister of Q by islands. This prediction
is borne out by the well-known possibility in Japanese wh-questions for the wh-
operator to be buried inside of an island.
Hagstrom (1998: 40–45) argues that such structures are derived via movement of the
Q-particle ka from a base position outside the island, as shown in (ii).
Thus, given the analysis in (ii), it appears that Japanese does allow wh-words to be
dominated by islands within the sister of Q, as predicted by our hypothesis in (14).
Furthermore, this fact raises the following, more general point. The ‘LF–pied-
piping’ analysis of sentences like (i) (Nishigauchi 1990) has often been criticized
on the grounds that its hypothesized pied-piping violates the observed constraints
on pied-piping in languages like English (Fiengo et al. 1988). However, the behav-
ior of pied-piping in Tlingit and the accompanying theory in (14) undermine these
criticisms. A proponent of Nishigauchi’s (1990) analysis could claim that the
differences between their hypothesized pied-piping and the pied-piping seen in
English follow simply from the fact that Japanese is not a Q/wh-Agreement
language.
11. A somewhat similar notion, that every nonphase is complement to a phase
head, appears in recent work of Boeckx (2009).
12. Note, however, that there is much controversy over the exact nature, and
therefore statement, of the PIC. For instance, Bošković(2005b) argues that Agreement
can apply across different spell-out domains. Some evidence in support of this
includes the phenomenon in Icelandic of verbal Agreement with nominative case-
marked objects:
13. However, for some recent thorough treatments, see Horvath (2007a) and
Heck (2008).
14. Curiously, we can see from sentences like the following (first introduced in
chapter 4) that the correlates of (21c) in German are well-formed.
226 NOTES TO PAGES 152–154
We will later see in section 5.4 how our Q-based theory of limited pied-piping licenses
the ‘DegP Inversion’ overtly seen in English (ii). That account will apply equally
well to the postulated German structure in (ii).
15. As reported by Coon (2009), this pattern is also found in the Mayan language
Chol. Moreover, Coon (2009) argues that a Q-based theory of pied-piping like the
one developed here can capture certain complex features of possessor pied-piping in
Chol. A related (although slightly different) account of the Chol facts will be pro-
vided in section 5.4.
16. As noted by Heck (2008: 243–244), there are some languages that permit
pied-piping by postnominal possessors. Greek, illustrated below, is one such
language.
Following Heck (2008), I will adopt the view that such languages allow SpecDP to
be rightward. That is, in these structures the postnominal possessor has not actually
remained within the lexical projection NP, but rather has moved up into a (rightward)
specifier of D.
17. However, it remains quite controversial whether these structures are indeed
instances of CP pied-piping, and thus possess the structure indicated in (26). In Eng-
lish, they appear to be much more limited than the Basque and Quechua structures in
(25). For example, the English structures seem to degrade when the ‘matrix’ subject
is an R-expression (i), and the ‘matrix’ verbs allowing such putative CP pied-piping
seem to be rather restricted ((ii) and (iii)).
NOTES TO PAGES 154–163 227
If these adverbials are contained within VP, then structures like (i) would seem to
falsify (50). However, another possibility is that they are adjuncts to vP, in which case
they are entirely consistent with (50):
228 NOTES TO PAGES 163–173
(ii) Fritz weiß [ [vP wie schön [VP geschrieben ] ] man haben muss . . .
The second counterexample comes from Basque. Heck (2008: 134) notes that it is
possible in Basque for complements of past participles to pied-pipe the entire parti-
ciple phrase.
Given our claim that Basque is a Q/wh-Agreement language (section 5.2.3), this
would appear to counterexemplify the generalization for Q/wh-Agreement languages
in (50).
There are two possible ways our Q-based account can approach these data. The
most radical is to adopt the position of Arregi (2003b) that Basque is, contrary to
appearance, not a wh-movement language. Another possibility is that examples like
(iii) involve movement of the wh-word to Spec-vP:
Such ‘secondary wh-fronting’, similar to that seen in Basque examples like (25a),
would position the wh-word within the same phase as the Agreeing Q-particle. Thus,
if such movement were to take place, our account would indeed predict the possibil-
ity of structures like (iii).
Of course, this analysis would raise the question of why such secondary wh-
fronting cannot save English structures like (41b–d). Given the theory of second-
ary wh-fronting presented later in this chapter, a mechanical solution would be to
stipulate that little-v/little-n/little-a generally cannot bear the feature ‘Wh[ ]’.
Basque’s divergent behavior would be due to its exceptionally allowing little-v to
bear ‘Wh[ ]’.
23. To my knowledge, all other purported functional heads (e.g., Foc, Agr, Num,
Loc) are a result of ‘exploding’ the five basic functional categories listed here.
24. The reader may observe that in both German (62) and Irish (63), there is
some kind of inversion between the wh-word and the preposition when the PP is
pied-piped. Our Q-based account has no special explanation of this inversion. Note,
however, that the example of Icelandic in (61) demonstrates that the optionality of PP
pied-piping is not necessarily tied to such inversion.
25. Similarly, Heck (2008: 146) presents evidence that CP pied-piping in Basque
(58) is an instance of ‘massive pied-piping’. Since Heck (2008) predicts that massive
pied-piping (unlike ‘normal’ pied-piping) is optional, he thereby captures the data in
(58) and (74).
26. Note that this ‘ambiguity account’ could also be adopted by Heck (2008) to
capture (75). Indeed, Heck (2008: 289) explicitly considers such approaches, label-
ing them ‘pseudo-optionality’.
27. Although our account does easily capture the cases of CP pied-piping in
(57)–(59).
NOTES TO PAGES 173–179 229
Heck (2008: 311) proposes that (i) and (ii) do not actually involve pied-piping of a
coordination, but instead involve coordination of two separate clauses, followed by
deletion within the first clause. It seems, though, that there is little justification for
treating (i) and (ii) as a separate phenomenon from that in (79)–(81).
31. Note, however, that this assumption is highly controversial, and is directly
challenged by the phenomenon of ‘First Conjunct Agreement (FCA)’. Babyonyshev
(1996) argues that FCA shows that Agreement is not sensitive to the CSC in the way
suggested by (83). On the other hand, Aoun et al. (1994, 1999) argue for a theory of
FCA that is consistent with our condition in (83).
32. However, Haraguchi (1973) observes that head-internal relative clauses in
Japanese seem to observe the CSC. Given the evidence that Japanese relatives are not
derived by movement (Matsumoto 1997; Davis 2006), the data Haraguchi (1973)
examines stand as a much broader puzzle.
33. Note, however, that this DegP inversion is also required for some Deg heads
other than the wh-head how. For example, demonstrative Deg heads in English must
also undergo this inversion.
We see then that, contrary to our later discussion, DegP inversion is not crucially tied
to the wh-feature of the Deg-head how. On the other hand, we will see in the follow-
ing sections some clearer cases where secondary wh-fronting is crucially tied to the
wh-feature. Since the facts surrounding these later examples are rather complicated,
I choose not to employ them in this introductory section. For that reason, I opt to
(somewhat misleadingly) use the simple case of DegP inversion in English to illus-
trate the basics of our account.
230 NOTES TO PAGES 180–185
34. Note, however, that something must also ensure that awh can only Agree with
DegPs, and not, for example, postnominal possessors. If awh could Agree with post-
nominal possessors, we would incorrectly derive structures like (i).
(ii) Mann [CP [CP den1 [ zu t1 überzeugen ] ]2 dir t2 vorzunehmen ]3 ich dir t3 empfehle ]man
who to convince you plan I you advise
Man that I advise you to plan to convince. (Heck 2008: 50)
In these structures, the relative pronoun den appears to first undergo secondary
wh-fronting to the left edge of the embedded infinitive zu überzeugen, ‘to
convince’. Importantly, the entire infinitive then seems to undergo secondary
whw-fronting to the left edge of the larger embedded infinitive dir vorzunehmen,
‘you plan’, from which position it then appears to pied-pipe the entire subordi-
nate clause den zu überzeugen dir vorzunehmen. It would appear, then, that these
German structures involve a ‘roll-up’ structure like the kind in (104).
NOTES TO PAGES 185–194 231
(iii) [CP [CP Hans1 [ zu t1 überzeugen ] ]2 dir t2 vorzunehmen ]3 ich dir t3 empfehle
Hans to convince you plan I you advise
I advise you to plan to convince Hans.
38. The Chol data here were provided to me by Jessica Coon (personal commu-
nication).
39. I thank Kirill Shklovsky for offering me his Russian judgments.
40. The Chol data here were provided to me by Jessica Coon (personal
communication). That Tzotzil behaves in a parallel fashion was reported to me by
Jessica Coon (personal communication), citing John Haviland (personal communi-
cation).
41. Note that this argument assumes that demonstratives in Russian must take
NPs as complements rather than adjoin to them, contrary to our earlier proposals in
chapter 4. We will later explain (in note 42) how the following account can be made
consistent with our earlier proposals concerning possessor extraction in Russian.
42. Observe, however, that the structure in (113) assumes that possessive nom-
inals in Russian can be DPs rather than bare NPs, contrary to our proposals in chap-
ter 4. On the other hand, we might follow the suggestion made in section 5.3.4 that
possessive nominals in Russian have the option of being DPs, as well as bare NPs.
However, the reader will observe that the logic of the following account assumes that
possessive nominals in CompPP are always DPs (and never bare NPs) in Russian.
We might render this assumption consistent with our wider proposals by assuming
that P in Russian selects for DPs. Consequently, possessive nominals in Russian
must be DPs when complement to P.
43. It should be noted that Heck’s own (2008) theory of massive pied-piping
requires a stipulation of equal complexity to (117) in order to derive generalization
(116). Although I do not wish to review the details of Heck’s account here, the inter-
ested reader should note that Heck (2008: 332) ultimately derives (116) from a stip-
ulation that features having undergone pure feature-movement are invisible for
selection (although not any other syntactic operations).
44. Curiously, such massive pied-piping does not seem to be possible when the
higher verbal category is an auxiliary.
I have no explanation of this apparent contrast between the pied-piping in (i) and that
in (130) and (131).
45. Heck (2008: 173–179) presents evidence that this constraint also holds in
Italian and Russian.
46. Heck (2008: 305) presents evidence that this generalization also holds for
Italian.
47. I use the term ‘path’ here in the sense of Pesetsky (1982). For a node A and
a node B, where A c-commands B, the path between A and B is the set of nodes
consisting of A, B, and the nodes dominating B but not dominating A.
48. As we see from (138), the chief benefit of our abandoning the distinction
between specifiers and adjuncts is that it permits us to derive both (133) and (134)
from the stipulation in (136). This abandonment, however, does come at a price. If
we do not distinguish adjuncts from specifiers, then the generalization in (132) pre-
dicts that massive pied-piping past adjuncts should be as ill-formed as massive
pied-piping past specifiers. As the following sentences show, however, this predic-
tion is not so.
Thus, unlike what we see for specifiers in (132), it is possible for massive pied-pipers
to be c-commanded by adjuncts. I currently have no solution for this straightforward
problem.
49. Admittedly, this is quite a ‘big if’, given the strong evidence that English
infinitivals contain phonologically null PRO in the subject position. For exactly this
reason, Heck (2008: 173) restricts his versions of (132)–(134) to overt specifiers. A
similar adjustment could be made to our (132)–(134) and (136). I have avoided doing
so simply to keep this exposition brief.
Chapter 6
1. Similarly, our Q-based analysis of relativization in (7) is independently pro-
posed by Sternefeld (2001). Sternefeld (2001) also develops a semantics for relative
clauses based on this syntax.
2. However, unlike our analysis in (6), Horvath (2000, 2007b) claims that this
focus-sensitive operator is adjoined to the fronted phrase, rather than heading that
phrase. Thus the analysis in Horvath (2000, 2007b) differs from (6) in that it must
still appeal to some mechanism of pied-piping.
3. Similarly, the work of Sternefeld (2001) demonstrates what role our putative
‘QREL’ might play in the semantics of relative clauses. We should note, however, that
as pointed out by Heck (2008: 39), it might be difficult to extend a Q-based account
to constructions such as topicalization, scrambling, and ‘Wackernagel-movement’ in
German, each of which exhibits pied-piping and an inability to strand prepositions.
The potential difficulty is that these movements seem not to possess any identifiable
semantics. Therefore a Q-based account would have to assume that the Qs associated
NOTES TO PAGES 205–208 233
with these movements are semantically empty, which seems to weaken the Q-based
theory.
4. Note that this hypothesis would not follow from the theory of the QP-Inter-
vention Condition sketched in section 2.6.3.1. Thus the following proposals must
assume a different theory of why the QP-Intervention Condition holds.
BIBLIOGRAPHY
Abels, Klaus. 2003. Successive cyclicity, anti-locality, and adposition stranding. Doctoral
dissertation. Storrs, CT: University of Connecticut.
Aissen, Judith. 1996. Pied piping, abstract agreement and functional projections in Tzotzil.
Natural Language and Linguistic Theory 14: 447–491.
Aoun, Joseph, Elabbas Benmamoun, and Dominique Sportiche. 1994. Agreement, word order
and conjunction in some varieties of Arabic. Linguistic Inquiry 25: 195–220.
Aoun, Joseph, Elabbas Benmamoun, and Dominique Sportiche. 1999. Further remarks on first
conjunct agreement. Linguistic Inquiry 30: 669–681.
Aronoff, Mark. 1976. Word formation in generative grammar. Cambridge, MA: MIT Press.
Arregi, Karlos. 2003a. Clausal pied-piping. Natural Language Semantics 11: 115–143.
Arregi, Karlos. 2003b. Focus on Basque movements. Doctoral dissertation. Cambridge, MA:
Massachusetts Institute of Technology.
Babyonyshev, Maria. 1996. Structural connections in syntax and processing: Studies in
Russian and Japanese. Doctoral dissertation. Cambridge, MA: Massachusetts Institute
of Technology.
Baker, C. L. 1970. Notes on the description of English questions: The role of an abstract ques-
tion morpheme. Foundations of Language 6: 197–219.
Baker, Mark. 1999. On the interplay of the universal and the particular: Case studies in Edo. In
Proceedings of the 35th Annual Meeting of the Chicago Linguistics Society, ed. by Sabrina
J. Billings, John P. Boyle, and Aaron M. Griffith. Chicago: University of Chicago Press.
Barss, Andrew, Ken Hale, Ellavina Tsosie Perkins, and Margret Speas. 1991. Logical form
and barriers and Navajo. In Logical structure and linguistic structure, ed. by C.-T. James
Huang and Robert May. Dordrecht: Kluwer Academic.
Beck, Sigrid. 1996. Quantified structures as barriers for LF movement. Natural Language
Semantics 4: 1–56.
Beck, Sigrid. 2006. Intervention effects follow from focus interpretation. Natural Language
Semantics 14: 1–56.
235
236 BIBLIOGRAPHY
Bennett, Paul. 1995. A course in Generalized Phrase Structure Grammar. London: University
College London Press.
Black, Lydia T., Nora Marks Dauenhauer, and Richard Dauenhauer. 2008. Classics of Tlin-
git oral literature, Volume 4: Anóoshi Lingít aaní ká, Russians in Tlingit America: The
Battles of Sitka, 1802 and 1804. Juneau, AK: Sealaska Heritage Foundation.
Boas, Franz. 1917. Grammatical notes on the language of the Tlingit Indians. University of
Pennsylvania University Museum Anthropological Publications 8:1.
Bobaljik, Jonathan. 2002. A-chains at the PF-interface: Copies and ‘covert’ movement. Natural
Language and Linguistic Theory 20: 197–267.
Boeckx, Cedric. 2009. On the locus of asymmetry in UG. Catalan Journal of Linguistics 8: 41–53.
Borsley, Robert. 1992. More on the difference between English restrictive and non-restrictive
relative clauses. Journal of Linguistics 28: 139–148.
Bošković, Željko. 2005a. On the locality of left branch extraction and the structure of NP.
Studia Linguistica 59: 1–45.
Bošković, Željko. 2005b. On the locality of Merge and Agree. Cambridge, MA: MIT Working
Papers in Linguistics.
Bresnan, Joan. 1976. On the form and function of transformations. Linguistic Inquiry 7: 3–40.
Bresnan, Joan, and Jane Grimshaw. 1978. The syntax of free relatives in English. Linguistic
Inquiry 9: 331–391.
Brody, Michael. 1997. Perfect chains. In Elements of grammar, ed. by Liliane Haegeman.
Dordrecht: Kluwer Academic.
Bury, Dirk, and Ad Neeleman. 1999. Projection of free relatives. Manuscript. University
College London.
Cable, Seth. 2006a. The interrogative words of Tlingit: An informal grammatical study.
Manuscript. University of Massachusetts, Amherst. Retrieved July 25, 2009, from http://
people.umass.edu/scable/papers/Tlingit-Interrogatives-Grammar.pdf.
Cable, Seth. 2006b. Syncope in the verbal prefixes of Tlingit: Meter and surface phonotactics.
Muenchen: Lincom Europa.
Cable, Seth. 2007. The grammar of Q: Q-particles and the nature of wh-fronting, as revealed
by the wh questions of Tlingit. Doctoral dissertation. Cambridge, MA: Massachusetts
Institute of Technology.
Cable, Seth. 2008. Question particles and the nature of wh-fronting. In Quantification: A
cross-linguistic perspective, ed. by Lisa Matthewson. North Holland Linguistic Series,
Linguistic Variations, Vol. 64. Bingly, UK: Emerald.
Campbell, Lyle. 1997. American Indian languages: The historical linguistics of Native
America. Oxford: Oxford University Press.
Carlson, Gregory N., and Francis Jeffry Pelletier. 1995. The generic book. Chicago: University
of Chicago Press.
Cheng, Lisa Lai-Shen. 1991. On the typology of wh-questions. Doctoral dissertation. Cambridge,
MA: Massachusetts Institute of Technology.
Cheung, Lawrence Y.-L. 2003. On certain (non-)violations of coordinate structure constraints.
Snippets 12: 9–11.
Chomsky, Noam. 1973. Conditions on transformations. In A festschrift for Morris Halle, ed.
by Stephen Anderson and Paul Kiparsky. New York: Holt, Reinhart & Winston.
Chomsky, Noam. 1975. Questions of form and interpretation. Linguistic Analysis 1: 75–109.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by step: Essays on
minimalism in honor of Howard Lasnik, ed. by Roger Martin, David Michaels, and Juan
Uriagereka. Cambridge, MA: MIT Press.
BIBLIOGRAPHY 237
Citko, Barbara. 2009. What don’t wh-questions, free relatives and correlatives have in common?
In Correlatives Cross-Linguistically, ed. by Anikó Lipták. Amsterdam: John Benjamins
Publishing.
Cole, Peter, and Gabriella Hermon. 1981. Subjecthood and islandhood: Evidence from Quechua.
Linguistic Inquiry 12: 1–30.
Cole, Peter, and Gabriella Hermon. 1998. The typology of wh-movement: Wh-questions in
Malay. Syntax 1: 221–258.
Comorovski, Ileana. 1996. Interrogative phrases and the syntax-semantics interface. Dordrecht:
Kluwer Academic.
Coon, Jessica. 2009. Interrogative possessors and the problem with pied-piping in Chol. Linguistic
Inquiry 40: 165–175.
Corver, Norbert. 1990. The syntax of left branch extractions. Katholieke Universiteit Brabant.
Corver, Norbert. 1992. Left branch extraction. In Proceedings of the Northeast Linguistics
Society 22. Amherst, MA: Graduate Linguistic Student Publications.
Corver, Norbert. 2007. Subextraction. In The Blackwell syntax companion (SYNCOM), ed. by
Martin Everaert and Henk van Riemsdijk. Oxford: Blackwell.
Cowper, Elizabeth. 1987. Pied-piping, feature percolation and the structure of the noun phrase.
Canadian Journal of Linguistics 32: 321–338.
Culicover, Peter. 1999. Syntactic nuts. Oxford: Oxford University Press.
Dauenhauer, Nora Marks, and Richard Dauenhauer. 1981. Because we cherish you: Sealaska
elders speak to the future. Juneau, AK: Sealaska Heritage Institute.
Dauenhauer, Nora Marks, and Richard Dauenhauer. 1987. Classics of Tlingit oral literature,
Volume 1: Haa shuká, Our ancestors: Tlingit oral narratives. Juneau, AK: Sealaska
Heritage Institute.
Dauenhauer, Nora Marks, and Richard Dauenhauer. 1990. Classics of Tlingit oral litera-
ture, Volume 2: Haa tuwunáagu yís, For healing our spirit: Tlingit oratory. Juneau, AK:
Sealaska Heritage Institute.
Dauenhauer, Nora Marks, and Richard Dauenhauer. 1994. Classics of Tlingit oral literature,
Volume 3: Haa kusteeyí, Our culture: Tlingit life stories. Juneau, AK: Sealaska Heritage
Institute.
Dauenhauer, Nora Marks, and Richard Dauenhauer. 2000. Beginning Tlingit. Juneau, AK:
Sealaska Heritage Institute.
Dauenhauer, Nora Marks, and Richard Dauenhauer. 2002. Lingít x’éináx sá! Say it in Tlingit:
A Tlingit phrase book. Juneau, AK: Sealaska Heritage Institute.
Dauenhauer, Nora Marks, and Richard Dauenhauer. Forthcoming. Intermediate Tlingit.
Juneau, AK: Sealaska Heritage Institute.
Davis, Christopher. 2006. Evidence against movement in Japanese relative clauses. Hand-
out from a talk presented at 2006 ECO5. University of Massachusetts, Amherst.
Retrieved July 25, 2009 from: http://people.umass.edu/cmdavis/papers/davis-
eco5-handout.pdf.
Dayal, Veneeta. 1996. Locality in wh-quantification: Questions and relative clauses in Hindi.
Dordrecht: Kluwer Academic.
Dryer, Matthew. 1985. Tlingit: An object-initial language? Canadian Journal of Linguistics
30: 1–13.
Dürr, Michael, and Egon Renner. 1995. The history of the Na-Dene controversy: A sketch.
In Language and culture in native North America: Studies in honor of Heinz-Jürgen
Pinnow, ed. by Michael Dürr, Egon Renner, and Wolfgang Oleschinski. Muenchen:
Lincom Europa.
Edwards, Keri. Forthcoming. Tlingit verb dictionary project. University of Alaska, Fairbanks.
238 BIBLIOGRAPHY
Edwards, Keri, John Marks, Nora Marks Dauenhauer, and Richard Dauenhauer. 2005. Lingít
x’éináx áx! Hear it in Tlingit: A Tlingit mini phrase book & CD. Juneau, AK: Sealaska
Heritage Institute.
Embick, David, and Alec Marantz. 2008. Architecture and blocking. Linguistic Inquiry 39:
1–53.
Emonds, Joseph. 1976. A transformational approach to English syntax. New York: Academic
Press.
Emonds, Joseph. 1979. Appositive relatives have no properties. Linguistic Inquiry 10:
211–243.
Enrico, John. 2003. Haida syntax. Lincoln, NE: University of Nebraska Press.
Fabb, Nigel. 1990. The difference between English restrictive and non-restrictive relative
clauses. Journal of Linguistics 26: 57–77.
Falk, Yehuda N. 2001. Lexical-Functional Grammar: An introduction to parallel constraint-
based syntax. Stanford, CA: CSLI Publications.
Fanselow, Gisbert. 1991. Minimale syntax. Habilitation thesis. University of Passau.
Fanselow, Gisbert. 1997. Minimal link effects in German (and other languages).” Manuscript.
University of Potsdam.
Featherston, Sam. 2005. Magnitude estimation and what it can do for your syntax: Some wh-
constraints in German. Lingua 115: 1525–1550.
Fiengo, Robert, C.-T. James Huang, Howard Lasnik, and Tanya Reinhart. 1988. The syntax of
wh-in-situ. In Proceedings of the Seventh West Coast Conference on Formal Linguistics
(WCCFL 7), ed. by Hagit Borer. Stanford, CA: CSLI Publications.
Fodor, Janet Dean, and Ivan A. Sag. 1982. Referential and quantificational indefinites. Lin-
guistics and Philosophy 5: 355–398.
Fox, Danny. 2000. Economy and semantic representation. Cambridge, MA: MIT Press.
Gazdar, Gerald, Ewan Klein, Geoffrey Pullum, and Ivan Sag. 1985. Generalized phrase struc-
ture grammar. Oxford: Blackwell.
Ginzburg, Jonathan, and Ivan Sag. 2000. Interrogative investigations: The form, meaning, and
use of English interrogatives. Stanford, CA: CLSI Publications.
Gordon, Raymond G., Jr. 2005. Ethnologue: Languages of the world, 15th ed. Dallas, TX:
SIL International.
Grimshaw, Jane. 1981. Form, function and the language acquisition device. In The logical
problem of language acquisition, ed. by C. L. Baker and John McCarthy. Cambridge,
MA: MIT Press.
Grimshaw, Jane. 2000. Locality and extended projections. In Lexical specification and in-
sertion, ed. by Peter Coopmans, Marin Everaert, and Jane Grimshaw. Amsterdam: John
Benjamins Publishing.
Groenendijk, Jeroen, and Martin Stokhof. 1982. Semantic analysis of WH-complements. Lin-
guistics and Philosophy 5: 175–233.
Groenendijk, Jeroen, and Martin Stokhof. 1984. Studies on the semantics of questions and the
pragmatics of answers. Doctoral dissertation. University of Amsterdam.
Groos, Anneke, and Henk van Riemsdijk. 1981. Matching effects in free relatives: A param-
eter of core grammar. In Theory of markedness in generative grammar, ed. by Andrea
Belletti. Pisa: Scuola Normale Superiora.
Hagstrom, Paul. 1998. Decomposing questions. Doctoral dissertation. Cambridge, MA:
Massachusetts Institute of Technology.
Hamblin, C. L. 1958. Questions. Australasian Journal of Philosophy 36: 159–168.
Hamblin, C. L. 1973. Questions in Montague English. Foundations of Language 10: 41–53.
Haraguchi, Shosuke. 1973. “Dragging” reconsidered. Linguistic Inquiry 4: 95–97.
Heck, Fabian. 2004. A Theory of Pied Piping. Doctoral dissertation. Universität Tübingen.
BIBLIOGRAPHY 239
Heck, Fabian. 2008. On pied-piping: Wh-movement and beyond. Berlin: Mouton de Gruyter.
Heck, Fabian. 2009. On certain properties of pied-piping. Linguistic Inquiry 40: 75–111.
Heim, Irene. 1994. Interrogative semantics and Karttunen’s semantics for ‘know.’ In Proceed-
ings of the 9th Annual Conference of the Israel Association for Theoretical Linguistics,
ed. by Rhonna Buchall and Anita Mittwoch. Somerville, MA: Cascadilla Press.
Heim, Irene, and Angelika Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell.
Holmberg, Anders. 2008. The syntax of the Finnish question particle. Manuscript. University
of Newcastle. Retrieved July 25, 2009, from http://ling.auf.net/lingbuzz/000596.
Horvath, Julia. 1986. FOCUS in the theory of grammar and the syntax of Hungarian.
Dordrecht: Foris.
Horvath, Julia. 2000. Interfaces vs. the computational system in the syntax of focus. In Inter-
face strategies, ed. by Hans Bennis, Martin Everaert, and Eric Reuland. Amsterdam:
Royal Netherlands Academy of Arts and Sciences.
Horvath, Julia. 2007a. Pied-piping. In The Blackwell syntax companion (SYNCOM), ed. by
Martin Everaert and Henk van Riemsdijk. Oxford: Blackwell.
Horvath, Julia. 2007b. Separating ‘focus movement’ from focus. In Phrasal and Clausal
Architecture: Syntactic Derivation and Interpretation; In Honor of Joseph E. Emonds,
ed. by Simin Karimi, Vida Samiian and Wendy K. Wilkins. Amsterdam: John Benjamins
Publishing.
Huddleston, Rodney D. and Geoffrey K. Pullum. 2002. The Cambridge Grammar of the Eng-
lish Language. Cambridge: Cambridge University Press.
Ishihara, Roberta. 1984. Clausal pied-piping: A problem for GB. Natural Language and Lin-
guistic Theory 2: 397–418.
Jackendoff, Ray. 1977. X-bar syntax: A study of phrase structure. Cambridge, MA: MIT
Press.
Jacobson, Pauline. 1995. On the quantificational force of English free relatives. In Quantifica-
tional in natural languages, ed. by Emmon Bach, Eloise Jelinek, Angelika Kratzer, and
Barbara Partee. Dordrecht: Kluwer Academic.
Jelinek, Eloise. 1984. Empty categories, case and configurationality. Natural Language and
Linguistic Theory 2: 39–76.
Karttunen, Lauri. 1977. Syntax and semantics of questions. Linguistics and Philosophy 1:
3–44.
Kayne, Richard. 1983. Connectedness. Linguistic Inquiry 14: 223–249.
Kayne , Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press.
Kayne, Richard. 2000. Parameters and universals. Oxford: Oxford University Press.
Kennedy, Christopher, and Jason Merchant. 2000. Attributive comparative deletion. Natural
Language and Linguistic Theory 18: 89–146.
Kishimoto, Hideki. 2005. Wh-in-situ and movement in Sinhala questions. Natural Language
and Linguistic Theory 23: 1–51.
Kiss, Katalin E. 1995. Discourse configurational languages. Oxford: Oxford University
Press.
Ko, Heejeong. 2005. Syntax of why-in-situ: Merge in [Spec, CP] in the overt syntax. Natural
Language and Linguistic Theory 23: 867–916.
Koopman, Hilda. 2000. Prepositions, postpositions, circumpositions and particles: The struc-
ture of Dutch PPs. In The Syntax of Specifiers and Heads: Collected Essays of Hilda J.
Koopman. London: Routledge.
Koopman, Hilda, and Anna Szabolcsi. 2000. Verbal complexes. Cambridge, MA: MIT Press.
Kratzer, Angelika. 1991. The representation of focus. In Semantics: An international hand-
book of contemporary research, ed. by Arnim von Stechow and Dieter Wunderlich. New
York: Walter de Gruyter.
240 BIBLIOGRAPHY
Kratzer, Angelika. 1998. Scope or psuedoscope? Are there wide-scope indefinites? In Events
in Grammar, ed. by Susan Rothstein. Dordrecht: Kluwer.
Kratzer, Angelika. 2006. Exhaustive questions. Handout from a talk presented at the Massachusetts
Institute of Technology on May 12, 2006. University of Massachusetts, Amherst.
Kratzer, Angelika, and Junko Shimoyama. 2002. Indeterminate pronouns: The view from
Japanese. In Proceedings of the 3rd Tokyo Conference on Psycholinguistics, ed. by Yukio
Otsu. Tokyo: Hituzi Syobo.
Kuno, Susumu, and Jane J. Robinson. 1972. Multiple wh-questions. Linguistic Inquiry 3:
463–487.
Lasnik, Howard, and Mamoru Saito. 1992. Move-alpha: Conditions on its application and
output. Cambridge, MA: MIT Press.
Leer, Jeff. 1991. The schetic categories of the Tlingit verb. Doctoral dissertation. University
of Chicago.
Leer, Jeff. 1993. The Tlingit anaphoric system and its roots in Na-Dene. Manuscript. Fair-
banks, AK: Alaska Native Language Center.
Leer, Jeff. 2000. The negative/irrealis category in Athabaskan-Eyak-Tlingit. In The Athabas-
kan languages: Perspectives on a Native American language family, ed. by Theodore
Fernald and Paul Platero. Oxford: Oxford University Press.
Leer, Jeff, Doug Hitch, and John Ritter. 2001. Interior Tlingit noun dictionary: The dialects
spoken by Tlingit elders of Carcross and Teslin, Yukon, and Atlin, British Columbia.
Whitehorse: Yukon Native Language Centre.
Lewis, David. 1970. General semantics. Synthese 22: 18–67.
Li, Charles N. 1976. Subject and topic. New York: Academic Press.
Matsumoto, Yoshiko. 1997. Noun modifying constructions in Japanese. Amsterdam: John
Benjamins Publishing.
Matthewson, Lisa. 1998. Determiner systems and quantificational strategies: Evidence from
Salish. The Hague: Holland Academic Graphics.
Matthewson, Lisa. 1999. On the interpretation of wide-scope indefinites. Natural Language
Semantics 7: 79–134.
May, Robert. 1985. Logical form: Its structure and derivation. Cambridge, MA: MIT Press.
Mithun, Mariane. 1999. The languages of native North America. Cambridge: Cambridge
University Press.
Naish, Constance. 1966. A syntactic study of Tlingit. Language data: Amerindian Series 6.
Dallas, TX: SIL International.
Nanni, Deborah, and Justine Stillings. 1978. Three remarks on pied piping. Linguistic Inquiry
9: 310–318.
Newman, Paul, and Martha Ratliff. 2001. Linguistic fieldwork. Cambridge; Cambridge University
Press.
Nichols, Johanna. 1986. Head-marking and dependent-marking grammar. Language 62:
56–119.
Nishigauchi, Taisuke. 1990. Quantification in the theory of grammar. Dordrecht: Kluwer
Academic.
Nyman, Elizabeth, and Jeff Leer. 1993. Gágiwdul.àt: Brought forth to reconfirm. The legacy of
a Taku River Tlingit clan. Fairbanks, AK: Alaska Native Language Center.
Pesetsky, David. 1982. Paths and categories. Doctoral dissertation. Cambridge, MA: Massachusetts
Institute of Technology.
Pesetsky, David. 1987. Wh-in-Situ: Movement and unselective binding. In The representation
of (in)definiteness, ed. by Eric Reuland and Alice G. B. ter Meulen. Cambridge, MA:
MIT Press.
Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, MA: MIT Press.
BIBLIOGRAPHY 241
Pesetsky, David, and Esther Torrego. 2001. T-to-C movement: Causes and consequences.
In Ken Hale: A life in language, ed. by Michael Kenstowicz. Cambridge, MA: MIT
Press.
Pesetsky, David, and Esther Torrego. 2007. The syntax of valuation and the interpretability
of features. In Phrasal and clausal architecture: Syntactic derivation and interpretation,
ed. by Simin Karimi, Vida Samiian, and Wendy K. Wilkins. Amsterdam: John Benjamins
Publishing.
Postal, Paul. 1972. Two remarks on dragging. Linguistic Inquiry 3: 130–136.
Quinn, Steven. 2007. Perseverance to do ‘Macbeth’ in Tlingit.” Associated Press article.
Retrieved July 25, 2009, from http://www.sealaskaheritage.org/news/news_article_tlingit
_macbeth_2_2007.htm.
Reinhart, Tanya. 1992. Wh-in-situ: An apparent paradox. In Proceedings of the Eighth Amsterdam
Colloquium, ed. by Paul Dekker and Martin Stokhof. Amsterdam: ILLC.
Reinhart, Tanya. 1995. Interface strategies. Utrecht: OTS Working Papers in Linguistics.
Reinhart, Tanya. 1997. Quantifier scope: How labor is divided between QR and choice func-
tions. Linguistics and Philosophy 20: 335–397.
Reinhart, Tanya. 2006. Interface strategies: optimal and costly computations. Cambridge,
MA: MIT Press.
Richards, Norvin. 1997. What moves where in which language? Doctoral dissertation. Cambridge,
MA: Massachusetts Institute of Technology.
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed. by
Liliane Haegeman. Dordrecht: Kluwer Academic.
Rooryck, Johan. 1994. Generalized transformation and the wh- cycle: Free relatives as bare
wh-CPs. Groninger Arbeiten zur Germanistischen Linguistik 37: 195–208.
Rooth, Mats. 1985. Association with focus. Doctoral Dissertation. University of Massachu-
setts, Amherst.
Rooth, Mats. 1992. A theory of focus interpretation. Natural Language Semantics 1: 75–116.
Ross, John Robert. 1967. Constraints on variables in syntax. Doctoral dissertation. Cambridge,
MA: Massachusetts Institute of Technology.
Ross, John Robert. 1979. Wem der Kasus schlagt. Linguistiche Berichte 63: 26–32.
Rullmann, Hotze, Lisa Matthewson, and Henry Davis. 2008. Modals as distributive indefi-
nites. Natural Language Semantics 16: 317–357.
Ruys, Eddy. 1992. The Scope of Indefinites. Doctoral dissertation. University of Utrecht.
Ruys, Eddy. 2000. Weak crossover as a scope phenomenon. Linguistic Inquiry 31: 513–539.
Safir, Kenneth. 1986. Relative clauses in a theory of binding and levels. Linguistic Inquiry 17:
663–689.
Safir, Kenneth. 2009. Viable syntax: Rethinking minimalist architecture. Manuscript. New
Brunswick, NJ: Rutgers University.
Sag, Ivan A., Inbal Arnon, Bruno Estigarribia, Philip Hofmeister, T. Florian Jaeger, Jeanette
Pettibone, and Neal Snider. 2006. Processing accounts for Superiority effects. Manu-
script. Stanford University. Retrieved July 25, 2009, from http://lingo.stanford.edu/sag/
papers/squib.pdf.
Samarin, William J. 1967. Field linguistics: A guide to linguistic fieldwork. New York: Holt,
Rinehart & Winston.
Sauerland, Uli, and Fabian Heck. 2003. LF-intervention effects in pied-piping. In Proceedings
of NELS 33, ed. by Makoto Kadowaki and Shigeto Kawahara. Amherst, MA: GLSA.
Schwarzschild, Roger. 2002. Singleton indefinites. Journal of Semantics 19: 289–314.
Sells, Peter. 1985. Pied-piping and the feature [WH]. Manuscript. Stanford University.
Sealaska Heritage Institute. 2003. Sealaska Heritage Institute website. Retrieved July 25,
2009, from http://www.sealaskaheritage.org.
242 BIBLIOGRAPHY
Williams, Frank, Emma Williams, and Jeff Leer. 1978. Tongass texts. Fairbanks, AK: Alaska
Native Language Center.
Winter, Yoad. 1997. Choice functions and the scopal semantics of indefinites. Linguistics and
Philosophy 20: 399–467.
Yatsushiro, Kazuko. 2001. The distribution of mo and ka and its implications. In Formal ap-
proaches to Japanese linguistics 3, ed. by Maria Christina Cuervo, Daniel Harbour, Ken
Hiraiwa, and Shinichiro Ishihara. MIT Working Papers in Linguistics 41. Cambridge,
MA: MIT Working Papers in Linguistics.
Yoshida, Keiko, and Tomoyuki Yoshida. 1996. Question marker drop in Japanese. ICU
Language Research Bulletin 11: 37–54.
This page intentionally left blank
INDEX
245
246 INDEX