Thinking Syntactically
TSA01 07/18/2005, 03:04PM1
Blackwell Textbooks in Linguistics
The books included in this series provide comprehensive accounts of some of the
most central and most rapidly developing areas of research in linguistics. Intended
primarily for introductory and post-introductory students, they include exercises,
discussion points and suggestions for further reading.
1. Liliane Haegeman, Introduction to Government and Binding Theory (Second
Edition)
2. Andrew Spencer, Morphological Theory
3. Helen Goodluck, Language Acquisition
4. Ronald Wardhaugh, Introduction to Sociolinguistics (Fifth Edition)
5. Martin Atkinson, Children’s Syntax
6. Diane Blakemore, Understanding Utterances
7. Michael Kenstowicz, Phonology in Generative Grammar
8. Deborah Schiffrin, Approaches to Discourse
9. John Clark and Colin Yallop, An Introduction to Phonetics and Phonology
(Second Edition)
10. Natsuko Tsujimura, An Introduction to Japanese Linguistics
11. Robert D. Borsley, Modern Phrase Structure Grammar
12. Nigel Fabb, Linguistics and Literature
13. Irene Heim and Angelika Kratzer, Semantics in Generative Grammar
14. Liliane Haegeman and Jacqueline Guéron, English Grammar: A Generative
Perspective
15. Stephen Crain and Diane Lillo-Martin, An Introduction to Linguistic Theory
and Language Acquisition
16. Joan Bresnan, Lexical-Functional Syntax
17. Barbara A. Fennell, A History of English: A Sociolinguistic Approach
18. Henry Rogers, Writing Systems: A Linguistic Approach
19. Benjamin W. Fortson IV, Indo-European Language and Culture: An
Introduction
20. Liliane Haegeman, Thinking Syntactically: A Guide to Argumentation and
Analysis
TSA01 07/18/2005, 03:04PM2
Thinking Syntactically
A Guide to Argumentation and Analysis
Liliane Haegeman
TSA01 07/18/2005, 03:05PM3
© 2006 by Liliane Haegeman
blackwell publishing
350 Main Street, Malden, MA 02148-5020, USA
9600 Garsington Road, Oxford OX4 2DQ, UK
550 Swanston Street, Carlton, Victoria 3053, Australia
The right of Liliane Haegeman to be identiﬁed as the Author of this Work has been asserted in
accordance with the UK Copyright, Designs, and Patents Act 1988.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or
otherwise, except as permitted by the UK Copyright, Designs, and Patents Act 1988, without the
prior permission of the publisher.
First published 2006 by Blackwell Publishing Ltd
1 2006
Library of Congress Cataloging-in-Publication Data
Haegeman, Liliane M. V.
Thinking syntactically : a guide to argumentation and analysis / Liliane Haegeman.
p. cm. — (Blackwell textbooks in linguistics ; 20)
Includes bibliographical references and index.
ISBN-13: 978-1-4051-1852-1 (hard cover : alk. paper)
ISBN-10: 1-4051-1852-0 (hard cover : alk. paper)
ISBN-13: 978-1-4051-1853-8 (pbk. : alk. paper)
ISBN-10: 1-4051-1853-9 (pbk. : alk. paper) 1. Grammar, Comparative and general—
Syntax. 2. Linguistic analysis (Linguistics) I. Title. II. Series.
P291.H234 2006
415—dc22
2005010992
A catalogue record for this title is available from the British Library.
Set in 10/13pt Sabon
by Graphicraft Limited, Hong Kong
Printed and bound in the United Kingdom
by T.J. International Ltd, Padstow, Cornwall
The publisher’s policy is to use permanent paper from mills that operate a sustainable forestry policy,
and which has been manufactured from pulp processed using acid-free and elementary chlorine-free
practices. Furthermore, the publisher ensures that the text paper and cover board used have met
acceptable environmental accreditation standards.
For further information on
Blackwell Publishing, visit our website:
www.blackwellpublishing.com
TSA01 07/18/2005, 03:05PM4
Contents
Preface and Acknowledgments vi
1 Introduction: The Scientiﬁc Study of Language 1
Discussion 1
Exercises 47
2 Diagnostics for Syntactic Structure 65
Discussion 65
Exercises 123
3 Lexical Projections and Functional Projections 155
Discussion 155
Exercises 210
4 Reﬁning Structures: From One Subject Position to Many 237
Discussion 237
Exercises 268
5 The Periphery of the Sentence 305
Discussion 305
Exercises 354
Bibliography 370
Index 381
TSA01 07/18/2005, 03:05PM5
Preface and Acknowledgments
The title of this book is Thinking Syntactically. As the title suggests, the focus of the
book is on “thinking about syntax.” Syntax is the component of linguistics that is
concerned with the way words are put together to form sentences. This book illustrates
one way of thinking about sentence formation.
The Goals of the Book
Over the years, many types of syntactic theories have been developed in an attempt
to explain how sentences are formed. An approach that has given rise to a lot of
exciting discoveries is the one initiated by the American linguist Noam Chomsky in
the 1950s and which is known as “generative grammar.” One of the properties of
generative grammar which I think makes it particularly attractive is that it uses a
methodology modeled on what is used in the natural sciences. Thus, generative
linguists try to “think” about syntax in a scientiﬁc way; they elaborate their analyses
using a scientiﬁc methodology. The emphasis on methodology entails that, when
confronted with a syntactic theory or a particular syntactic analysis, syntacticians
do not have to accept the proposals as they are, unthinkingly and blindly. Rather,
they can examine the logic behind the proposals, evaluate it, and decide on its merits.
Ideally, then, learning generative syntax should imply learning this way of thinking
about syntax. It should deﬁnitely not be rote learning. In practice, I feel syntax has
often been reduced to rote learning, and that is why I have written this book.
The goal of the book is not to present all the intricacies of one syntactic theory.
Rather, its aim is to reconstruct and to illustrate as explicitly as possible the thinking
behind generative syntax. In other words, the aim is to illustrate how to “think
syntactically.” Generative syntax is not a spectator sport, where you sit on the
sidelines and watch others perform. Rather, I would like to get you involved. I
would like you to enter the world and the mindset of the practicing generative
syntacticians, to think with them and follow the argumentation as it develops. For
instance, sometimes when arguing in favor of one analysis over another, syntacticians
will use arguments drawn from language data; such arguments are called empirical
arguments. At other times, the syntactician will use arguments which themselves are
TSA01 07/18/2005, 03:05PM6
drawn from the theory he or she is working in; such arguments are theoretical
arguments. Ideally, these empirical and theoretical arguments should converge, but
that is not always the case. In such circumstances, in order to evaluate one analysis
over another, it is important to be able to assess the nature of the argumentation
itself and to compare different arguments.
The result of working your way through this book should be that when you are
confronted with syntactic analyses you are able to evaluate the arguments that have
led to the analyses, to check the way the arguments have been built up, to examine
the argumentation. Indeed, observe in passing that the kind of rigorous thinking
explored here may well come in handy in everyday life, as, for instance, when you
are deciding who to vote for, whether to buy a house or to rent one, or which job to
apply for.
Another aspect that distinguishes this book from many introductions to generative
syntax is the kinds of examples used. Very often, syntactic analyses are based
on a small set of home-made examples, which seem to have little or no bearing on
any kind of language that we meet in everyday life. Though this is a perfectly
legitimate move and one that we will sometimes also adopt in this book, to the
beginning students of syntax such an approach to language may look rather dry and
totally irrelevant. Because of the exclusive use of artiﬁcial examples, a syntax course
often seems to belong in a separate world, unconnected to the daily linguistic reality.
In this book, there will be arguments based on home-made “artiﬁcial” examples,
but in addition we will also be using a lot of attested examples mainly taken from
recent journalistic prose. The reason for introducing such examples is to show how
concepts that are relevant to syntactic theory are not outside the real world, but,
rather, drawn from and part of the real world.
To my mind, thinking syntactically should not be conﬁned to syntax classes. It
should be a way of thinking that is available to you in your daily life, that makes
you curious about linguistic phenomena, that makes you interested in the language
used around you, and that even makes you more aware of the language you use
yourself. I hope that having worked your way through this book, you will have
acquired a new linguistic sensitivity, and that in everyday life you will recognize
certain patterns discussed in the book and that you will also spot new and different
patterns that would perhaps not be accounted for in the book. I hope that in the
latter case you become so intrigued by these new data that you will try to ﬁgure out
how these new data should be analyzed in terms of the system elaborated in this book.
In addition to the many attested examples, it will also often be necessary to
construct our own examples in order to test certain hypotheses. In the ﬁnal chapter
of the book we will pay some attention to how such examples are constructed.
Though most examples discussed in this book are drawn from English, there is
also material drawn from other languages. The goal is to show that just as we can
think in a formal way about the structure of English, we can do the same for other
languages. If you are a native speaker of a language other than English you are
encouraged to think about your own language in similar terms as those laid out in
the book.
Preface vii
TSA01 07/18/2005, 03:05PM7
The book does not aim at providing a complete survey of a particular theory.
Rather, it shows that a theory is the result of a particular way of thinking. But the
book also shows that the thinking is never ﬁnished. At the end of the book, we will
have outlined some components of a theory about sentence formation, but as will
become clear in the exercises throughout the book, there remain many questions
and problems, and the theory presented is by no means complete. However, this is
not only due to the limited scope of this introduction. Even if I had written a book
twice as long, and even if I had been able to incorporate all the current proposals in
syntactic theory, still, in a few months’ time, if not sooner, there would have come
along new proposals challenging some of the hypotheses presented here and invalidating
others. Syntactic research is a continuous and continuing enterprise shared by
many enthusiastic researchers across the world. If syntacticians really had already
formulated an exhaustive and perfect theory of sentence formation, if there really
were no questions left, then there would be no practicing syntacticians left, either.
The Organization of the Book
The exercises
The book contains ﬁve chapters, each elaborating a step toward the formulation
of a theory of sentence structure. With each chapter comes a set of exercises. The
exercise headings are accompanied by the abbreviations (T), (L), and (E). The
abbreviation (T) stands for “tie in,” and indicates that a particular exercise ties in
with the material in the preceding chapter. Tie-in exercises are signaled by footnotes
in the chapter. Whenever a footnote points toward an exercise, it means that the
exercise can be tackled at that point in the chapter. The abbreviation (L) stands for
“look ahead” and it signals that the material covered in the exercise will be taken
up in a later chapter of the book. Look-ahead exercises also contain cross-references
to the later point at which the material is tackled. The abbreviation (E) stands for
“expansion” and signals that the material covered in these exercises goes beyond
that covered in the book. Again references to further reading will be included
in them. Since the material contained in T-exercises has been covered in the text,
T-exercises will tend to be “easier” than L-exercises or E-exercises.
The format of some of the E-exercises and the L-exercises is quite different from
the standard exercise format that you may expect to ﬁnd in a textbook. In particular,
some exercises are longer, they contain lots of text, and they look more like workbook
sections. The reason why such discursive exercises have not been included
in the main body of the text is that they are only intended here as additional
illustrations of how certain issues are problematic and how they can be or have
been pursued using the argumentation developed in the associated chapter. These
discursive exercises typically will not offer an exhaustive or deﬁnitive treatment of
the issues in question. Rather, they illustrate how a hypothesis is challenged and
viii Preface
TSA01 07/18/2005, 03:05PM8
how it may have to be reworked in the light of new data or of new theoretical
proposals.
When, having worked your way through a chapter, you want a quick rehearsal
of the material in the chapter, you will probably mainly want to revise using the
T-exercises. If you want to know what is to come later in the book, you could also
try the L-exercises. If you want to discover more intriguing problems which go
beyond the discussions in the present book, you should try the E-exercises.
The footnotes in the chapters and in the exercises also contain references to the
scientiﬁc linguistics literature. However, for the student-reader many of the publications
referred to will be too advanced and too technical and they should not be
tackled until you have reached the end of the book. Some more accessible references
are pointed out when they are available.
The chapters
The ﬁrst chapter of the book offers an introduction to scientiﬁc methodology and
how it can be applied to the study of syntax. Among other things, this chapter
introduces the hypothesis that the meaning of a sentence is calculated on the basis
of its component parts and their relations in the structure. This hypothesis about
the mapping of form onto meaning will be one of our guidelines throughout the
book. The ﬁrst chapter also provides an overview of some patterns of question
formation in English and French.
Chapter 2 introduces the key tools for identifying the constituents of a sentence.
It is shown that two of the main constituents of the sentence are its subject and its
verb phrase. The verb phrase is a constituent whose head is a verb. It is a “projection”
of the verb. The verb denotes the action or state depicted by the sentence; it
has a lot of descriptive content and it is called a lexical head. The projection of the
verb is a lexical projection.
Chapter 3 shows how subject and verb phrase are related through a linking
element, the inﬂection of the verb. This chapter introduces the hypothesis that the
inﬂection of the ﬁnite verb heads its own projection. The inﬂection is a “functional”
head; it does not have the same kind of descriptive content as a lexical head. Projections
of functional elements are called functional projections.
In Chapter 4 we pursue one of the consequences of the hypothesis that the meaning
of the sentence is worked out on the basis of its component parts and their
structural relations. We will discover that for this hypothesis to be maintained, the
sentences must have more than one subject position. We introduce the hypothesis
that the subject is ﬁrst inserted inside the VP and is then moved to the subject
position outside the VP.
The ﬁnal chapter of the book returns to question formation and we show how the
system elaborated in the ﬁrst four chapters of the book can be implemented to
derive the word order in English questions. This chapter focuses on the importance
of the movement operation for the formation of sentences.
Preface ix
TSA01 07/18/2005, 03:05PM9
A Note to the Teacher
This book targets introductory syntax classes. It could be the ﬁrst step in a syntax
program that will lead onto more theoretical work or it could be the starting point
of a more empirically oriented approach with a generative basis. The exercises try
to illustrate these two directions.
Though there are many exercises in the book, I hope that the exercises will also
provide inspiration for additional exercises along the format of those in the book.
This may be particularly relevant for teachers whose students are native speakers of
languages other than English. Exercises in the students’ own language can be provided
modeled on those in the book. One type of exercise which is not provided in
the exercise sections but is a natural spin-off from the way the book is written is to
ask students to look for particular patterns in their own reading. From my own
experience, though, I have found that it is important to deﬁne such research tasks
rather narrowly, so that they can be tied to the teaching. The attested data in the
exercises in this book can be taken as a guideline for the students’ own search. Such
research exercises can be devised both for English and for other languages.
References in footnotes of the text signal the relevant literature and they are
intended to make up for the inevitable shortcuts that have to be part and parcel of
a fairly basic introduction. Both older “classic” texts in the generative literature and
more recent minimalist texts have been included.
The textbook should cover an introductory semester-long course in syntax. The
chapters can also be the basis for self-study. The text can be complemented with
additional readings, and suitable supplementary reading can be of various types. By
way of illustration, I offer some suggestions here, but the choice will depend very
much on the overall orientation of the linguistics program into which this book is
being integrated. For instance, since a lot of the discussion hinges around functional
structure and the subject, the course could lead up to a study of some of the recent
discussions of the position of subjects or of verbs. Accessible overview papers
on this area can be found in many of the syntax handbooks that have been published
recently. McCloskey (1997), for instance, would be a very good follow-up to
Chapter 4. Another possible extension would be to take the students beyond the
proposals in the book and to explore the concept of “Predicate Phrase” (Bowers
2001). Yet another possibility would be to extend the discussion to the structure of
the nominal projection, an issue which is not touched upon very much here. Bernstein
(2001) could be the basis for such an extension. Some more advanced theoretical
papers written against a Minimalist background might also be used, though these
will probably require more input from the teacher.
The book might be suitably complemented with papers in neighboring areas of
interest. For instance, the discussion of functional categories might be linked to
papers on the question of language acquisition and on the question of how much of
such structure is present in the early grammar. To mention but two examples, one
might choose some of the papers in Clahsen (1996) or in Friedemann and Rizzi
x Preface
TSA01 07/18/2005, 03:05PM10
(1999). The text could also be complemented with material on language variation
or on creolization (cf. DeGraff (1997), and the papers in DeGraff (1999)). Alternatively,
the course could be accompanied by papers on processing such as Frazier and
Clifton (1989), or Gibson and Warren (2004) to mention one recent example.
The textbook should also enable the student to move easily on to introductory
textbooks such as my own Introduction to Government and Binding Theory (1994)
or Haegeman and Guéron’s (1999) English Grammar: A Generative Perspective.
The book could also lead onto any of the recent introductions to Minimalist syntax
such as Andrew Carnie’s (2002) Syntax: A Generative Introduction, David Adger’s
(2003) Core Syntax, Andrew Radford’s (2004) Syntactic Theory and English Syntax,
or Norbert Hornstein, Jairo Nunes, and Kleanthes Grohmann’s (forthcoming)
Understanding Minimalism: An Introduction to Minimalist Syntax.
Acknowledgments
This book owes a lot to my students and my colleagues. I mention in particular
the licence, maîtrise, and DEA students at the university of Lille, who were a skeptical
and challenging audience at ﬁrst. They were the ideal audience to try out this
material. Through them, I discovered the shortcomings of many textbooks, especially
my own earlier ones. My students were the good-humoured, willing, and challenging
testing ground for various parts of this book. Among my students, I mention in
particular Elodie Gauchet and Virgine Marant, who pointed out some of the typos
in Chapters 4 and 5.
Among the colleagues who encouraged me to write this book I mention David
Adger, Siobhan Cottell, Ilse Depraetere, Eric Haeberli, Chad Langford, Philip Miller,
Kathleen O’Connor, and Paul Rowlett. Thanks to David and Paul for comments on
some sections. I thank Ruth Huart for providing the critical outside view that made
me rethink textbook writing. Warm thanks also go to three anonymous reviewers
for Blackwell: their encouraging remarks and suggestions helped reshape this book.
I also thank Jacqueline Guéron for very useful comments on an earlier version and
Artemis Alexiadou, Frank Nuyts, Eric Haeberli, Suzan Pintzuk, and Ur Shlonsky
for help with the examples. Special thanks go to Katya Paykin-Arroues, who proofread
parts of the ﬁnal version, and to John Wakeﬁeld whose careful editing also
improved the text a lot. I also thank Siobhan Cottell for being a patient friend who
was there to listen when I needed a sympathetic ear, who encouraged me to write
this book and who painstakingly provided detailed comments on a preﬁnal version.
Siobhan helped me to devise ways of making this book more student-friendly.
I hope she will not be too disappointed by the ﬁnal version. Thanks to three people
at Blackwell who made writing this book so much fun: Philip Carpenter, who certainly
knows how to tempt his authors into writing books, and Ada Burnstein and
Sarah Coleman who acted like my guardian angels. I also owe thanks to Etienne
Vermeersch, who, about 30 years ago now, was the ﬁrst to make me realize the
Preface xi
TSA01 07/18/2005, 03:05PM11
importance of the difference between inductive and deductive thinking, and to
David Lightfoot, Henk van Riemsdijk, and Neil Smith, who, also some time ago
now, were the ﬁrst to teach me how to think syntactically.
Finally, I thank Hedwig and Nelson for being there, and for preventing me from
thinking syntactically all the time.
Liliane Haegeman
Gentbrugge
xii Preface
TSA01 07/18/2005, 03:05PM12
Introduction: The Scientiﬁc Study of Language 1
1 Introduction: The Scientiﬁc Study
of Language
Discussion
Contents
0 Introduction: Scope of the Chapter 3
1 Linguistics as the Science of Language 3
1.1 Linguistics as a science 3
1.1.1 Some deﬁnitions 3
1.1.2 Explanation: An example 6
1.1.3 Language phenomena: An example 7
1.1.3.1 Ambiguity 7
1.1.3.2 The data 11
1.1.3.3 Predictions 12
1.2 How to go about it 13
1.2.1 Induction and deduction 13
1.2.2 Explicitness, systematicity 15
1.2.3 Elegance, parsimony, economy 16
1.2.4 Doubt 18
1.2.5 Summary 18
2 From Raw Linguistic Data to Generalizations: Word Order in English
Questions 19
2.1 Introduction: Sentence meaning and word meaning 19
2.2 Question formation 21
2.2.1 Subject-auxiliary inversion 21
2.2.2 When there is no auxiliary 24
2.2.3 Language and economy 27
2.3 From form to meaning: Subject-auxiliary inversion and question
formation 28
2.3.1 Introduction 28
2.3.2 SAI can give rise to interrogative interpretation/SAI
always gives rise to interrogative interpretation 29
TSC01 07/18/2005, 03:05PM1
2 Chapter 1
2.3.3 Interrogative sentences can be formed by means of SAI/
Interrogative sentences are always formed by means
of SAI 32
2.3.4 Verbs and inversion 35
2.4 A brief discussion of deﬁnitions 36
3 Language and Languages 41
3.1 Going further aﬁeld: Comparative syntax 41
3.1.1 French 42
3.1.2 German and Dutch 43
3.2 Going back in time: Diachronic variation 44
3.3 Comparative data: Conclusion 45
4 Summary 46
TSC01 07/18/2005, 03:05PM2
Introduction: The Scientiﬁc Study of Language 3
1
In this book footnotes will be used for the following purposes:
• to add various comments to the text – notes 2 and 6 of this chapter are examples;
• to refer to earlier or later sections in the book in which the issue under consideration or a
related issue is discussed – notes 8 and 9 of this chapter are examples;
• to point the reader to relevant exercises – note 4 of this chapter is an example;
• to refer to the literature for more extensive discussion of issues dealt with in the text – notes
3 and 5 of this chapter are examples. In general the references will offer a more complete
survey of the data and/or a more sophisticated theoretical analysis. The texts referred to will
usually be more advanced and will probably not be accessible to the student-reader, at least
not at the early stages of the book. When a text is itself introductory (and hence accessible)
this will be signaled in the note.
0 Introduction: Scope of the Chapter
This chapter is an introduction: it sets the scene for the remainder of the book. The
focus of our enquiry in this book is language and in particular we will be interested
in the way that words are put together to form sentences. The study of sentence
formation is usually referred to as syntax.
Syntax is a branch of linguistics. In this chapter we discuss the main properties of
the methodology used in linguistics. We set the scene for the later chapters in that
we will determine how we ought to go about it when studying syntax. The chapter
is divided into three sections. In section 1 we discuss the methodological implications
of the idea that linguistics is a scientiﬁc discipline. We will try to determine what the
deﬁning properties of scientiﬁc work are and to formulate some guidelines for our
own work. Using the example of question formation in English, section 2 offers an
illustration of the scientiﬁc methodology used in linguistics. Section 3 shows why,
even when concentrating on the formation of English sentences, it is important to
extend the data we examine beyond Modern English. Section 4 is a summary.1
1 Linguistics as the Science of Language
1.1 Linguistics as a science
1.1.1 SOME DEFINITIONS
Syntax, the area of study we are concerned with in this book, is a domain of linguistics.
When we look up the word linguistics in a dictionary we ﬁnd deﬁnitions
such as the following:
TSC01 07/18/2005, 03:05PM3
4 Chapter 1
Linguistics (i) The science of language(s), esp. as regards nature and structure.
(Concise Oxford Dictionary (COD) 1976: 632)
(ii) The study of human speech in its various aspects (as the units,
nature, structure, and modiﬁcation of language, languages, or a
language including esp. such factors as phonetics, phonology, morphology,
accent syntax, semantics, general or philosophical grammar,
and the relation between writing and speech) – called also
linguistic science, science of language. (Webster’s Third New International
Dictionary of the English Language 1981: vol. II, 1317)
(iii) The study of language in general and of particular languages,
their structures, grammar etc. (Longman Dictionary of English
Language and Culture (LDOCE) 1998: 767)
The three deﬁnitions are similar, but careful readers may have observed that
deﬁnitions (i) and (ii) contain the word science, and that the word is absent from
deﬁnition (iii). Before we conclude that this means that the compilers of the COD
and those of Webster’s dictionary used to think that linguistics was a scientiﬁc
enterprise but that those compiling the Longman dictionary no longer do, consider
that in English other scientiﬁc ﬁelds of study are also referred to by words ending
in -ics: physics and mathematics, for instance. The gloss for the ending -ics in the
Longman dictionary is as follows:
-ics 1. The scientiﬁc study or use of ___: linguistics (the study of language), electronics
(the study or making of apparatus that uses chips, transistors etc.),
acoustics . . . (LDOCE: 1566)
In other words, combining Longman’s deﬁnition (iii) of linguistics with its gloss for
the ending -ics, we can conclude that the Longman dictionary makers also consider
linguistics to be the scientiﬁc study of language.
Since dictionary makers try to reﬂect actual usage of language, linguistics can
plausibly be deﬁned as the science of language or the scientiﬁc study of language.
However, while it is easy to provide such a deﬁnition of the discipline, it is much
harder to go beyond that and to explain what it is that linguists do and in what way
their work is supposed to be “scientiﬁc.” Commenting on this point the English
linguist David Crystal says:
Linguistics, indeed, usually deﬁnes itself with reference to this criterion [being scientiﬁc]:
it is the scientiﬁc study of language. But this is a deceptively simple statement; and
understanding exactly what anyone is committed to once he decides to do linguistics is
an important step, an essential preliminary to any insight into the essence of the subject.
What are the scientiﬁc characteristics that make the modern approach to language
study what it is? (Crystal 1971: 77)
Before embarking on the study of syntax, which is the branch of linguistics that
concentrates on the formation of sentences, we should try to clarify what makes a
TSC01 07/18/2005, 03:05PM4
Introduction: The Scientiﬁc Study of Language 5
branch of study scientiﬁc. Once we have done that, it will be easier to understand why
linguists in general, and syntacticians in particular, go about their work the way they
do. Note that the brief presentation of our interpretation of the concept “science” is
not at all an attempt to offer an introduction into the philosophy of science. Rather,
by stepping back and reﬂecting for a moment on what we normally see as the
deﬁning properties of science, we can try to isolate the main features of the scientiﬁc
method and then try to implement these same features when studying syntax.
Below are some deﬁnitions of the notion “science,” taken from various written
sources. Read them carefully and identify what you think the key concepts in these
deﬁnitions are. Pay particular attention to concepts that occur more than once.
(1) Systematic and formulated knowledge, pursuit of this or principles regulating
such pursuit. Branch of knowledge (esp. one that can be conducted on scientiﬁc
principles), or organised body of knowledge that has been accumulated on a
subject. (COD: 1066)
(2) Accumulated and accepted knowledge that has been systematized and formulated
with reference to the discovery of general truths or the operation of natural
laws; knowledge classiﬁed and made available in work, life, or the search
for truth; . . . knowledge obtained and tested through the scientiﬁc method.
(Webster’s Third New International Dictionary of the English Language 1981:
vol. II, 2032)
(3) Science is a hunt for order, explanation and regularity. It explains the anomalous
by reference to the law it seeks to establish. (Hywel Williams, Guardian,
7.8.2002, p. 8, col. 7)
(4) Science, by deﬁnition, is the search for order in nature. (Newmeyer 1983: 41)
The concepts that occur frequently in the deﬁnitions above have been isolated and
grouped:
knowledge (1), (2);
pursuit (1), hunt (3), search (2), (4), seek (3);
explanation (3);
laws of nature, natural laws (2), general truths (2), law (3);
order (3), (4), regularity (3), systematic (1), (2);
formulate/formulation (1), (2).
Not surprisingly, these extracts converge on the key concepts associated with science.
They all agree that science aims at achieving knowledge and that science is an activity.
Science is not an inert state of knowledge; science means doing something, engaging
in some activity. Scientiﬁc activity is deﬁned as a “search,” a “hunt,” a “pursuit”;
in other words science is the active pursuit of a goal. Combining these two concepts
TSC01 07/18/2005, 03:05PM5
6 Chapter 1
we can say that the search undertaken by the scientist has as its goal “knowledge,”
but the kind of knowledge that is achieved is in itself dynamic. The goal of scientists
is not merely taking note of and recording certain phenomena and thus “knowing”
about them: scientists want to explain the phenomena they have observed. Explanation
leads to understanding: scientists want to understand why the phenomena
observed are the way they are.
1.1.2 EXPLANATION: AN EXAMPLE
To clarify the notion “explanation” let us look at an example. We start from the
following very simple observation. Snow that has fallen overnight often turns into
water during the day. We refer to this natural phenomenon as “melting”: a solid
matter gradually turns into a liquid. When dealing with such a natural phenomenon,
scientists will not be satisﬁed with mere observation. They will want to understand
it. They will want to explain why the snow has melted and why other solid matters,
say, a glass or a plastic cup or the mud in the garden or the sand on the beach or the
tarmac on the roads, have not melted at the same time and/or in the same manner.
Scientists will also want to understand why snow melts on certain days, but does
not melt on other days. In order to explain the phenomenon observed scientists will
try to relate it to other phenomena. So the goal of scientists will be to ﬁnd the cause
of the phenomenon observed. For our example, a fairly plausible hypothesis could
be that snow melts on a certain day because during the day the temperature has
risen, and as a result the snow reaches the critical temperature at which it turns into
water, its melting point. If that particular temperature is not attained, snow will not
melt. Scientists might formulate the hypothesis that there is a causal link between
temperature and the solid/liquid states observed.
Scientists will not stop at snow turning into water. They will view the melting of
snow in more general terms; they will look at other solids and examine whether
these also change into liquids when heated. Metals, for instance, such as iron or
steel or copper, also melt, but they require a much higher temperature than snow.
In order to ﬁnd out whether particular metals melt or not, scientists cannot just
patiently wait and hope to come across them melting. For instance, if the melting
point of a particular solid matter is 100 degrees centigrade, this temperature cannot
be met with in everyday circumstances, even on a hot day. To go beyond the mere
observation of phenomena in the natural environment and to ﬁnd out more
about melting temperatures, scientists can resort to experiments: they heat solids
to a certain temperature and observe and record what happens. While doing so,
scientists rely on the generalized hypothesis that all solids will melt under certain
well-deﬁned conditions, namely when they reach a critical temperature, their
melting point.
As mentioned, when trying to assess the melting points of individual matters,
scientists do not just wait for things to happen. Rather, what they do is create the
relevant circumstances that can trigger the process under examination, in other
words they will run an experiment. But note that before doing the experiment,
TSC01 07/18/2005, 03:05PM6
Introduction: The Scientiﬁc Study of Language 7
scientists must already have some idea what the relevant factors will be. For instance,
if scientists think that heat is responsible for the melting process, they will apply heat
to the material and they will keep all other elements constant. The experiment is
guided by a hypothesis, namely that solids melt when heated to a critical point. The
goal of the experiment is (i) to test the general hypothesis that all solids melt when
they are at some particular temperature, and (ii) to identify the relevant critical
temperature.
What scientists are doing is looking for regularities (here that all solids liquefy at
a certain point), for systematic patterns. Scientists try to formulate general laws to
cover the facts they observe. They are looking for order. In our example, these laws
establish relations between temperature–matter–melting. We provide an explanation
if we can account for the phenomena, if we can say that snow melts because the
temperature rises above 0°C and that 0°C is the melting point of snow. On the
other hand, a silver bracelet will not melt in the same circumstances because its
melting point is much higher.
Scientists will not stop at the inventory of melting points. Having conﬁrmed that
a series of solids melt when heated to certain temperatures, they will then want to
explain why different materials have different melting points. Again they will try to
answer this question by observation, experimentation, and by forming hypotheses
which they put to the test.
As a further step scientists will try to explain the difference in the melting points
by looking more closely at the nature of the different materials under examination.
Ultimately, they will devise an account which not only explains why the matters
that have been observed melt at a particular temperature but they will also try to
predict melting points for matters that they may come across in future. For instance,
they will predict the melting point of a metal that consists of two parts zinc and one
part copper. Note that this means in fact that by identifying a melting point for a
solid matter scientists predict when the solid matter will melt and they also predict
when it will not melt, i.e. when it remains solid. Once again, the prediction will be
tested by experimentation.
1.1.3 LANGUAGE PHENOMENA: AN EXAMPLE
1.1.3.1 Ambiguity
The object matter that is studied in linguistics is language. If linguistics is a science,
then we should not simply make an inventory of linguistic phenomena (i.e. language
facts) and describe them but we also want to explain them. Let us just look at a
simple point here to illustrate the nature of the task that awaits the linguist. Consider
example (5a), taken from a British newspaper. How does this extract refer to
the protesters? What kind of individuals would qualify as the relevant protesters?
(5) a Manchester’s morning rush-hour trafﬁc was brought to a near standstill
yesterday as 150 black cab drivers staged a go-slow protest calculated to cause
maximum disruption to commuters. (Guardian, 14.9.2000, p. 4, cols 2–3)
TSC01 07/18/2005, 03:05PM7
8 Chapter 1
In the extract, the protesters are described by means of the string of words 150
black cab drivers. What kind of individuals does this string pick out? The string of
words 150 black cab drivers has two interpretations or two readings: in one reading
we are referring to ‘those who drive cabs and are of a speciﬁc ethnic origin’, and in
the other we are referring to ‘those who drive cabs which are of a particular color’.
In both readings, the adjective black distinguishes the drivers in question from
others: in the ﬁrst reading the distinctive feature is the color of the driver’s skin, and
in the second it is the color of his cab. In example (5a) both readings are available.
Observe that the extract above is taken from a British newspaper. In Britain,
taxis are indeed often black. But even in a context in which taxis tend to be a
different color, say yellow, the string 150 black cab drivers still potentially has the
two interpretations described above. Linguistically speaking, the string is ambiguous
regardless of which color taxis actually are.
The question arises why the string 150 black cab drivers has these two interpretations.
Are all strings of words necessarily ambiguous in this way? If not, what is the
cause of the ambiguity of this example? Could it be the word cab, another word for
taxi, that causes the ambiguity? To ﬁnd out if the use of the word cab is at the basis
of the ambiguity, we can experiment with the sentence and replace the word cab
with the word taxi. Consider (5b): is this sentence ambiguous?
(5) b Manchester’s morning rush-hour trafﬁc was brought to a near standstill
yesterday as 150 black taxi drivers staged a go-slow protest calculated to
cause maximum disruption to commuters.
(5b) remains ambiguous. The presence of the word cab in (5a) as such is not the
cause of the ambiguity. Does the presence of the numeral 150 have anything to do
with the ambiguity? Or could the ambiguity be due to the fact that the noun driver
is in the plural? Neither of these is probably at the basis of the ambiguity; to
conﬁrm this intuition let us again experiment with the sentences above. It is clear
that both (5c), without the numeral 150, and (5d), with a singular noun driver,
remain ambiguous.
(5) c Manchester’s morning rush-hour trafﬁc was brought to a near standstill
yesterday as black cab drivers staged a go-slow protest calculated to cause
maximum disruption to commuters.
d Manchester’s morning rush-hour trafﬁc was brought to a near standstill
yesterday as a black cab driver staged a go-slow protest calculated to cause
maximum disruption to commuters.
Can we reword the string 150 black cab drivers and make it unambiguous? One
option is shown in (5e):
(5) e Manchester’s morning rush-hour trafﬁc was brought to a near standstill
yesterday as 150 drivers of black cabs staged a go-slow protest calculated
to cause maximum disruption to commuters.
TSC01 07/18/2005, 03:05PM8
Introduction: The Scientiﬁc Study of Language 9
Table 1 Classiﬁcation of examples
Number Example Ambiguous?
(5a) 150 black cab drivers +
(5b) 150 black taxi drivers +
(5c) black cab drivers +
(5d) a black cab driver +
(5e) 150 drivers of black cabs −
At this point, we could inventorize our observations and come up with the classiﬁcation
in Table 1. Why is (5e) no longer ambiguous? And why are the other examples
ambiguous? The ambiguity relates to the position of the adjective black in relation
to the other words of the segment. In the ambiguous cases black precedes cab driver(s)
and it may either be taken to modify a string cab driver(s), in which case black
refers to the ethnic origin of the driver(s), or it may be taken to modify the noun
cab, in which case it refers to the color of the cab. We can show these relations by
using square brackets as in (6).
(6) a 150 [[black cab] drivers]
b 150 [black [cab drivers]]
Square brackets show the grouping of words into larger units: in (6a) black is
combined with cab, giving the unit [black cab]. The meaning of the unit [black cab]
is calculated on the basis of the combination of the meanings of its component
parts, the words black and cab. The meaning of black combines with the meaning
of cab: in this grouping black refers to the color of the cab. The unit [black cab] is
then grouped with drivers to form a more comprehensive unit [[black cab] drivers].
The meaning of the resulting unit is again based on that of its component parts:
(i) black cab, and (ii) drivers. With the grouping in (6a), black cab drivers denotes
a driver of black cabs.
In (6b) on the other hand, cab is ﬁrst combined with drivers to form [cab drivers].
The meaning of this unit is calculated on the basis of the meaning of its two
component parts cab and drivers: here cab drivers denotes people who drive cabs.
Then we combine the unit [cab drivers] with the adjective black to form [black
[cab drivers]]. Again the meaning of black cab drivers is based on that of its component
parts, (i) black and (ii) cab drivers. In the grouping in (6b), the adjective
black modiﬁes the unit cab drivers; black cab drivers now denotes cab drivers who
are black.
The fact that two groupings of words are available for one string of words is
the cause of the ambiguity of the string. So we explain the observed ambiguity by
relating it to a particular cause: the internal organization or structure of the string.
TSC01 07/18/2005, 03:05PM9
10 Chapter 1
The ambiguity in the relevant examples is said to be structural. This means that we
must assume that the relations between words have an impact on their interpretation:
the string 150 + black + cab + drivers has two meanings because the words in the
string can be combined with each other in two different ways.
To remove the ambiguity we can combine the words black, cab, drivers differently,
as shown in (5e). In (5e) the adjective black precedes cabs and it does not
precede the noun drivers. In this example the adjective black is related uniquely to
cab, and only one reading is available, the reading corresponding to that of (6a).
(6) c 150 drivers of [black cabs]
The following extract conﬁrms the potential for ambiguity of the example in (5a):
(7) a A few years ago a newspaper article about the dangers of women riding alone
in cabs brought a long and furious tirade from a reader incensed by the way
the drivers had been racially described. In fact the article had been using the
phrase “black cab drivers” to differentiate those working in hackney cabs
from mini-cab drivers. (Independent, 13.10.2000, Review, p. 5, col. 2)
Example (5a) actually appeared in the context (7b). In that context, a reader confronted
with the ambiguous sentence (5a) would immediately have been able to
select the appropriate grouping of the words with the associated reading: in (7b)
reference is made to “black cabs,” making black distinctive as a color of cabs.
(7) b More than 70 black cabs travelled under police escort from Manchester
airport to the city, driving four abreast and slowing early morning trafﬁc to
a 10mph crawl. (Guardian, 14.9.2000, p. 4, cols 2–3)
Examine the caption in (8a) which was used to characterize a person on TV: in
what way is it ambiguous? What could be the cause of the ambiguity?
(8) a a tall rose grower (BBC 1 television, 31.7.2002 (News, South))
The person we are talking about, the “referent” of the string of words in (8a), could
be either a person of any height who grows tall roses (8b), or a tall person who
grows roses of any height (8c).2
We can again relate the ambiguity of (8a) to the
structure of the sequence of words: that is, to the different ways the words tall, rose,
and grower can be combined. In (8b) and (8c) square brackets again represent the
two structures. In (8b) we ﬁrst combine tall with rose, giving the unit tall rose. In
this unit the adjective tall modiﬁes rose: it denotes the size of the rose. This unit is in
2
In the particular BBC broadcast the ﬁrst reading was intended: the speciality of the particular
gardener was growing tall roses.
TSC01 07/18/2005, 03:05PM10
Introduction: The Scientiﬁc Study of Language 11
turn combined with grower. The person denoted by this string of words grows tall
roses. According to the grouping in (8c), rose ﬁrst combines with grower, giving the
unit rose grower. This unit denotes a person who grows roses. The adjective tall
then combines with rose grower. In the second combination, tall modiﬁes the unit
rose grower, the adjective indicates the size of the rose grower.
(8) b a [[tall rose] grower]
c a [tall [rose grower]]
Strings of words are sometimes ambiguous, and the ambiguity of the particular
examples examined above was due to the organization of these words into larger
units, their structure. In both the examples, black cab drivers and tall rose grower,
the ambiguity is related to the sequencing of the combination of the elements. This
means that the interpretation of a string of words is not merely the left-to-right sum
of the interpretations of the individual words. It also depends on how the words
are put together. We could think of a mathematical analogy here. The formula
(A − B) − C is not identical to the formula A − (B − C). When A = 6, B = 3, and
C = 2, for instance, the ﬁrst equation equals 1, and the second equals 5. We can
make this observation into a more general hypothesis and propose that in language,
interpretation depends on the way the strings of words are composed, namely their
structure:
(9) Compositionality
The meaning of a string of words is determined compositionally; i.e. it is
determined by its component parts and by their relations.
1.1.3.2 The data
Out of context, the string 150 black cab drivers (5a) has two interpretations; the
string 150 drivers of black cabs (5e) does not. This is a fact of language. We offered
a ﬁrst explanation in terms of the grouping of the words contained in the string.
Before we continue the discussion, it is useful to think again about the kind of
language material we have been using. Did we restrict ourselves to observing the
language material available? Or did we also use experimental facts?
Sentence (5a) is an attested example, it was found in a newspaper. As speakers of
English we are able to interpret it and we can assign two interpretations to it. In
other words, we use our intuitions about the interpretation of the string. The dual
reading of the example is due to the fact that there is an ambiguous string in the
sentence, 150 black cab drivers. We have relied on material found, an attested
sentence, but not only that: we also rely on our linguistic competence. As speakers
of English, we can work with the observed material: we assign an interpretation
to the strings of words, and, using our knowledge of the language, we are able
to reformulate these strings and compare the interpretations of various strings.
Sentence (5a) does not come with a warning that it is ambiguous. We rely on our
intuitions about the language to decide on its interpretation.
TSC01 07/18/2005, 03:05PM11
12 Chapter 1
Sentences (5b–e) are not attested examples. Relying on our competence as
speakers of English, we have constructed these sentences ourselves, using (5a) as
our inspiration. If we want to understand why a sentence is ambiguous, we will not
just examine it as it is. We will play around with the example, to see whether we
can construct similarly ambiguous sentences, or ones that are unambiguous. We
experiment with the data, relying on our competence of the language.3
As linguists we will, among other things, want to look at data such as those in
(5) and try to explain why examples (5a–d) are ambiguous and why example (5e)
is not. We rely on our own intuitions concerning attested data, and also on experimental
data (sentences which we construct ourselves).
Though attested data may be useful, we deﬁnitely cannot conﬁne or research to
them. In addition to playing around with attested examples, as we have done above,
we can also just construct examples “out of the blue” and experiment with them.
For example, the string in (10a), which is again ambiguous, is not an attested
example. It is a constructed example which serves to illustrate once again how the
different groupings of words lead to ambiguity:
(10) a a Flemish language teacher
b a [Flemish language] teacher
c a Flemish [language teacher]
1.1.3.3 Predictions
Recall that one of the goals of a scientiﬁc approach is also to predict what is
possible and what is not possible. For instance, the melting point of a metal predicts
both at which temperature the metal will melt and when it will not melt. Similarly,
when dealing with language data we want to elaborate predictions. For instance,
taking our example above, we don’t only want to account for the ambiguity of a
particular example, but we also want to predict when strings of words will be
ambiguous. Based on the attested examples in (5a–d) and in (8a) and on the constructed
example in (10a) we could formulate a ﬁrst hypothesis that a string of
words composed of the sequence adjective – noun – noun may lead to ambiguity.
The ambiguity of such sequences is due to the fact that the adjective either bears on
the noun that it immediately precedes or it bears on the combination of the two
nouns that it precedes:
(11) a [[adjective noun] noun]
b [adjective [noun noun]
Thus we generalize our ﬁndings and go beyond the description of some individual
examples (attested or constructed) to formulate general principles. (11c–g) contains
3
On the use of intuitions and attested data see also the recent (and fairly accessible) discussions
in Borsley and Ingham (2002, 2003), Stubbs (2002), Lehmann (2004), and the papers in Penke
and Rosenbach (2004).
TSC01 07/18/2005, 03:05PM12
Introduction: The Scientiﬁc Study of Language 13
some additional constructed examples of the same sequence adjective – noun –
noun and indeed these examples are also ambiguous.4
(11) c a French art student
d an American literature teacher
e an Italian restaurant owner
f a Dutch bicycle maker
g a trendy furniture designer
The example in (5e), which was not ambiguous and which we repeat here in (11h),
does not display the relevant sequence. Here the adjective black preceded just the
one noun cabs, which it modiﬁes.
(11) h 150 drivers of black cabs
1.2 How to go about it
1.2.1 INDUCTION AND DEDUCTION
From the descriptions above we can also infer how not to proceed in scientiﬁc
work. To reach the goal of explaining the data that we observe we cannot simply
draw up a list of interesting observations. A mere list of phenomena does not lead
to any understanding. When discussing an example such as (5a), for instance, we
cannot satisfy ourselves with a mere anecdotal description of the example and how
it may give rise to ambiguity and to misunderstanding (cf. (7a)). We should try to
relate the observed language fact, the ambiguity of the example, to other language
facts and to elaborate an explanation that goes beyond example (5a).
A starting point is identiﬁcation and classiﬁcation of the data, the material we
wish to examine. We may, for instance, identify a set of ambiguous examples and
oppose them to a set of non-ambiguous examples. Classiﬁcation is followed by an
attempt at explanation.
4
Exercises 1, 2, and 3. Among other things, footnotes will be used to refer to the exercises. When
a footnote reads “Exercise 1” this means that you can try Exercise 1 at that point in the
chapter. You are advised to tackle the exercises at two points in time. First you can do each
exercise at the point in the chapter when it is signaled by a note. The exercise will allow you
to apply what you have just learnt and will provide more illustrations of the concept being
discussed. You can also try to do the same exercise later on, when you have covered more
ground. Doing this will ensure that you still remember the notions which you have learnt
previously.
Sometimes a (partial) key will be provided in the exercises and additional discussion will be
added under the heading “Key and comments.” These supplementary discussions will alert you
to speciﬁc points that have not been tackled in the main body of the text. In particular, sometimes
such discussions will answer questions that you may have been wondering about.
TSC01 07/18/2005, 03:05PM13
14 Chapter 1
When research starts from observation of empirical data, the procedure we adopt
is referred to as induction. This type of approach is captured by Webster’s dictionary
in the following way:
natural science
A branch of study that is concerned with observation and classiﬁcation of facts and
esp. with the establishment or strictly with the quantitative formulation of veriﬁable
general laws chieﬂy by induction and hypotheses. (Webster’s Third New International
Dictionary of the English Language 1981: vol. III, 2032)
By means of induction we attempt to uncover general principles (or “laws”) that
underlie the observed phenomena. We formulate hypotheses whose ﬁrst goal is to
account for the observed phenomena. Ideally, however, the hypotheses must always
go beyond providing an account for what is observed. We also want to understand
why we have observed just those phenomena and not others. We want to be able
to predict which alternative phenomena could have been observed and which
ones would never arise. Put differently, we set out to deﬁne the bounds of what
is possible.
In our melting point example discussed in section 1.1.2, scientists ﬁrst observe
and classify data in relation to the natural phenomenon of melting. At some point
they will have established an inventory of melting points: for instance silver melts at
961°C, while gold only melts at 1063°C and platinum melts at 1769°C. As a second
step, an attempt is made to provide an explanation for why silver melts at a lower
temperature than gold. This difference will be related to the internal composition of
the solid materials studied. A successful analysis should be able to account for the
melting temperatures observed and it should also predict when solid matters will
melt and when they will not melt. Similarly, when dealing with the ambiguity of
(5a) we ﬁrst classify a sample of language data with respect to their potential for
ambiguity. The ambiguity is related to the internal composition of the data analyzed,
in particular the ambiguous strings allow for two possible groupings of the sequence
adjective – noun – noun. In so doing, we deﬁne the bounds of what is possible. We
predict that 150 black cab drivers is ambiguous, because it has the relevant structural
property, and that 150 drivers of black cabs is not ambiguous, because it lacks
those properties. Thus, we go beyond the data observed and formulate predictions
about what can arise and what will not arise.
When working on the linguistic examples we appealed to some hypotheses about
language. For instance, we proposed that words are grouped, that language is structured.
We needed these concepts to be able to isolate a string of words 150 black
cab drivers, from a sentence. We appealed to a general concept “structure” to refer
to groupings of words in the string. But if we appeal to the concept structure, then
we need to clarify at least two points. (i) We have to deﬁne the nature of linguistic
structures, and (ii) we have to be able to make precise how “structure” is mapped
into meaning or interpretation. In other words we have to elaborate a theory of
language; we need a theory about how linguistic forms are structured and how
TSC01 07/18/2005, 03:05PM14
Introduction: The Scientiﬁc Study of Language 15
these structures relate to interpretation. Such a theory will provide the framework
for the discussion and explanation of the data examined.
Scientiﬁc work is guided both by empirical considerations (observation of data
and experimentation) and by theoretical concepts. With respect to linguistics, the
interplay between empirical data and theory is expressed very clearly in the following
extract by the Dutch linguist Simon Dik:
In linguistics, as in other sciences, there is an essential interaction between data
analysis and theory formation: an adequate analysis of the data of some particular
language is impossible without some general theoretical insight into the principles
underlying the structure and functioning of language in general; on the other hand, an
adequate development of general linguistic theory presupposes the meticulous analysis
of the facts of particular languages. (Dik 1989: 33)
In our example above, we proceeded from the observation of empirical data to the
formulation of a hypothesis which provides an explanation of these data. This way
of working is called induction. The combination of several hypotheses about a certain
domain of enquiry (here language) gives rise to a more comprehensive network of
hypotheses, a theory.
Having formulated a set of principles that are part of a theory, scientists (and hence
linguists) may also proceed to working “deductively.” That means that they examine
a particular component of their theory (that is the network of hypotheses). Their aim
will be to examine how the hypotheses that have been formulated interact with each
other. For instance, they may look for internal inconsistencies that arise when two
hypotheses lead to contradictory predictions. They may also examine whether there
is any overlap between the different components of the theory, when the same facts
are explained by two different hypotheses. This type of theoretical work may lead
to the reformulation of some components of the theory. Thus novel hypotheses may
emerge from theoretically oriented work and these new hypotheses will themselves
have to be tested on the basis of the empirical data. Once again the data examined
may consist of attested language material or of constructed language material.
In science, experimental, data-driven work and theoretical work continuously
interact. Hypotheses are formulated on the basis of the observed data and these
hypotheses are integrated into the theory. The theory itself is examined and streamlined;
theoreticians formulate predictions on the basis of the reformulated theories
and their predictions are tested by observations and experiments.5
1.2.2 EXPLICITNESS, SYSTEMATICITY
In our list of essential concepts in the deﬁnitions (1)–(4), we also signaled the terms
formulate and formulation ((1), (2)). Some extracts from the dictionary deﬁnitions
of the verb formulate that are relevant here are given in (12):
5
For a general discussion of the relative impact of induction and deduction in various present
day approaches to linguistics see also Stuurman (1989).
TSC01 07/18/2005, 03:05PM15
16 Chapter 1
(12) formulate a Longman to express in an exact way
b COD set forth systematically
c Webster put into a systematized statement or expression
The deﬁnitions of formulate refer to expressing something systematically and in
an exact way. Scientists have to formulate, i.e. to state, their basic assumptions,
their hypotheses, their procedures, and their results precisely and explicitly. This
will enable other scientists working in the same area to evaluate the work, to repeat
experiments on which the research is based, and either to accept and implement
(parts of) the ﬁndings contained in the work or to challenge them. To put it more
succinctly: scientiﬁc research is “capable of replication and subject to peer review.”6
In order to guarantee that their research can be replicated and reviewed by their
peers, linguists also have to formulate their ﬁndings as precisely and explicitly as possible.
Sometimes, mathematical types of formulae are used in linguistics. This is not
really a requirement of scientiﬁc methodology, but it is a natural by-product of the
wish to be as precise as possible. By using exact and generally unambiguous formulae,
scientists ensure that there is clarity as to the interpretation of their statements.
Note that the term formulate implies a pre-requirement that scientists be able to
deﬁne the terms they use. They must be able to describe their procedures, argumentation,
etc. They cannot satisfy themselves with a vague description of results without,
for instance, stating exactly how experiments were run and how the results were
obtained. For linguistics, the same requirements of explicitness apply. Simply saying
that (5a) has two interpretations and that this is due to the adjective black modifying
either the noun cab or the noun cab drivers is not going to be sufﬁcient. We must
express quite precisely how the relation of modiﬁcation is encoded in language. In
our representation of the structure, we have tried to represent this by the squared
bracketing convention, which is used to represent the grouping of words into units.7
Another point that comes up regularly in the deﬁnitions of the scientiﬁc enterprise
is the concept systematicity (see deﬁnitions (1) and (2) and also (12c)). The linguist
David Crystal (1971: 90) says: “The need to study phenomena using a procedure
which is as methodical and standardized as possible is . . . obvious enough.” He
goes on to underline the importance of an underlying descriptive framework that provides
the system in which the research is inscribed. Systematicity implies systemizing,
i.e., looking at things against the background of a system. “Systemizing is the drive
to analyse and explore a system, to extract underlying rules that govern the behaviour
of a system” (Simon Baron-Cohen, Guardian, G2, 17.4.2003, p. 12, col. 1).
1.2.3 ELEGANCE, PARSIMONY, ECONOMY
It often happens that a number of scientists (or linguists) are simultaneously trying
to account for a particular set of data and that each comes up with a different account
6
Citation due to Dr David Gosling, letter to the editor: Independent, 15.7.2004, p. 22, col. 2.
7
A very accessible preliminary discussion of requirements in scientiﬁc work is given in Crystal
(1971: 77–127).
TSC01 07/18/2005, 03:06PM16
Introduction: The Scientiﬁc Study of Language 17
for these data. Scientists will often be seen to elaborate competing accounts. The
question arises of how to choose between competing accounts. What would make
one explanation or one theory better than another? Newmeyer says:
Science, by deﬁnition, is the search for order in nature. Scientists take it for granted
that their goal is to formulate the most elegant (i.e. the most order reﬂecting) hypothesis
possible, consistent with the data, about the particular area under investigation.
(Newmeyer 1983: 41)
Let us go back to our example of the melting of snow. The initial observation was
that the snow that had fallen overnight may melt during the day. Scientists working
on this issue and who observe that snow melts when the sun comes up might have
proposed that the melting process is due to the length of exposure to sunlight. In
other words, they explain the melting by two factors: (i) sunlight, (ii) time. Even
though these scientists might also be able to account for the observed fact, snow
melting during the day, their account is not as highly valued as the one we elaborated
above because it invokes two factors, sunlight and time, rather than one,
temperature. If two accounts cover the same sets of facts, then an account relying
on one factor is better than an account that requires two. Ultimately, in fact, the
account which accounts for the melting of snow relying on sunlight and time can be
reduced to an account in terms of temperature, because sunlight will give rise to an
increase in the temperature. But we know that sunlight as such is not essential for
snow to melt: a sudden increase in the temperature overnight will also make snow
melt. Explanations and theories should use as few rules/principles as possible to
account for the data.
The idea that scientiﬁc explanation should be as simple as possible is not new,
it is sometimes referred to as “Ockham’s Razor,” due to the English theologian
and philosopher William of Ockham (c.1285–1349), who said that entia non sunt
multiplicanda praeter necessitatem – ‘entities are not to be multiplied unnecessarily.’
This means that, other things beings equal, the simpler of two explanations
is to be preferred. In the same vein, Newmeyer writes:
Certain points, I think, are uncontroversial. One is that, given two theories that cover
the same range of facts, the one in which the facts follow from a small number of
general principles is better than the one that embodies myriad disparate statements and
auxiliary hypotheses. Another is that it is methodologically correct to reduce redundancy
within a theory, to reduce the number of postulates while preserving the scope
of the predictions. (Newmeyer 1983: 41)
Einstein put it more succinctly:
The grand aim of all science is to cover the greatest possible number of experimental
facts by logical deduction from the smallest number of hypotheses or axioms. (Einstein
1954, cited in Abraham et al., 1996: 4)
TSC01 07/18/2005, 03:06PM17
18 Chapter 1
In linguistics too, we will value an account with a smaller number of rules more
than one which requires more rules to explain the same set of data.8
1.2.4 DOUBT
A ﬁnal essential ingredient of scientiﬁc work is doubt. This statement may come as
a surprise, since a search for knowledge and understanding would at ﬁrst sight seem
to aim at certainty rather than doubt. When we say that doubt is an important
component in scientiﬁc work this means that we should always remain aware that
our answers to problems and the knowledge we acquire are hypotheses. New insights
or new developments in research may well mean that we must go back on what we
think we know and revise earlier proposals. The journalist Tim Radford cites the
scientist Tom McLeish:
Doubt, expressed most potently 3,000 years ago in the biblical book of Job, is the
greatest scientiﬁc tool ever invented . . . To do good science you have to doubt everything,
including your ideas, your experiments, and your conclusions. (Guardian, G2,
4.9.2003, p. 12, col. 4)
1.2.5 SUMMARY
In this section we have looked at the idea that linguistics is a science and we have
gone over the main properties of the scientiﬁc method. Science is based on the
interaction of the observation of phenomena (“data”) and theory. The observation
of data may lead to theoretical proposals or hypotheses. This is described by the
term induction. Sets of hypotheses, or theories, may themselves also lead to new
hypotheses; this is referred to as deduction. We have also seen that scientiﬁc work is
systematic and explicit. It aims at providing simple explanations for complex data.
We have mentioned that one should not take for granted whatever results one has
arrived at and that any kind of research implies that the researcher is willing to
question and challenge the results of his or her own work.
In the remainder of this chapter we will illustrate the kind of phenomena, the data,
that are dealt with in syntax. We will look at a set of language phenomena and we
will try to describe the data and evaluate some explanations for them. In the later
chapters of this book, we will elaborate step by step analyses of speciﬁc problems,
focusing on the overall question of how a sentence is structured. We will try to show
in as precise a way as possible how a hypothesis can be developed and evaluated in
linguistics, and how the proposals elaborated will lead to the formulation of a more
comprehensive theory. We will repeatedly show that once we have developed a
certain hypothesis we need to examine its consequences and that we continuously
need to reconsider and revise the results of earlier work. The role of doubt in
scientiﬁc work will thus be made clear throughout the discussion. Note that though
8
For a concrete illustration of how the criterion of economy or simplicity can apply in syntactic
theory see Chapter 2, section 2.4.2.
TSC01 07/18/2005, 03:06PM18
Introduction: The Scientiﬁc Study of Language 19
we will end up providing some interesting insights into language, in this book we
are mainly interested in the process of the research, that is, how we have arrived at
the results.
2 From Raw Linguistic Data to Generalizations:
Word Order in English Questions
In this section we examine another concrete example of how we could go about
analyzing language in a scientiﬁc way. We choose what looks like a well-known and
very simple domain of enquiry, that of English question formation. The goal of the
section is to show how even apparently simple linguistic phenomena require the
greatest care when it comes to formulating hypotheses. It is important to bear in
mind that the goal of this section (and indeed of this book) is not to elaborate a fullﬂedged
and ﬁnished analysis of sentence formation, nor do we pretend to arrive
at a complete explanation, but rather we try to illustrate one way we can “think”
scientiﬁcally about language. What we will try to do is to “unpick” our thinking
about a phenomenon, to dissect the argumentation into smaller building blocks. We
will also show how we may compare various formulations.
In the discussion below it is important to actively try to do the thinking. When a
question is raised in the text, ﬁrst try to answer it before reading the account. It is
important, then, to be an active reader who does not simply follow the text but who
tries to carefully monitor each step of the discussion.
2.1 Introduction: Sentence meaning and word meaning
In the discussion above, we have introduced the idea that language somehow unites
“form” with “interpretation.” The “forms” of language are ultimately either sounds,
or symbols on paper, “letters.” Linguistic entities may be associated with interpretation.
We say “may” because sounds as such do not necessarily have meaning. For
instance, though the sound [a:] happens to correspond to a meaningful unit in many
variants of English (“are”), other sounds [b] or [p] do not. Words, on the other hand,
are meaningful units: dog, cat, nose, etc. are all words with an interpretation.9
Some
words may have the same meaning, for instance cab and taxi. Such words are said
to be synonyms. Some elements may correspond to more than one meaning. A
9
In Chapter 3, section 3 we will discuss how the kinds of meanings conveyed by words may
be made more precise. For instance, the verb examined in example (i) seems, at ﬁrst sight, to
contribute more to the message conveyed by the sentence than the auxiliary have.
(i) The students have examined the documents.
TSC01 07/18/2005, 03:06PM19
20 Chapter 1
classic example is the word bank, which may refer to a riverbank or to a ﬁnancial
institution.
Sentences consist of words, and the interpretation of a sentence is calculated
(or “computed”) on the basis of the combined meanings of the individual words.
The words contribute their own meaning to the sentence, and the combination of
these individual meanings provides us with the sentence meaning. Going back to
(5a), for instance, if you replace the word cab by its synonym taxi the meaning of
the sentence does not change. This is so because the contribution of cab to the
meaning of the sentence is the same as that of the word taxi. If you replace cab
by a word with a different meaning, say bus, then the meaning of the sentence
will change:
(5) f Manchester’s morning rush-hour trafﬁc was brought to a near standstill
yesterday as 150 black bus drivers staged a go-slow protest calculated to
cause maximum disruption to commuters.
Observe that (5f) remains ambiguous. In the sequence black bus drivers the adjective
black might again refer to the ethnic origin of bus drivers, or it might set off black
buses from other buses. The latter reading does presuppose that such a taxonomy
of buses makes sense; one can easily imagine a context in which black buses might
for instance be run by a low-budget company, or that they are speciﬁcally used for
long-distance travel, etc.
Sentence meaning derives from word meaning. However, we have seen that
the meaning of sentences is not simply attained by adding up the meanings of
the individual words. The meaning of a sentence is also determined by how the
sentence is assembled, how the words are put together. This was illustrated by
the discussion of (5a). In order to account for the ambiguity of this example, we
elaborated the hypothesis that words in a sentence are grouped; in other words,
they form units, which we indicated by means of square brackets: we repeat (6a–b)
here as (13a–b):
(13) a 150 [[black cab] drivers]
b 150 [black [cab drivers]]
(13a) serves to show that in the sequence adjective – noun – noun, the adjective
black is grouped with the noun cab: the string black cab is a unit inside the larger
unit black cab drivers. In this structuring of the words, black modiﬁes cab; black
refers to the color of the cab. In (13b) the noun cab is structured with the noun
drivers, and to this unit is added a speciﬁcation of color. Cab narrows down the
type of driver we are talking about. In (13b) black modiﬁes cab driver; black indicates
the ethnic origin of the drivers. So sentence meaning is based (i) on the meaning
of the individual words, and (ii) on the way these words have been assembled
into larger units. The technical term to refer to the way words are assembled
into sentences is syntax, which is based on Greek συν (“sun”) ‘together’ and τασσω
TSC01 07/18/2005, 03:06PM20
Introduction: The Scientiﬁc Study of Language 21
(“tassoo”) ‘put, arrange in a particular order’. In the next section we will explore
further the idea that syntax determines sentence meaning.
2.2 Question formation
2.2.1 SUBJECT-AUXILIARY INVERSION
Consider the underlined sections in the following short extract: What does the
punctuation mark “?” at the end signal? Suppose you replace the symbol “?” by the
full stop. How would you minimally have to change the sentences?
(14) She had meant to drive down to the quay and regain the yacht, but she
now had the immediate impression that something more was to happen
ﬁrst. “Which way are you going? Shall we walk a bit?” he began . . . (Edith
Wharton, The House of Mirth, 1998: 201)
We refer to the symbol “?” as a “question mark” because this symbol occurs at the
end of a sentence which is used to ask a question. Sentences ending in a question mark
convey that there is a certain amount of information which the speaker/writer doesn’t
have and he or she is trying to make the interlocutor supply that missing information.
In the extract (14) the speaker (“he”) asks two questions of his interlocutor
(“she”). Let us isolate the questions in (14) and look at their form more closely.
(15) a Shall we walk a bit?
b Which way are you going?
If we merely replace the question mark by a full stop and do nothing else, the
sentences do not really work any more. There is something wrong with them;
they are not really acceptable sentences of English. In (15c, d) below we use the
asterisk (*) to signal that we ﬁnd a sentence unacceptable. In fact, care must be
taken here. Of course, the sequences of words in (15c, d) are as such not unacceptable,
since they are perfectly natural questions (15a, b), but the problem with
(15c, d) is that by removing the question marks and replacing them with full stops,
we signal that the sentences should no longer be interpreted as questions. The
asterisks in (15c, d) mean that these sentences become unacceptable if not interpreted
as questions.
(15) c *Shall we walk a bit.
d *Which way are you going.
To repair the sentences in (15c, d) we could propose the rewordings in (15e, f). These
sentences can be used as assertions; they are not normally used as questions. The
speaker does not indicate that he or she expects a response from the interlocutor; he
TSC01 07/18/2005, 03:06PM21
22 Chapter 1
or she simply afﬁrms something. The sentences in (15a, b) are interrogative sentences,
those in (15e, f) are declarative.10
(15) e We shall walk a bit.
f You are going this way.
Compare the forms of (15c, d) and of (15e, f). We see that the sentences in (15e, f)
begin with the pronouns we and you; these pronouns function as the subjects of the
sentence. The subjects are followed by shall and are, elements referred to as auxiliaries.
A provisional (and very approximate) characterization of auxiliaries, to be reﬁned
in Chapter 3, is that they are elements that are typically followed by a verb: shall is
followed by the verb walk, are is followed by the verb going.11
The examples above
show that the relative positions of the subject and the auxiliary in a declarative sentence
are different from those in an interrogative sentence. Consider the extracts in
(16). Identify the questions. For each question locate the subject and the auxiliary.
(16) a Eventually the waitress came out of the kitchen with a tray the size of a
table-top . . . “Can I get you anything else?” she said. “No, this is just ﬁne,
thank you.” . . . “Would you like some ketchup?” “No, thank you.” (Bill
Bryson, The Lost Continent, 1990: 159)
b The people of Toronto are not wearing masks . . . Are we taking precautions
such as washing our hands? Of course. Are we stopping our lives
because of this? Certainly not. (Guardian, 26.4.2003, p. 11, col. 6, letter to
the editor from Michelle Lee, Toronto)
c What are my borrowing options? . . . How much can I afford? . . . Where
do I begin? (New York Times, 28.4.2003, p. A22, advertisement Fleet)
d Can she be held accountable for the problems that today’s nurses are grappling
with? (Washington Post, 29.4.2003, p. F1, col. 2)
The declarative sentences contain no special marking of the declarative force, the
interrogative pattern is signaled by the word order: the auxiliary precedes the subject.
We might propose that the interrogative pattern is formed by changing the
position of the auxiliary with respect to that of the subject. We refer to this process
as subject-auxiliary inversion, abbreviated as SAI. Now how exactly does SAI work?
In a declarative sentence we ﬁnd the order in (17a), in a question we get (17b):
(17) a declarative: subject – auxiliary
b interrogative: auxiliary – subject (SAI)
10
In the following discussion we will often equate the concepts “question” and “interrogative
sentence.” This equation would have to be challenged in a more careful analysis and we should
make a distinction between the two concepts. However in the framework of what we are
trying to do in this chapter the distinction is not crucial. For a good and accessible discussion
see Huddleston (1994).
11
Exercise 4.
TSC01 07/18/2005, 03:06PM22
Introduction: The Scientiﬁc Study of Language 23
Suppose we start from the order in the declarative sentence (17a) and try to attain
(“derive” to use the technical term) the order in the question (17b). How can we
relate the order in (17b) to that in (17a)? There are basically three options, which
are schematically summarized in (18). (18) contains three hypotheses about how
the order auxiliary – subject is formed or derived. The arrows are intended to show
the derivations, that is, which constituent is moved and where it is moved. According
to (18a), the auxiliary is moved to a position to the left of the subject; according
to (18b), the subject moves to a position to the right of the auxiliary; according to
(18c) subject and auxiliary switch places.
(18) SAI
a declarative sentence subject auxiliary verb
interrogative sentence auxiliary subject verb
b declarative sentence subject auxiliary verb
interrogative sentence auxiliary subject verb
c declarative sentence subject auxiliary verb
interrogative sentence auxiliary subject verb
How can we decide between these derivations? For examples like those in (16) it
is not clear how to decide. The three alternatives will produce the same end result:
the subject will end up to the right of the auxiliary. How could we differentiate the
three alternatives? In order to ﬁnd out which of the three hypotheses is preferable
we will run an experiment. We will create a sentence in which the outcome of
the three procedures in (18) would be different. Here’s an idea. Suppose we had a
declarative sentence in which something intervened between the subject and the
auxiliary, then the outcome of the different operations in (18) would be distinct.
Consider the following example:
(19) These new shops deﬁnitely are doing well.
Let us try out the three derivations for SAI illustrated in (18) on the basis of example
(19). Each derivation leads to a different pattern, as illustrated by (20). In (20a) the
auxiliary moves to a position to the left of the subject; in (20b) the subject moves to
a position to the right of the auxiliary; in (20c) the subject and the auxiliary switch
places. The acceptable word order is that in (20a). What would you conclude with
respect to the precise formulation of SAI?
(20) a Are these new shops deﬁnitely —— doing well?
b *—— Deﬁnitely are these new shops doing well?
TSC01 07/18/2005, 03:07PM23
24 Chapter 1
c *Are deﬁnitely these new shops doing well?
The data in (20) show that SAI is an operation in which the auxiliary moves from a
position to the right of the subject to a position to its left. We can now formulate a
rule for the derivation of interrogative sentences in English as in (21a). To this we
also add a speciﬁc formulation for deriving the order auxiliary – subject. (21b)
makes explicit what the process of SAI involves.
(21) a Interrogative sentences are formed by means of SAI.
b SAI: move the auxiliary leftward across the subject.12
To further test (21), we can invent additional examples with auxiliaries and check
whether the corresponding questions would be formed by moving the auxiliary to
the left of the subject. For example:
(22) a The murderer has broken the window.
b The murderer was arrested last night.
c We really must go to that meeting.
The prediction of (21) is that questions corresponding to (22) will be as in (22′):
(22) a′ Has the murderer broken the window?
b′ Was the murderer arrested last night?
c′ Must we really go to that meeting?
(21) formulates a hypothesis for turning a declarative sentence into an interrogative
sentence. Examples (22a′–c′) are compatible with this hypothesis. Observe that
underlying the hypothesis is a much more general hypothesis that form (word
order) and meaning are related. An additional underlying assumption in (21b) is
that in SAI the position of the subject and that of the verb are themselves unaffected
by SAI, only the auxiliary moves. SAI affects certain elements of the sentence but
not others.
2.2.2 WHEN THERE IS NO AUXILIARY
We started out from the observation that sentences may serve to make a statement,
in which case they are declarative, and they may be used to ask a question, in which
case they are interrogative. The form of the sentence encodes the difference in
interpretation: questions are formed by SAI, that is moving the auxiliary across the
subject. The examples in (23) pose a problem for applying SAI (21). Why is that?
How would we form the interrogative variant of the sentences?
12
Exercise 12 provides an additional speciﬁcation concerning the application of SAI.
TSC01 07/18/2005, 03:07PM24
Introduction: The Scientiﬁc Study of Language 25
(23) a He wants to buy a house this year.
b She wanted to become a policewoman.
The problem with (23) is that the sentences lack an auxiliary. Their interrogative
form is given in (24):
(24) a Does he want to buy a house this year?
b Did she want to become a policewoman?
Once again, we see that an element precedes the subject, does in (24a) and did
in (24b). Let us experiment with these sentences. Could the added elements have
occurred in the position to the right of the subject?
(25) a He does want to buy a house this year.
b She did want to become a policewoman.
Observe that the additional element do is inﬂected. The form of the ending of the
verb in (23) corresponds to that of do in (24) and (25): in (23a) the verb wants has
the third person singular ending -s; in (24a) and in (25a) does is a third person
singular of do. Similarly, in (23b) wanted is a past tense form of the verb; in (24b)
and (25b) did is a past tense form of do. The present tense form and the past tense
form of the verb are called the ﬁnite forms of the verb. When a sentence contains a
ﬁnite verb it is called a ﬁnite sentence. In the examples in (24) and (25) a ﬁnite form
of do is accompanied by a non-ﬁnite form of the verb want.
Apparently, both in questions and in declarative sentences, the elements does and
did can occupy the same positions as elements such as shall and is, the auxiliaries.
We will assume that does/did are also auxiliaries. So in interrogative sentences with
do, an auxiliary element is used to signal interrogative force, and the positions of
the subject and the verb (want in (24)) do not change.
Things are becoming complex here. When there is an auxiliary in the sentence we
move that auxiliary to the left of the subject to form a question. When there is no
auxiliary, we insert a form of the auxiliary do and invert that with the subject.
Could we have inserted do in sentences with auxiliaries? If you form questions on
the basis of the declaratives in (22), but by inserting do to the left of the subject, the
resulting patterns are those in (26), none of which is acceptable.
(26) a *Do we must go to that meeting?
*Do must we go to that meeting?
*Must do we go to that meeting?
b *Does the murderer has broken the window?
*Does has the murderer broken the window?
*Has does the murderer broken the window?
c *Did the murderer was arrested last night?
*Did was the murderer arrested last night?
*Was did the murderer arrested last night?
TSC01 07/18/2005, 03:07PM25
26 Chapter 1
Table 2 English question formation
Sentences with auxiliary Sentences without auxiliary
Insert do in the auxiliary position
Subject-auxiliary inversion (21b) Subject-auxiliary inversion (21b)
13
For more discussion of this idea see Chapter 5.
On the basis of the examples given, we could conclude that questions are encoded
by SAI. If there is already an auxiliary in the sentence we invert that auxiliary with
the subject, if there is no auxiliary available, we choose a variant of the sentence
with the auxiliary do, and invert do with the subject. We sum up our ﬁndings in
Table 2. Again, the formulations in this table rely on additional tacit assumptions.
For instance, the instruction to “insert do in the auxiliary position” implies that a
sentence has an “auxiliary position.” In more general terms, this implies that we
think of sentences in terms of particular positions or slots into which elements are
inserted and that certain types of units belong to certain types of positions. In the
next chapters we will make these assumptions more explicit.
Obviously, we also want to know why questions are formed using SAI. We want
to know why we insert do in sentences without auxiliaries. And we want to know
why we do not insert do in sentences with an auxiliary. We will attempt to formulate
a ﬁrst rough hypothesis, to be reﬁned in later chapters. We will say that questions
can be formed by subject-auxiliary inversion. We will further propose that this is
because the position to the left of the subject encodes question force.13
If there is no
auxiliary in the sentence, do is inserted as a sort of saving device, to enable us to
form the question. If there is an auxiliary in the sentence, inserting do is uneconomical,
since we already have all the ingredients to form a question. We only insert do as a
last resort. We return to this issue later, but it is important to signal here that while
formulating the proposal above we have introduced yet another general hypothesis.
The idea that we only introduce do as a last resort suggests that question formation
is somehow guided by a principle of “economy,” which says “Do not insert elements
if you don’t need them.”
At this stage the discussion of question formation remains highly informal, but
hopefully it can serve to illustrate how we proceed when elaborating hypotheses
in syntax. We start from some data, either attested data or constructed data, or a
mixture, and we move on to formulate one or more hypotheses to account for the
data. Then we increase the size of the data set and we test our hypothesis, modifying
it whenever necessary. While formulating our hypothesis we will probably introduce
further theoretical assumptions. We can introduce additional assumptions to
enable ourselves to formulate a general rule. For instance, we assume that there is a
relation between linguistic form and meaning. However, we must remain vigilant:
TSC01 07/18/2005, 03:07PM26
Introduction: The Scientiﬁc Study of Language 27
we should be aware of any additional assumptions that we have been relying on and
we should be prepared to examine these additional hypotheses themselves, possibly
at some later stage. When evaluating a particular hypothesis, we examine its empirical
coverage (the data which it can account for) and we also have to examine what
additional assumptions we have been relying on. It will be important to keep track
of any additional hypotheses because we need to make sure, for instance, that they
do not lead to contradictions in our system.
2.2.3 LANGUAGE AND ECONOMY
The idea that we only introduce the auxiliary do as a “last resort” suggests that
question formation is somehow driven by a principle of “economy”: “Don’t insert
forms if you don’t need them.” If we adopt this principle, another question arises:
Is the scope of the principle of economy conﬁned to question formation or does
it apply more generally?
Actually, keeping strictly to the use of English do, we have already come across
examples in which do occurs in a non-interrogative form. Was the use of do essential
in (25)? Or to put it differently: Are the sentences with do in (25) and those without
do in (23) exactly equivalent? If inserting do in non-interrogative sentences did
not make any difference, then the examples in (25) would contradict the economy
principle we have hinted at. They would be counter-evidence for the principle of
economy. However, when we study the relevant examples carefully we note that the
insertion of do in (25) (as compared to the original examples (23)) has some interpretive
effect, though it may be hard to pin down. Try to think of circumstances
where (25a) with do would be appropriate. One might imagine this in a context
such as the conversation in (27), in which doubts have been raised about Bill’s
intention to buy a house:
(27) Speaker A: I think Bill has changed his mind about buying a house. He is
redecorating his ﬂat.
Speaker B: He does want to buy a house this year. The redecoration of his
ﬂat is because he wants to add to its sales value.
The auxiliary do is inserted to strengthen an afﬁrmation against a background in
which some doubt has been raised about it. In (27B), the speaker uses do to counteract
the doubt expressed by the preceding utterance. This suggests that the auxiliary
do is not completely redundant in declarative examples and declarative sentences
containing the auxiliary do are not in contradiction with the hypothesis that there
is some principle of economy at work in language. Let us therefore maintain the
hypothesis that economy is a guiding principle in the formation of sentences.
Consider the underlined examples of do in the following extracts. What effect
does the presence of do have for the interpretation of the sentence?
(28) a Photographers aren’t allowed to alter their photos in a way that misleads
you, from posing a photo to digitally deleting a stray hair or telephone wire.
TSC01 07/18/2005, 03:07PM27
28 Chapter 1
The Post does allow photographers to do some things to their pictures.
They can “enhance for reproduction,” meaning they can adjust the colors
slightly so they will print better on the paper’s presses. (Washington Post,
10.12.2002, p. C14, col. 3)
b I am glad that Roy Grimwood points out the advantages our generation
(1960s) has had from university and which, thanks to the Thatcherite legacy,
we would deny others. However, while no doubt many graduates do earn
extra because of their qualiﬁcations, it must not be assumed that all do.
(Guardian, 7.12.2002, p. 11, col. 5, letter to the editor from Robert
Bracegirdle, Rothley, Leicestershire)
In both examples, the underlined auxiliary (does, do) serves to oppose the afﬁrmative
content of the sentence to a denial explicit or implicit in the context. If we delete
do we alter the meaning slightly in that we weaken the contrastive effect of the
sentences.14
2.3 From form to meaning: Subject-auxiliary inversion and
question formation
2.3.1 INTRODUCTION
We have seen that subject-auxiliary inversion (SAI) is used to form interrogative
sentences in English. There is a relation between form, the position of the auxiliary
in the sentence, and interpretation: SAI helps to show the difference between statements
and questions. We have not been fully explicit, though, about the nature of
the relation between SAI and interrogative interpretation. A more precise formulation
is called for. Is the relation between SAI and interrogative interpretation a strict
relation of cause and effect? Does the correlation imply that all English interrogative
sentences are necessarily formed by SAI? Does the correlation mean that SAI
necessarily gives rise to questions? Let us try to make the nature of the correlation
more exact.
We are investigating a form–interpretation relation: the form concerns a particular
word order pattern: SAI. We have interpreted it as a leftward movement of the
auxiliary across the subject. What exactly is the relation between SAI and interrogative
interpretation? There are a number of possible relations that might obtain.
We will compare the statements in (29) and try to assess which kinds of sentences
would be covered by each of the statements. Though the statements are similar, it
will soon turn out that the linguistic data they cover differ considerably. The statements
lead to different predictions.
14
Exercise 7.
For more discussion of examples with do see also Chapter 3, section 1.2.3.2.
TSC01 07/18/2005, 03:08PM28
Introduction: The Scientiﬁc Study of Language 29
(29) a SAI can give rise to interrogative interpretation.
b SAI always gives rise to interrogative interpretation.
c Interrogative sentences can be formed by means of SAI.
d Interrogative sentences are always formed by means of SAI.
e SAI is always used to form an interrogative sentence, and interrogative
sentences are always formed by means of SAI.
We will test the impact of these statements below. For each statement we will ﬁrst
make explicit its predictions and then we will try to test these predictions on the
basis of empirical data.
2.3.2 SAI CAN GIVE RISE TO INTERROGATIVE INTERPRETATION/
SAI ALWAYS GIVES RISE TO INTERROGATIVE
INTERPRETATION
The discussion above has shown that statement (29a) is valid. How does (29b)
differ from (29a)? (29a) is a statement about a potential (“can”) use of SAI. If we
think about it carefully, we will conclude that (29a) is equivalent to (30a): the
predictions of (29a) and (30a) are identical. (29a) says that SAI can be associated
with question interpretation.
(30) a Some sentences with SAI are interrogative.
To formulate (29b) we have removed the modal auxiliary can and we have added
the adverb always. Compared to (29a), (29b) is stronger: it says that whenever SAI
applies you create an interrogative sentence. (29b) does not imply a potential
outcome of SAI, it implies a deﬁnite outcome. (29b) is equivalent to (30b):
(30) b All sentences with SAI are interrogative.
Consider the underlined fragments in the examples in (31). Do they bear on the
statements we are examining? Let us ﬁrst examine if they are compatible with the
formulation in (29a).
(31) a All he wants to know is which boxes have I ticked on the forms he keeps
giving me to ﬁll in. (Guardian, G2, 15.3.2001, p. 9, col. 8)
b People ask why was I not at Coniston when Bluebird was raised, but I
would have been far too emotional. (Guardian, 15.3.2001, p. 5, col. 8)
With respect to (29a) there is no problem: the underlined strings in (31a) and (31b)
display SAI and they are interrogative.15
The interrogative sequences in (31) have
been integrated into a larger structure. Sentences that become part of larger sentences
15
We will see later that there is something special about the word order in these sentences
(Chapter 5, section 2.6), but this point is not relevant to the current discussion.
TSC01 07/18/2005, 03:08PM29
30 Chapter 1
are said to be embedded or subordinate clauses.16
In (31a) the string which boxes
have I ticked on the forms he keeps giving me to ﬁll in is an embedded clause which
is also interrogative, an embedded interrogative for short. Embedded questions are
also called indirect questions, or reported questions. (31c) is an independent question
or a direct question: the sentence is addressed by a speaker to an interlocutor and
the speaker expects an answer from the interlocutor.
(31) c Which boxes have you ticked on the forms he keeps giving you to ﬁll in?
We can form additional examples of direct questions of our own, which will again
be compatible with (29a):
(31) d Why were you not at Coniston when Bluebird was raised?
e Have you done linguistics before?
f Would all linguists agree that linguistics is a science?
Let us return to (29b). How could we evaluate its empirical coverage? We have seen
that the claim in (29b) is stronger than that in (29a). (29b) says that whenever you
have SAI you have an interrogative sentence, i.e. a sentence that can be used as a
question. (29b) is a general rule for the interpretation of SAI. We reworded (29b) in
(30b) above. (29b)/(30b) implies that there is a necessary link between the pattern
of SAI and the interrogative force of the sentence. How can we evaluate this type of
general statement?
Let us think of similar general statements that are not about language. Consider
the statement: “All swans are white.” It is a universal claim about the color of
swans. To prove the universal claim true it is not sufﬁcient to ﬁnd a number of
white swans. Even having found thousands of white swans, we cannot be sure that
the next swan might not be non-white. On the other hand, if we ﬁnd just one nonwhite
swan, we will have shown that the statement “All swans are white” is false.
In other words, in order to evaluate a general statement, we do not really show the
statement to be true, but we look for evidence to falsify it, for “counter-evidence.”
Now we return once again to (29b)/(30b). What kind of data would count as
counter-examples to the generalization? Relevant counter-evidence would be sentences
that display SAI and which are not interrogative. Are there such sentences?
To test the claim, we could try to devise sentences of our own with SAI and which
are not interrogative or we could try to ﬁnd attested examples that contradict the
claim. In the ﬁrst procedure, we rely purely on our intuitions as speakers of English.
This procedure might appear risky and less objective than the second procedure
where we could claim the evidence is “objective” because it exists independently.
However, note that this is not necessarily true: even in the second procedure we still
need to rely on our intuitions because as speakers we have to decide whether the
function of SAI in the relevant sentences is or is not that of forming a question. We
16
We return to the form of embedded clauses in more detail in Chapter 5.
TSC01 07/18/2005, 03:08PM30
Introduction: The Scientiﬁc Study of Language 31
still need to interpret the evidence and for that, we need to rely on our intuitions. So
whichever line of attack we use, our own intuitions are involved.17
Are there any sentences that display the SAI pattern and which are not interrogative?
Consider the following examples: (32a) is constructed, (32b–c) are
attested. Does SAI give rise to an interrogative interpretation in these examples? If
not, what is the interpretive contribution of SAI in these sentences?
(32) a Had I known you were coming, I would have baked a cake.
b The guests were being offered cushions to take with them should their free
seats prove insufﬁciently comfortable. (Guardian, 1.7.2002, p. 4, col. 8)
c Had the money not been returned, the evidence would have pointed strongly
to a conclusion that the NRCC “ﬁnanced” the Forum. (Washington Post,
29.4.2003, p. A18, col. 3)
In (32a–c) SAI is used in conditional clauses. These examples can be paraphrased
as in (33):
(33) a If I had known you were coming, I would have baked a cake.
b The guests were being offered cushions to take with them in case their free
seats should prove insufﬁciently comfortable.
c If the money had not been returned, the evidence would have pointed
strongly to a conclusion that the NRCC “ﬁnanced” the Forum.
Now consider the following examples:
(34) a Not one word of evidence have they brought to support that. (Guardian,
11.12.2001, p. 4, col. 7)
b Within a year of Hague becoming leader, the party had a ballot of its
membership to say that not within the lifetime of this parliament would
Britain enter the Euro. (Guardian, G2, 13.5.2002, p. 7, col. 2)
In (34a) and (34b) the inverted auxiliary (have, would) is preceded by a negative
element (not one word of evidence, not within the lifetime of this parliament).
If these negative elements are removed from the initial positions, the SAI pattern is
no longer possible (35). In (34a) and (34b) SAI seems to be caused (“triggered”) by
the presence of the negative element in initial position.18
(35) a They have not brought one word of evidence to support that.
b Within a year of Hague becoming leader, the party had a ballot of its
membership to say that Britain would not enter the Euro within the lifetime
of this parliament.
17
See also the discussions in Borsley and Ingham (2002, 2003) and Stubbs (2002).
18
This pattern is sometimes called negative inversion. Exercise 6 of Chapter 5 is an exercise on
negative inversion.
TSC01 07/18/2005, 03:08PM31
32 Chapter 1
We conclude that SAI may give rise to question formation but that it is not exclusively
used for that purpose. SAI may also be used to form conditional clauses and
in sentences with negative elements in the initial position.19
We conclude from the discussion that (29b) and its paraphrase (30b) are not empirically
adequate. If (29b)/(30b) were adequate we should not be able to create sentences
like those in (32) and in (34), where SAI occurs in non-interrogative sentences.
2.3.3 INTERROGATIVE SENTENCES CAN BE FORMED BY MEANS
OF SAI/INTERROGATIVE SENTENCES ARE ALWAYS FORMED
BY MEANS OF SAI
Let us turn to the second set of statements, (29c) and (29d). They are repeated here,
with (29a), for convenience’s sake. First compare the formulation of (29c) with
(29a). What makes them different?
(29) a SAI can give rise to interrogative interpretation.
c Interrogative sentences can be formed by means of SAI.
d Interrogative sentences are always formed by means of SAI.
Both statements imply a causal relation: some factor A gives rise to/causes some
result B. To bring out the causal relation more clearly, let us try to paraphrase the
statements using a conditional sentence, i.e. using the pattern “if A then B.” (29a)
corresponds to (36a), (29c) to (36b):
(36) a If we apply SAI to a sentence, it can become interrogative.
SAI → question
b If we want to form an interrogative sentence, we can use SAI.
Question → SAI
The paraphrases reveal that the two statements differ in the way they conceive of
the cause-effect sequence. Statement (29a) takes the form (the word order arising by
SAI) as the starting point and conceives of the interpretation (“question”) as the
result of that form; statement (29c) takes the interpretation as the starting point and
predicts the form that will be used to convey it. We have looked at the empirical
adequacy of (29a)/(36a) already. What kind of examples would support (36b)?
(36b) predicts that interrogative sentences may display SAI. The examples discussed
in the preceding sections are compatible with the statement. Some of them are
straightforward illustrations of (36b) in that they illustrate questions formed by
means of SAI. Examples in which SAI does not lead to interrogative sentences are
irrelevant because like (29a), (29c) does not make a general over-arching statement
about the shape of interrogative sentences. It does not raise the “white swan”
problem posed by (29b) that was discussed above.
19
One may try to propose a more general explanation to account for why inversion applies in
just these three environments. We will not do this here.
TSC01 07/18/2005, 03:08PM32
Introduction: The Scientiﬁc Study of Language 33
(29c) says that if we want to form interrogative sentences we sometimes use SAI.
By comparison, (29d) is much stronger. This is shown by the conditional paraphrase
(36c):
(36) c If we want to form an interrogative sentence we always use SAI.
Both (29c) and (29d) concern properties of interrogative sentences. (29c) says that
an interrogative sentence may display SAI. (29d) says that it must display SAI, or,
put differently, that SAI is a necessary property of questions. (29c) corresponds to
(37a), (29d) to (37b):
(37) a Some interrogative sentences display SAI.
b All interrogative sentences display SAI.
How can we test these statements? We have assumed that the validity of (29c) was
conﬁrmed by the fact that some interrogative sentences have SAI. To test (29d),
would it be sufﬁcient to be able to construct or to ﬁnd interrogative sentences with
SAI? Would sentences such as those in (32) and (34), in which SAI is used in noninterrogative
sentences, be relevant? The answer to both questions is negative. Take
the second point ﬁrst. The fact that SAI may be used in sentences which are not
interrogative does not bear on (29d). It is irrelevant to the issue since (29d) is only
concerned with sentences that are interrogative. Secondly, ﬁnding or constructing
examples of interrogative sentences which display SAI does not “prove” that interrogative
sentences must always be formed by SAI. We need to show that all interrogative
sentences are formed by SAI. Again we are in a position in which we have
to test a universal claim. What then would be counter-evidence for (29d)? What we
need to ﬁnd is interrogative sentences (i.e. questions) without SAI. Do such data
exist? Consider the examples in (38) and compare them with (31a, b):
(38) a All he wants to know is which boxes I have ticked on the forms he keeps
giving me to ﬁll in.
b People ask why I was not at Coniston when Bluebird was raised, but I
would have been far too emotional.
In (38a) the underlined string is an indirect question: (38a) reports a question
uttered by the person referred to by the pronoun he. Similarly (38b) reports a question
asked by people. The indirect questions in the underlined sections in (38a) and
(38b) differ from those in (31a, b) in that they do not display SAI. The auxiliaries
have and was follow their respective subjects. In fact, there is dialectal variation in
terms of the form of embedded questions. For many speakers of English, the SAI
pattern in the embedded interrogatives in (31a, b) is slightly unusual and the noninverted
pattern in (38) would be more usual. To form an (embedded) interrogative
sentence, SAI is not required. Statement (29d) is too general: it does not allow us to
predict that the underlined strings in (38a, b) also qualify as interrogative.
TSC01 07/18/2005, 03:08PM33
34 Chapter 1
Would the examples in (38) pose a problem for statement (29c)? In other words,
do they constitute counter-evidence to the claim that “some interrogative sentences
have SAI”? Clearly not, (29c) leaves it open that there may be questions without
SAI. The weaker formulation in (29c) is compatible with the data; as far as we can
assess at this point, it is empirically adequate. The stronger formulation in (29d) is
not empirically adequate because it would lead us to expect SAI always to apply in
embedded questions, which is not the case. Because of these observations, we might
wish to narrow down statement (29d) to apply only to direct questions. This is
shown in (39a) and in its conditional paraphrase (39b):
(39) a Direct questions are always formed by means of SAI.
b If you want to form a direct question you always use SAI.
The new formulations predict (39c):
(39) c All direct questions display SAI.
At ﬁrst sight, this universal claim may seem plausible, and we have found many
examples that seem to conﬁrm it. But consider the examples in (40). Are they direct
questions? The answer is positive. Now let us locate the subject and the auxiliary
and evaluate the empirical coverage of (39c):
(40) a Which student will ﬁnish the exam ﬁrst?
b How many images will be remembered or become symbols of the war in
Iraq? (Guardian, G2, 16.4.2003, p. 11, col. 1)
(40) illustrates direct questions. In these examples, however, the subject is not
preceded by the auxiliary. In (40a) the subject is which student and the auxiliary
will follows it. Similarly, in (40b) the subject how many images is followed by the
auxiliary will. In other words, these direct questions do not display SAI. So we
cannot claim that all direct questions necessitate SAI.
If we turn to the last statement in (29) repeated below, it is clear that this cannot
work either:
(29) e SAI is always used to form an interrogative sentence, and interrogative
sentences are always formed by means of SAI.
(29e) conjoins statements (29b) and (29d), both of which were independently shown
to give rise to counter-examples. The counter-evidence raised against the independent
statements also falsiﬁes (29e). We could replace (29e) by a weaker statement,
the conjunction of (29a) and (29c):
(29) f SAI can give rise to interrogative sentences, and interrogative sentences can
be formed by means of SAI.
TSC01 07/18/2005, 03:08PM34
Introduction: The Scientiﬁc Study of Language 35
2.3.4 VERBS AND INVERSION
Statement (29f) summarizes our ﬁndings, but unfortunately it is relatively weak and
not very insightful. It says that interrogative sentences may display SAI and that
we may ﬁnd SAI in interrogative sentences. We do not claim a strong correlation
between the form (SAI) and the interpretation (question) and we also do not claim
to be providing an explanation. We know that the statement is an appropriate
summary of the data, it is descriptively adequate. But disappointingly, we have not
really made any predictions as to the difference between data which will be found
and those that will not be found. (29f) is not a general rule or a law, since it allows
both for cases in which SAI does not give rise to interrogative sentences and for
cases in which interrogative sentences do not display SAI.
However, in the discussion we did actually come across a pattern that is much
closer to a general rule than (29f). Recall our discussion of question formation with
respect to examples without auxiliaries such as (23), repeated here as (41) for
convenience:
(41) a He wants to buy a house this year.
b She wanted to become a policewoman.
We said that in order to form a question with inversion we inserted do, “as a last
resort.” Apparently when there is no auxiliary in a sentence we cannot simply invert
the verb with the subject to form a question: the examples in (42) are unacceptable.
(42) a *Wants he to buy a house this year?
b *Wanted she to become a policewoman?
Recall that we discovered that SAI is used in some other contexts. In the examples
in (32) in section 2, SAI was shown to be used to form conditional clauses and in
(34) we found it used in contexts with a sentence-initial negative element. In the
underlined conditional clauses in (43) there is no auxiliary. Would these contexts
allow a verb to invert with the subject?
(43) a If I knew you were coming, I could have baked a cake.
b The guests were being offered cushions to take with them, in case their free
seats proved insufﬁciently comfortable.
c If the money remained unaccounted for, the evidence would have pointed
strongly to a conclusion that the NRCC “ﬁnanced” the Forum.
It is not possible to rephrase these examples using inversion of the verb and the
subject:
(44) a *Knew I you were coming, I could have baked a cake.
b *The guests were being offered cushions to take with them proved their free
seats insufﬁciently comfortable.
TSC01 07/18/2005, 03:08PM35
36 Chapter 1
c *Remained the money unaccounted for, the evidence would have pointed
strongly to a conclusion that the NRCC “ﬁnanced” the Forum.
The underlined elements in (45) are negative. Would fronting that constituent entail
inversion of the verb and the subject?
(45) a They brought not a shred of evidence to support that claim.
b Within a year of Hague becoming leader, the party had a ballot of its
membership to guarantee that Britain never enters the Euro.
Once again, even if the negative elements were fronted, the verbs would not invert
with their subjects:
(46) a *Not a shred of evidence brought they to support that claim.
b *The party had a ballot of its membership to guarantee that never enters
Britain the Euro.
These data do allow the formulation of a “general law”: verbs, as opposed to
auxiliaries, cannot invert with the subject in interrogative sentences, in conditionals
or as a result of a sentence-initial negative constituent.
(47) Verbs that are not auxiliaries do not invert with the subject.20
(47) is general and it does succeed in deﬁning the bounds of possibilities in that it
radically excludes sentences such as (42), (44), and (46). The statement allows us to
predict that we will never ﬁnd such sentences and that if we construct them they
will be unacceptable.
(47) raises questions. First we need to try to deﬁne auxiliaries and we need to
make explicit how they differ from non-auxiliaries. We turn to this point in section
2.4.21
Second (47) is still only a generalization of facts, (47) is not an explanation.
We should not stop here but we should ask ourselves why it is that auxiliaries invert
with the subject and verbs do not, in other words why they differ in their distribution.
The question is like asking ourselves why snow melts in the sunshine and glass
does not. We will return to an explanation of the difference in distribution between
auxiliaries and verbs in Chapter 3, section 3.4.
2.4 A brief discussion of deﬁnitions
We have mentioned that scientiﬁc research has to be explicit. Being explicit means,
among other things, that the terminology used is transparent, that terms are deﬁned
20
Exercise 11 deals with an apparent counter-example to (47).
21
We will come back to the distinction in Chapter 3, section 3.4.
TSC01 07/18/2005, 03:08PM36
Introduction: The Scientiﬁc Study of Language 37
clearly and unambiguously and that they are used in a systematic way. Obviously,
in the preceding section we have not been fully explicit ourselves: we have not
deﬁned all the terms that we have been using, we have used some terms relying on
a previous understanding of them, in a rather vague pre-theoretical way. This is
because we all come to the task with some terminology and we have implemented
that in order to be able to start dealing with some data. This type of approach is just
about acceptable as long as we bear in mind that we have not yet deﬁned our terms
and that sooner or later (and preferably sooner rather than later) we will deﬁne the
technical terms used.
In this section we brieﬂy look at two terms that we have been using in a pretheoretical
sense and we will try to make them precise: these terms are auxiliary and
verb. What is a verb? Consider the following examples and identify which words
you think are verbs:
(48) a They bought books for their children.
b They wrote novels about the war.
Probably you will pick out the words bought and wrote. Why do we call these
words verbs? Often verbs are deﬁned as words denoting an action. Using this deﬁnition
as a guideline, which words would you identify as verbs in (49)?
(49) a His actions seemed incoherent.
b Her friends’ complaints remained secret.
In (49a) the word actions denotes actions but it will not normally be considered a
verb. Similarly in (49b) the word complaint may be said to denote an action, but it
is not a verb. On the other hand, words like seemed in (49a) and remained in (49b)
are normally classiﬁed as verbs even though they do not denote actions. Why do we
classify seemed and remained as verbs, with bought and wrote? What do these
words have in common? What these words share is that they have a set of forms
that differentiate them from the words that are not classiﬁed as verbs. For example,
verbs have a form referred to as the “past tense”; bought in (48a), wrote in (48b),
seemed in (49a), and remained in (49b) are all past tense forms. We experiment
with the sentences in (48) and in (49) and remove the past tense form of the verbs.
What does the difference in verb form do? Does it have an interpretive effect?
(50) a They buy books for their children.
b They write novels about the war.
c His actions seem incoherent.
d Her friends’ complaints remain secret.
While the sentences in (48) and (49) situate the state of affairs they denote in the
past time sphere, those in (50) situate it in the present time sphere. The -ed ending
of the past tense form here serves to indicate that the situation described by the
TSC01 07/18/2005, 03:09PM37
38 Chapter 1
sentence is not to be situated in the present. The form remained is composed of two
units: the verb remain, and the inﬂectional ending -ed. Remain means roughly ‘stay’,
-ed means roughly ‘not situated in the present/situated in the past’. Both components
are units of form; they can be used in other circumstances: remain does not need to
be associated with -ed, and -ed combines with other verbs than remain. The units
remain and -ed themselves are units of meaning and they cannot be decomposed
into smaller units with identiﬁable meanings. We call such units morphemes and we
will say that -ed is an inﬂectional morpheme. The morpheme -ed is also called a
bound morpheme because it can never be used on its own; it must be attached to
another morpheme. Remain is a free morpheme: it can be used on its own.
Often grammarians will refer to the verb forms without -ed in (50) as the present
tense. But note that there is in fact no real tense marking at all in the form of the
verb. The past tense ending -ed is not replaced by a speciﬁc ending for the present, the
form remain is simply the base form of the verb. In the present tense, the verb gets
a special ending only in examples with third person subjects such as those in (51).
(51) a She buys books for her children.
b She writes novels about the war.
c His action seems incoherent.
d Her friend’s complaint remains secret.
The so-called present tense and the past tense are said to be ﬁnite forms of the verb.
In the examples in (52a, b) the verbs buy and write combine with the ending -ing.
(52) a She is buying books for her children.
b She is writing novels about the war.
-ing is also an inﬂectional morpheme. In (52a, b) it serves to create the so-called
present participle of the verb, which follows is, itself a ﬁnite form of the auxiliary
be. This auxiliary cannot be followed by the ﬁnite forms of the verb (present tense
or past tense) (52c, d) and we have to use the -ing form. The -ing form of the verb
is a non-ﬁnite form. Finite forms vary for tense (is is a present tense, was is a past
tense), non-ﬁnite forms do not vary for tense.
(52) c *She is buys books for her children.
d *She is writes novels about the war.
Observe that the past tense ending (-ed), the third person singular ending (-s), and
the -ing ending typically are attached to verbs. They are not found on other categories
such as nouns, such as cat, or girl.22
So we can identify verbs by looking at
their morphology, at the kinds of inﬂectional morphemes they combine with.
22
When we ﬁnd the -s ending on a noun it is used either for a plural (cats, girls) or to form a
genitive (cat’s, girl’s). We assume that plural -s and genitive -s are different entities.
TSC01 07/18/2005, 03:09PM38
Introduction: The Scientiﬁc Study of Language 39
The uninﬂected base form of the verb may typically be found in contexts such as
these illustrated in (53):
(53) a She will buy books for her children.
b To write a novel about the war would be hard.
Buy and write in (53) do not have any endings. Even though the subject in (53a) is
a third person, she, the verb could not possibly be marked with the -s ending, nor
could we replace buy by the past tense form.
(53) c *She will buys books for her children.
d *She will bought books for her children.
Typically, words like will and the inﬁnitive marker to are followed by the uninﬂected
base form of the verb. This base form is also referred to as the inﬁnitive of the
verb. We can also use this distributional information to identify a verb: if a word can
follow the auxiliary will or the inﬁnitive marker to, then that word will be a verb.
To ﬁnd out if a word qualiﬁes as a verb, we can check whether it has the
morphological and distributional properties of verbs such as buy or remain in
the examples above: that is to say we examine whether the word can associate
with the typical verb endings (-s, -ed, -ing), and whether it can occupy positions
typically occupied by verbs. Identify the verbs in the following example (which
corresponds to example (31a) in section 2). You may obviously experiment with the
sentence to check whether the items that you have identiﬁed as verbs have the
various forms we have been talking about.
(54) a All he wants to know is which boxes have I ticked on the forms he keeps
giving me to ﬁll in. (Guardian, G2, 15.3.2001, p. 9, col. 8)
In (54a) wants is a verb. For one thing it has the third person ending -s, and it can
also be used in a past tense. In the non-ﬁnite form want can be used after will or
after to:
(54) b All he wanted to know is which boxes have I ticked on the forms he keeps
giving me to ﬁll in.
c All he will want to know is which boxes have I ticked on the forms he
keeps giving me to ﬁll in.
d All he seems to want to know is which boxes have I ticked on the forms he
keeps giving me to ﬁll in.
Other verbs in (54a) are know, have, tick, keep, give, ﬁll. Recall that (54a) contains
an indirect question signaled by the inversion of the auxiliary have. Using the formal
criteria set out above, words that are referred to as “auxiliaries” are also verbs. The
auxiliary have displays the morphological and distributional properties of a verb:
TSC01 07/18/2005, 03:10PM39
40 Chapter 1
(55) a He has ticked the boxes.
b I had ticked the boxes.
c Having ticked the boxes, I handed in the form.
d I will have ticked all the right boxes.
e I expect to have ticked the wrong boxes.
(56) illustrates the so-called progressive use of be: the auxiliary be is used in combination
with the present participle to express (roughly) an ongoing activity.23
(56) a The teacher is meeting the students.
b The teacher was meeting the students.
c The teacher has been meeting the students for a while.
d The teacher will be meeting the students.
e I expect to be meeting the students.
Morphologically and distributionally, auxiliaries are a subclass of verbs. Why then
do we not just call them verbs? We saw in the preceding sections that auxiliaries
have distinctive distributional properties that set them apart from the other verbs.
Typically, auxiliaries can invert with the subject – for instance in interrogative
sentences, in conditional clauses, or if a negative element has been fronted.24
In the
same circumstances, non-auxiliary verbs do not invert with their subjects.
There are further properties that set apart auxiliaries among the class of verbs.
Auxiliaries can be followed immediately by the negation marker not, and they can
also contract with not. Ordinary verbs like meet do not invert with the subject and
the use of the negation marker not in a ﬁnite sentence gives rise to the insertion of
the auxiliary do.
(57) a The teacher is meeting the student.
b Is the teacher meeting the student?
c The teacher is not meeting the student.
d The teacher isn’t meeting the student.
(58) a The teacher met the student.
b *Met the teacher the student?
b′ Did the teacher meet the student?
c *The teacher met not the student.
c′ The teacher did not meet the student.
d *The teacher meetn’t the student.
d′ The teacher didn’t meet the student.
23
Exercises 5 and 6.
24
Exercise 12 shows that not every sequence consisting of auxiliary be + subject instantiates
SAI.
TSC01 07/18/2005, 03:10PM40
Introduction: The Scientiﬁc Study of Language 41
In what follows we will occasionally (re)deﬁne some traditional terms.25
Since
auxiliaries are verbs, it is difﬁcult to set up the opposition shown in (59a). We need
to differentiate verbs that are not auxiliaries from those that are. The terms that are
usually used for this opposition are given in (59b, c).
(59) a auxiliary ↔ verb
b auxiliary ↔ lexical verb
c auxiliary ↔ full verb
The term auxiliary is related to a Latin verb auxiliare, which means ‘help’. Informally,
we could say that auxiliaries “help” the lexical verb that follows them. The term
“full verb” suggests that auxiliaries, in contrast to full verbs, are “not full.” Full verbs
have a lot of descriptive content: a change of full verb in a sentence may radically
change the action referred to in that sentence. A change of auxiliary will not alter
the action referred to, but it will perhaps shift the action in time or make it less
plausible. The term lexical is related to the word lexicon. The lexicon of a language
is its dictionary, the list of its words. The list of what we call lexical verbs is open
ended. We can add new lexical verbs to the language. Recent English words are
pedestrianize, download, digitize, etc. Verbs are open class words (and so are nouns,
adjectives, and adverbs). On the other hand, auxiliaries belong to a closed class. We
do not normally expect people to start creating new auxiliaries.26
3 Language and Languages
3.1 Going further aﬁeld: Comparative syntax
We have discovered that the distribution of English auxiliaries differs from that of
English lexical verbs. Section 2.4 brieﬂy discusses how to set off the two subclasses
of verbs by looking at their distributional properties. The question arises why lexical
verbs cannot invert with their subjects in questions and why auxiliaries can. Before
we can address this question, though, we must check how general this restriction on
inversion is.
In the introduction to this chapter we said that linguistics is the scientiﬁc study of
language. Language as an abstract concept manifests itself through individual languages:
English, French, German, etc. We have discovered that English verbs do not
invert with their subjects in questions. Before trying to account for this restriction,
we might do well to check whether this ban on verb inversion is general. The reason
why we should do this is the following: if we discover that the ban on inversion is
universal, i.e. if it applies to all languages, then we can try to explain it by a very
25
For a general discussion of the complexities of deﬁning word classes such as verbs and auxiliaries
see for instance Aarts and Haegeman (forthcoming).
26
We return to the oppositions in (59) in Chapter 3, section 3.4.
TSC01 07/18/2005, 03:10PM41
42 Chapter 1
powerful linguistic principle, one that will not vary cross-linguistically. On the
other hand, if we discover that subject-verb inversion is banned in English but is
possible in other languages, then we know that we must provide a ﬂexible explanation,
one that can be adjusted to the properties of the individual languages. In other
words, the following possibilities could be considered:
(60) a There is a universal ban on inverting the verb with the subject.
b There is a language speciﬁc ban on inverting the verb with the subject. The
ban applies in the following languages: English, . . .
To account for a universal pattern, we will have to invoke a universal principle that
rigidly applies to all languages. If there is no such universal pattern, we will have to
devise a parameter along which languages may vary. We will then try to relate the
observed cross-linguistic variation to speciﬁc properties of individual languages. In
what follows we will try to assess which of the two statements in (60) is correct.
3.1.1 FRENCH
Consider the French examples in (61) and their word-for-word glosses in English.
Verbs and auxiliaries are given in bold. It is clear that for many of the examples,
the word-for-word glosses would not qualify as English sentences. In the light of
the discussion above, what conclusions could we draw with respect to inversion
in French? Would you say that French is more “liberal” than English in its use of
inversion? Or is it more restricted? Motivate your answer.
(61) a Achetait-elle le journal tous les jours? A-t-elle acheté le journal?
bought she the paper all the days has she bought the paper
‘Did she buy the paper every day?’ ‘Has she bought the paper?’
b Ecrit-elle des romans? A-t-elle écrit des romans?
writes she novels has she written novels
‘Does she write novels?’ ‘Has she written any novels?’
c Vient-il demain? Est-il venu hier?
comes he tomorrow is he come yesterday
‘Is he coming tomorrow?’ ‘Did he come yesterday?’
d Que dit-il? Qu’a-t-il dit?
what says he what has he said
‘What does he say?’ ‘What has he said?’
The English versions of the examples in the left most column read as in (62).
(62) a Did she buy the paper every day?
b Does she write novels?
TSC01 07/18/2005, 03:10PM42
Introduction: The Scientiﬁc Study of Language 43
c Does he come tomorrow?
d What does he say?
With respect to the position of lexical verbs, French seems to be more liberal than
English in that such verbs invert with the subjects, as do auxiliaries. Let us refer to
a pattern in which the full verb inverts with the subject as subject-verb inversion,
or SVI. Languages differ with respect to whether they allow SVI: English doesn’t
allow SVI, French does.
3.1.2 GERMAN AND DUTCH
We conclude that the difference in inversion patterns distinguishes French from
English. French is a Romance language, English is a Germanic language. Could it be
that this difference accounts for the difference observed here? Perhaps Germanic
languages in general never allow SVI. Let us look at some Germanic languages like
German or Dutch. Do they allow verbs to invert with their subjects? If we turn to
German, we ﬁnd that there too the verb can precede the subject. Verbs and auxiliaries
are in bold.
(63) a Kaufte sie jeden Tag die Zeitung? Hat sie jeden Tag die Zeitung gekauft?
bought she every day the paper has she every day the paper bought
‘Did she buy the paper every day?’ ‘Has she bought the paper every day?’
b Schreibt sie Romane? Hat sie Romane geschrieben?
writes she novels has she novels written
‘Does she write novels?’ ‘Has she written any novels?’
c Kommt er morgen? Wird er morgen kommen?
comes he tomorrow will he tomorrow come
‘Is he coming tomorrow?’ ‘Will he come tomorrow?’
d Was sagt er? Was hat er gesagt?
what says he what has he said
‘What does he say?’ ‘What has he said?’
Similarly, the Dutch examples in (64) display SVI:
(64) a Koopt zij elke dag de krant? Heeft zij de krant elke dag gekocht?
bought she every day the paper has she the paper every day bought
‘Did she buy the paper every day?’ ‘Has she bought the paper every day?’
b Schrijft zij romans? Heeft zij romans geschreven?
writes she novels has she novels written
‘Does she write novels?’ ‘Has she written any novels?’
TSC01 07/18/2005, 03:10PM43
44 Chapter 1
c Komt zij morgen? Zal zij morgen komen?
comes she tomorrow will she tomorrow come
‘Is she coming tomorrow?’ ‘Will she come tomorrow?’
d Wat zegt zij? Wat heeft zij gezegd?
what says she what has she said
‘What does she say?’ ‘What has she said?’
From the comparative data above we conclude that there is cross-linguistic
variation with respect to the possibility of SVI. This means that postulating a universal
ban on SVI would lead to the wrong predictions. Any attempt at explaining
the ban on SVI in English will have to take into account speciﬁc properties of that
language.
3.2 Going back in time: Diachronic variation
We have shown the relevance of using comparative data for linguistics research.
On the basis of the small sample of languages discussed in the previous section,
we concluded that the ban on inverting English verbs with subjects should not be
stated as a “universal law of language.” Languages vary with respect to their word
orders. Even if we are mainly (or only) interested in English, it is still useful to
introduce the comparative angle, because this will mean that we can situate the
English data in a wider perspective and we know what kind of explanation to
look for.
Could we conclude that the ban on verb inversion is an inherent rule of English?
Consider the data from Old English (or Anglo-Saxon) in (65). Auxiliaries and verbs
are in bold. Do auxiliaries invert with their subjects in questions? Does Old English
have verb inversion?
(65) a Hwi sceole we oqres mannes niman?
why should-1pl we another man’s take
‘Why should we take what belongs to another?’
(Ælfric, Lives of Saints, 24, 188, Haeberli 2000: 110)
b Hwi noldest Du hyt secgan me?
why neg+would-2sg you it say me
‘Why would you not tell it to me?’
(Gen, 31.27, Kroch and Taylor 2000: 152)
c Hwilcne oqerne sige sceolde ure drihten syllan?
what other victory should-3sg our Lord give
‘What other victory should our Lord give . . . ?’
(ÆLS, 31.128–9, Pintzuk 1991: 53)
TSC01 07/18/2005, 03:11PM44
Introduction: The Scientiﬁc Study of Language 45
d Hwæt sægest qu, yrDling?
what say-2sg you, peasant
‘What do you say, farmer?’
(Acoll, 22, Van Kemenade 1987: 138)
e Hu begæst qu work qin?
how beget-2sg you work your
‘How do you carry out your work?’
(Acoll, 22, Van Kemenade 1987: 138)
f To hwæm locige ic buton to Dæm eaDmodum?
to whom look-1sg I except to the humble
‘To whom do I look except to the humble?’
(CP, 41.299.18, Fischer et al. 2000: 54)
In each of the examples (65d–f) a full verb inverts with the subject. Old English
seems to behave more like German, Dutch, and French than like Modern English.
In particular, Old English displays SVI in questions. Why should that be? In order
to understand this, we would have to examine Old English in more detail to detect
what property or properties will set it off from present-day English and what makes
it more similar to French or to Dutch and German.27
When we consider the historical development of a language we engage in what
is called diachronic linguistics. We examine earlier stages of the language and
compare them with later stages. Obviously, since there are no native speakers of
Old English we have to turn to attested material.
3.3 Comparative data: Conclusion
We conclude that languages vary with respect to inversion patterns. Inversion affects
full verbs in some languages but not in others. This is summarized in Table 3. A
27
We will attempt to offer an explanation for the difference in inversion patterns between Old
English and Modern English in Chapter 3, sections 1.2.4.2, 1.2.4.3, and 1.2.4.4.
Table 3 Inversion patterns: classiﬁcation of languages
Language SAI SVI
Modern English + −
Old English + +
French + +
German + +
Dutch + +
TSC01 07/18/2005, 03:11PM45
46 Chapter 1
classiﬁcation like that in Table 3 is only a ﬁrst step in our research. Remember that
our aim is to formulate the “laws” of language. In the next chapters we will, among
other things, try to explain why languages differ in this particular way.
4 Summary
This chapter sets the scene for the remainder of the book. We have ﬁrst discussed
the properties of the scientiﬁc method. One core concern is that science presupposes
that we aim at providing an explanation of data. We set out to provide general
accounts for particular data. We try to attain this goal by formulating hypotheses
that relate sets of empirical data in terms of cause and effect. Such hypotheses lead
to the elaboration of theories, sets of interacting hypotheses. The hypotheses and
the theory also allow us to make predictions about what will be possible (the data
that we may ﬁnd) and what will be impossible (the data that we should never ﬁnd).
It is important to formulate the empirical observations and the hypotheses that
account for them as precisely and explicitly as possible. The importance of precise
and explicit formulation is illustrated in a discussion of the relation between a
formal property of English – subject-auxiliary inversion – and an interpretative
property – interrogative force. It has become clear from the data that we cannot
strongly correlate SAI and interrogative force: not all examples displaying SAI are
interrogative and interrogative force does not always lead to SAI. Any formulation
of that correlation must refer to potential rather than general links between form
and meaning.
One empirical generalization about English that does emerge from the discussion
was that verbs do not invert with their subjects in English. In the ﬁnal part of the
chapter, we saw that this generalization cannot be stated as a universal linguistic
ban on subject verb inversion and that in many languages, including earlier stages
of English, verbs do indeed invert with their subjects.
In the discussion of the relation between form and meaning we have underlined
the importance of structure, that is, the way that words are combined to form units
and sentences. Our discussion of inversion patterns implies that we think of sentences
as containing particular positions or slots into which elements are inserted
and that certain types of units belong to certain types of positions. In the next
chapters we will explore these underlying assumptions about sentences structure
and we develop an explicit and systematic theory of the structure of sentences.
TSC01 07/18/2005, 03:11PM46
Introduction: The Scientiﬁc Study of Language 47Exercises 47
Exercises
Contents
Introductory note to the exercises 47
Exercise 1 Exploiting multiple meaning (T) 48
Exercise 2 Ambiguity (T) 49
Exercise 3 Time speciﬁcations and their interpretations (T) 50
Exercise 4 Auxiliary (T) 51
Exercise 5 Copula be (T) 52
Exercise 6 Copula be and other linking verbs (T) 52
Exercise 7 “Emphatic do” (T) 52
Exercise 8 Contextually related ellipsis (L) 54
Exercise 9 Substitutes for units containing a verb (L) 55
Exercise 10 Substitution by so (L) 57
Exercise 11 Classifying examples: Locative inversion (E) 57
Exercise 12 Classifying examples: Predicate inversion (E, L) 61
Introductory note to the exercises
The exercises in this book will be accompanied by the abbreviations (T), (L), and
(E). The abbreviation (T) stands for “tie in,” and it serves to signal that a particular
exercise ties in with the material in the preceding chapter. Such tie-in exercises are
signaled by footnotes in the chapter. The abbreviation (L) stands for “look ahead”
and it signals that the material covered in the exercise will be taken up in a later
chapter. The abbreviation (E) stands for “expansion” and it signals that the material
covered in the exercise goes beyond that covered in the book. Since the material
contained in them has been covered, T-exercises will tend to be “easier” than
L-exercises or E-exercises. Sometimes exercises will combine tie-in elements with
new material that is to be treated in a later chapter, in which case we will label
them as (T, L). Alternatively they may partly be exercises of the material in the
chapter to which new material is added. Such exercises are labeled (T, E).
Some of the E-exercises will include longer discussion of particular points. Exercises
11 and 12 of this chapter are examples. The reason why the discussion in these
exercises is not included in the main body of the text is that the exercises are intended
only as illustrations of how research topics can be pursued in linguistics. Using the
argumentation developed in the text, the discussions show how particular issues
can be picked up and developed further. These “discursive exercises” typically will
TSC01 07/18/2005, 03:11PM47
48 Chapter 1
not offer an exhaustive or deﬁnitive treatment of the issues in question. Rather, they
illustrate how an analysis can be called into question and may have to be reworked in
the light of new data or of new theoretical proposals. Recall that scientiﬁc theories are
not static. A scientiﬁc theory is not a painting that you can admire but that essentially
is “dead” in that nothing in it can alter. Science is a search, it is active and alive.
In any branch of science there are only two possibilities. There is either nothing left to
discover, in which case why work on it, or there are big discoveries yet to be made, in
which case, what the scientists say now is likely to be false. (Nigel Calder, author of
Magic Universe: The Oxford Guide to Modern Science. Cited in the Guardian, 3.6.2004,
p. 6, col. 2)
Exercise 1 Exploiting multiple meaning (T)
(1) The following is a letter to the editor published in The Guardian (13.9.1997,
p. 8). Discuss the interpretation of the extract.
Branching out
I was amazed to read in your article about Marc Bolan (The King and I,
September 11) that “. . . a headstone is to be erected by the tree that killed
him.” I presume you will be giving full coverage to this example of arboreal
largesse and perhaps you will even print extracts of the tree’s speech?
Francis Quinn, 52c James Street, Cookstown, Co. Tyrone, BT80 8LT.
Discuss how the potential for multiple interpretation is exploited in the following extracts.
(2) A couple of hunters are out in the woods when one of them falls to the
ground. He doesn’t seem to be breathing. The other whips out his mobile
phone and calls the emergency services. He gasps out to the operator: “My
friend is dead! What can I do?” The operator, in a calm soothing voice says:
“Just take it easy. First, let’s make sure he’s dead.” There is silence, then a
shot is heard. The guy’s voice comes back on the line. He says “OK, now
what?” (Guardian, G2, 20.12.2001, p. 4, col. 6; “The world’s funniest joke?”)
(3) A man spots a farmer standing in a ﬁeld in the rain. “Why?” he asks. The
farmer replies: “I am trying to win a Nobel Prize. You get one for being out
standing in your ﬁeld.” (Observer, 10.10.2004, p. 19, col. 5)
(4) My most vivid memory of him and us students was waiting for a train at
Victoria and telling jokes in the manner of “Will the people who took the train
TSC01 07/18/2005, 03:11PM48
Introduction: The Scientiﬁc Study of Language 49
on platform seven please bring it back?” (Guardian, 22.11.2002, p. 9, col. 6,
letter to the editor from Edward Lynton)
(5) “George Best was a fantastic football player and he would have been even
better if he’d been able to pass night-clubs the way he passed the ball,”
Docherty said of the errant star. (Based on Guardian, 31.7.2002, p. 3, col. 4)
(6) Mr Howard said that under Labour a teacher is assaulted every seven minutes
(as in the old joke, “and he’s getting pretty damned sick of it”). (Guardian,
1.7.2004, p. 2, col. 7)
(7) a I have always found the advice on medicine bottles to “keep out of reach
of children” to be advice well worth following. (Guardian, 18.2.2002, p. 13,
col. 4, letter to the editor from David Carter)
b Another unintended message: Marks and Sparks’ advice to its customers, as
printed on all their plastic shopping bags, is: “To avoid suffocation, keep
away from children.” (Guardian, 18.2.2002, p. 13, col. 4, letter to the editor
from Dick Brown)
Exercise 2 Ambiguity (T)
The following passages are extracts from published written material. In each of the
extracts some segment gives rise to more than one interpretation and could potentially
lead to misunderstandings. Sometimes the ambiguity is highlighted in the passage
because it is exploited by the author, but in most cases the ambiguity may well go
unnoticed because the context of the extract will privilege one reading rather than
the other. Discuss the ambiguities that arise in the examples and identify the linguistic
elements that give rise to the multiple interpretations. After you have dealt with the
examples one by one try to classify them in terms of the cause of the ambiguity.
(1) Jackie Child’s youngest daughter was just two when she was jailed for manslaughter
nine years ago. (Guardian, G2, 27.7.2001, p. 10, col. 1)
(2) “Have a fag.”
“You’re making me into a smoker, Mrs Anthony. Thanks, I will. But you
should try to cut them down, they aren’t too good for you.”
“Twenty a day since I was twenty-ﬁve and seventy yesterday,” said Mrs
Anthony.
“Seventy! Gracious, you’ll be . . .”
“Seventy years of age yesterday.”
“Oh, seventy. Isn’t it time you had a rest then?”
(Muriel Spark, Memento Mori, 1977: 54)
Exercises 49
TSC01 07/18/2005, 03:11PM49
50 Chapter 1
(3) If you feel threatened in a taxi, ﬁrmly ask the driver to stop and get out.
(based on Guardian, G2, 7.3.2003, p. 7, col. 2)
(4) Perhaps they hadn’t intended to kill, only confront him jointly, threaten and
shock. But the French cook’s knife had been handy, lying on the table maybe.
(Ruth Rendell, An Unkindness of Ravens, 1994: 213)
(5) “I can’t get used to wearing my engagement ring yet. The other day I even
scratched my nose with it because it’s so big – the ring I mean.” (based on
“Diana, a tribute.” Sunday Times Supplement, Style, 7.9.1997, p. 11)
(6) In the survey, 200 couples were asked to keep reading diaries for three weeks.
(Guardian, 27.5.2002, p. 8, col. 8)
(7) What funny story about your life do you tell your grandkids (if you’ve got
any)? Do you mean, if I’ve got any funny stories, or any grandchildren? As it
happens I have three grandchildren and no funny stories. (Independent, interview
Gore Vidal, 11.8.1999, p. 7, col. 6)
(8) We need more robust measures. (Headline, Guardian, 29.11.2003, p. 20)
(9) Error lets bad meat trader off the hook. (Headline, Guardian, 24.5.2004,
p. 6, col. 7)
Comment on the interpretation of the underlined sections in the following passages:
(10) a Since 2003 individual drivers have also been subject to a licensing regime
and it is a source of some satisfaction to drivers of black cabs that many
mini cab drivers and ﬁrms have struggled to comply. “The mini cabs are in
dire straits,” claimed Mr. Oddy. “In reality people need moving around
London and mini cabs don’t to the same sort of work black cab drivers do.”
(Guardian, 3.9.2004, p. 14, col. 6)
b Rhys Jones lived in the penthouse and a swimming pool was built in
the basement. “Black-cab drivers ask me if he still lives there,” says Ellis.
(Observer Magazine, 25.7.2004, p. 38, col. 3)
Discuss the orthography in (10b).
Exercise 3 Time speciﬁcations and their
interpretations (T)
Discuss the interpretation of the underlined time speciﬁcations in the following
examples.
TSC01 07/18/2005, 03:11PM50
Introduction: The Scientiﬁc Study of Language 51
(1) Tony Blair admitted that he had run into “tough times” in recent months
yesterday. (Independent, 5.9.2003, p. 2, col. 1)
(2) Mr Straw decided to appoint a panel of independent doctors to examine
General Pinochet on January 5. (Guardian, 13.1.2000, p. 1, col. 3)1
(3) We found out that he had been given a scholarship to the RAF just after the
accident, which was an awful timing. (Guardian, 19.5.2001, p. 4, col. 8)
(4) Men who use internet chatrooms to “groom” young girls for sex were warned
that they face long jail terms yesterday. (Guardian, 16.10.2003, p. 5, col. 1)
(5) George Carey, the Archbishop of Canterbury, formally made the long anticipated
announcement of his plan to retire from ofﬁce in the autumn yesterday.
(Guardian, 9.1.2002, p. 3, col. 1)
Exercise 4 Auxiliary (T)
In section 2.2.1 of the chapter we roughly deﬁned auxiliaries as follows:
A provisional (and very approximate) characterization of auxiliaries, to be reﬁned
in Chapter 3, is that they are elements that are typically followed by a verb.
Discuss the appropriateness of the deﬁnition on the basis of the following
examples:
(1) James is deﬁnitely writing another novel.
(2) Has James already ﬁnished his novel?
(3) He hasn’t talked to his publisher yet but he will soon.
(4) I promised I would get you a present and get you a present, I will.
(5) The baby is asleep in its cot.2
1
For discussion of this example see also Chapter 2, section 1.1.
2
See also Exercise 5.
Exercises 51
TSC01 07/18/2005, 03:12PM51
52 Chapter 1
Exercise 5 Copula be (T)
On the basis of (1) show that be is an auxiliary when followed by the progressive
participle:
(1) The baby is sleeping in its cot.
Consider now (2):
(2) The baby is asleep.
In this example be relates a subject, the baby and an adjective asleep. The adjective
expresses a property of the subject and this use of the adjective is often referred
to as predicative. Copula be links a subject and a predicate. In (2) copula be is not
followed by another verb. Examine the morphological and distributional properties
of be when used as a copula. Does copula be behave as a lexical verb or as
an auxiliary? Provide arguments for your answer.
Exercise 6 Copula be and other linking verbs (T)
Consider the examples below.
(1) Mary was very tense.
(2) Mary seemed very tense.
(3) Mary remained very tense.
(4) Mary became very tense.
(5) Mary looked very tense.
In each of these examples we basically ascribe a property ‘very tense’ to the referent
of the name Mary. The link between the property and the subject is established by
means of the words was, seemed, remained, became, looked. What is the category
of these words? Are they auxiliaries or full verbs? Motivate your answer.
Exercise 7 “Emphatic do” (T)
Consider the underlined occurrences of do in the following examples.3
Can we
eliminate do and preserve a grammatical sentence? It turns out that if we eliminate
3
See also Chapter 3, section 1.2.3.2 for discussion.
TSC01 07/18/2005, 03:12PM52
Introduction: The Scientiﬁc Study of Language 53
do we must attach the inﬂection that is associated with do to the verb itself and the
resulting sentences will be acceptable. Looking at the contexts in which the sentences
with do are used, try to identify a common contextual factor that relates all these
examples.
(1) I don’t remember much of anything she said in the church foyer or what I
uttered back. She had that dazzling effect on me. Truth is, she still does.
What I do recall is that she invited me to a holiday party two nights later at
the mutual friend’s place. (Chicago Tribune, 22.12.2003, section 13, p. 9,
col. 1)
(2) Coleman, who describes himself as a “semi-professional punter”, gave evidence
at a trial in Southampton in October 2001 and his statements to the
court then will form the basis of the case against him. It is still not clear if he
will turn up for the 10 a.m. hearing at the club’s headquarters in London, but
the feeling at Portman Square yesterday was that he would indeed appear
to defend himself . . . If Coleman does appear this morning, the Jockey Club
may also wish to inquire about another part of the evidence. (Guardian,
22.1.2003, p. 14, cols 1 and 2)
(3) On Tuesday Clarett disputed the contention of university ofﬁcials that he had
failed to ﬁle the proper paperwork that would have allowed him to attend
the funeral . . . Each side is right, Clarett did ﬁll out the papers but ﬁlled them
out too late to receive tickets to ﬂy home. (New York Times, 2.1.2003,
p. D1, col. 1)
(4) Jackson is hardly a virgin forest. Like most of the state’s redwood land, it has
been logged intermittently since about the middle of the 19th century . . . But
the forest does have thousands of acres of 80- to 100-year old redwoods.
(San Francisco Chronicle, 28.11.2002, p. A34, col. 1)
(5) I’m probably more benevolent towards Mr Livingstone than a lot of
people and I actually do think he’s very brave in trying congestion charging.
(Guardian, 3.1.2003, p. 3, col. 4)
(6) People close to Senate leader Tom Daschle say he should be considered a
possible candidate, but many Democrats say they would be surprised if he
does run. (Atlanta Journal-Constitution, 1.12.2002, p. A6, col. 5)
(7) But that’s the trouble with middle-aged men these days: they’re so busy
trying to convince the world that they really do like Eminem that they have
forgotten several decades of their past. (Los Angeles Times, 26.11.2002,
p. E13, col. 3)
Exercises 53
TSC01 07/18/2005, 03:12PM53
54 Chapter 1
(8) In the new report, mice that were fed only every other day – but could gorge
on the days they did eat – saw similar health beneﬁts to ones that had their
diet reduced by 40 percent. (Washington Post, 29.4.2003, p. A3, col. 5)
(9) We were told journalism is a science. It didn’t make sense then nor does it
now. But it does make sense that we were learning a profession. (adapted
from Washington Post, 29.4.2003, p. A22, col. 4)
(10) When it was ﬁrst established in 1900, the Nobel committee clearly thought
[the Peace Prize] should be awarded to people who really did believe in
peaceful solutions and non-violence. (Guardian, 7.12.2002, p. 10, col. 1)
Exercise 8 Contextually related ellipsis (L)
As shown by examples (2) and (7) in Exercise 2, material from a sentence may
sometimes be omitted or deleted. The omitted material can usually be recovered
from the context. In the following examples locate all instances of ellipsis. Identify
which material has been omitted. Indicate the ellipsis by means of the symbol ∅
and consider which element immediately precedes it.4
Example
I asked him to write the report but he wouldn’t.
• I asked him to write the report but he wouldn’t ∅.
• Omitted material: write the report.
∅ is preceded by the contraction of the auxiliary would + negation.
(1) He wants to be the boss. In Silver Spring yesterday, he was. (Guardian,
23.10.2002, p. 1, col. 5)
(2) When he ﬁrst ran for ofﬁce four years ago, Gov. Gray Davis vowed to save
California’s old-growth forests. He hasn’t, as Moloney sees it, and she wants
him to live up to that long-ago campaign promise. (Los Angeles Times,
26.11.2002, p. B7, col. 2)
(3) A lot of prisoners lie and say they are sorry about something when they are
not. (Guardian, 17.1.2003, p. 1, col. 30)
4
We will be looking at the relevance of ellipsis for determining structure in Chapter 2,
section 1.6.
TSC01 07/18/2005, 03:12PM54
Introduction: The Scientiﬁc Study of Language 55
(4) This study compared complications in 552 ARF patients in the intensive care
unit at four academic hospitals, 326 of whom received diuretics on a particular
day and 226 who did not. (Los Angeles Times, 26.11.2002, p. F7, col. 1)
(5) Whitelaw had given his word to be loyal to her, and he was. (Guardian, G2,
11.9.2002, p. 4, col. 4)
(6) Sometimes I feel like I would like to crawl away and hide. But I will not.
(Guardian, 11.12.2002, p. 1, col. 2)
(7) I have never been to Australia, but a friend who has assures me that Moody
is quite correct. (Guardian, Sport, 14.12.2002, p. 4, col. 3)
(8) After all, Francesca’s hardly news any more. We are all trying to forget her.
As if we could. Although we should. I can’t. (Francis Fyﬁeld, Undercurrents,
2001: 50)
(9) We’re also keen to have a meeting with all parties and ﬁnd out what’s
gone wrong, because it’s obvious something has. (Guardian, 13.12.2002,
p. 15, col. 4)
(10) Under government policy, Cubans who make it to shore are generally allowed
to stay, while those who do not are sent back to their homeland. (New York
Times, 28.11.2002, p. A26, col. 2)
Exercise 9 Substitutes for units containing a verb (L)
In the previous exercise we saw that to avoid repetition, a string of words is
sometimes omitted.5
In examples (2), (4), (6 –10), the omitted strings contained a
verb. For instance: (8) has 3 ellipsis sites:
(1) a We are all trying to forget her. As if we could ∅. Although we should ∅. I
can’t ∅.
The symbol ∅ stands for forget her:
(1) b We are all trying to forget her. As if we could forget her. Although we should
forget her. I can’t forget her.
In each case, the ellipsis site is preceded by an auxiliary: in (i) the relevant auxiliaries
are could, should, can’t.
Exercises 55
5
For the role of substitution in syntactic analysis see Chapter 2, section 1.3.
TSC01 07/18/2005, 03:12PM55
56 Chapter 1
In the examples below the repetition of a string of words is avoided not by
omitting it but by substituting one word or a short string of words. The substitutes
are underlined in the examples. Identify which strings have been replaced by the
underlined words:
(2) When he was named chief by Mayor James K. Hahn, Bratton told The Times
that he wanted to establish close ties with prominent leaders in the city’s
minority communities. In doing so, he said, he would be better able to keep
local leaders informed of police action and reduce the likelihood of communities
“exploding in anger.” (Los Angeles Times, 26.11.2002, p. B10, col. 5)
(3) Is there anything that can prevent Hurricanes? To date, science and technology
have not given us the ability to do so. (Chicago Tribune, 3.1.2004,
section 1, p. 28, col. 1)
(4) Can stout shoes save you during a nuclear attack? They might do, providing
you shake the radioactive dust from them before going inside. (Guardian,
G2, 1.4.2004, p. 15, col. 4)
(5) I believe that if I were to continue to play for Zimbabwe I would do so only
by neglecting the voice of my conscience. (Guardian, Sports, 17.3.2003,
p. 6, col. 2)
(6) Your leader (Local voters must use their power, April 30) is right. So they
must, but why then stop at suggesting the proportional representation that
alone can make it worthwhile to vote? (Guardian, 5.5.2003, p. 19, col. 4,
letter to the editor from Prof. George Hutchinson, Southampton)
(7) During the appearances, Bratton rejected the idea of ﬂooding South Los
Angeles streets with ofﬁcers. Doing so would probably raise the ire of a
community with a long history of confrontation with police rather than solve
any problems. (Los Angeles Times, 26.11.2002, p. B1 + 10, col. 1)
(8) The national primary care research and development centres at Manchester
and York universities, which carried out both this study and 1998’s, acknowledge
that not all the doctors who say they want to leave will. But previous
research has shown that many will do so. (Guardian, 3.1.2003, p. 1, col. 1)
(9) We save life ﬁrst and we do the rest if we possibly can. The priority is to save
life. If we can put the ﬁre out, we will do so. (Guardian, 14.11.2002, p. 1,
col. 4)
(10) I’m sure that neither of them could have murdered Brooks. It’s a physical
impossibility, knowing about dates and times. But she could have done. Ellie
TSC01 07/18/2005, 03:12PM56
Introduction: The Scientiﬁc Study of Language 57
Smith could have done – if only just. (Colin Dexter, The Daughters of Cain,
1995: 330)
(11) “What was your accent like?” “Southern Welsh; You can still hear the trace
of it, mind you.” “So you can,” said Isobel. (Muriel Spark, The Bachelors,
1963: 101)
Exercise 10 Substitution by so (L)
In Exercise 9 we have seen that strings of words containing a verb may be replaced
by do (4), (10), and by do so (2), (3), (5), (7), (8), (9).6
Do and do so in fact always
replace a string containing a verb. In examples (6) and (11) of Exercise 9, a string
of words containing a verb is replaced by so. The relevant part of (11) is repeated
below:
(i) “You can still hear the trace of it.” “So you can,” said Isobel.
So = still hear the trace of it.
Hear is a verb.
Examine the examples below. Could we generalize the pattern illustrated in (i) and
say that so always substitutes for a string of words containing a verb?
(1) We have counselled against the war, but once it’s a reality the story moves on
and so will we. (Guardian, G2, 17.3.2003, p. 7, col. 4)
(2) The towns [Paula Radcliffe] has lived in (Nantwich, Bedford and Loughborough)
are the epitome of Middle England. And so is she. (Guardian,
17.12.2002, p. 13, col. 1)
(3) Willie Whitelaw on a meeting with the West German interior minister: “He’s
very keen on terrorism. So am I.” (Guardian, G2, 11.9.2002, p. 5, col. 4)
Exercise 11 Classifying examples: Locative
inversion (E)
In this exercise we return to the classiﬁcation of inversion patterns. The exercise is
longer and more discursive than the preceding exercises. It probably is also slightly
more demanding. The goal is to carry further the kind of investigations started
6
For the role of substitution in syntactic analysis see Chapter 2, section 1.3.
Exercises 57
TSC01 07/18/2005, 03:12PM57
58 Chapter 1
in the chapter and see where that would lead us. In this particular exercise we
will discover that English has more than one type of inversion. Here we look at
patterns in which a verb appears to the left of the subject. Exercise 12 ties in with
Exercise 11.
Recall that we discussed the derivation of word order patterns in which the
subject of a sentence is preceded by an auxiliary. Let’s start from (1a) below.
Identify the subject and replace it by a pronoun. Does the example contain any
auxiliaries? Is the auxiliary immediately followed by a full verb? Using SAI, form a
direct question based on (1a).7
(1) a This startling insight will naturally emerge from doing the syntax course.
The subject of the sentence is this startling insight. The sentence refers to a future
event, futurity being signaled by the auxiliary will, which is followed by the inﬁnitive
of the verb. The various modiﬁcations suggested above are illustrated in the
following sentences:
(1) b It will naturally emerge from doing the syntax course.
c Will this startling insight naturally emerge from doing the syntax course?
Observe that when we apply SAI, we only move the auxiliary will in front of the
subject; we cannot also move the verb emerge in front of the subject, regardless of
whether we take the adverb naturally along:
(1) d *Will emerge this startling insight naturally from doing the syntax course?
e *Will naturally emerge this startling insight from doing the syntax course?
Observe ﬁnally that we can also apply SAI to (1b), whose subject is a pronoun:
(1) f Will it naturally emerge from doing the syntax course?
Now consider (2a). What is the subject? Is there an auxiliary? How would we form
a direct question?
(2) a This startling insight naturally emerges from doing the syntax course.
The subject of (2a) is again this startling insight, there is no auxiliary. We can
replace the subject by a pronoun. When we want to ask a question we apply SAI,
inserting the auxiliary do as a last resort.8
(2) b It naturally emerges from doing the syntax course.
c Does this startling insight naturally emerge from doing the syntax course?
7
See Chapter 1, section 2.3.2.
8
See Chapter 1, section 2.2.2.
TSC01 07/18/2005, 03:12PM58
Introduction: The Scientiﬁc Study of Language 59
Again, we can also apply SAI with do insertion to (2b), with a pronominal subject:
(2) d Does it naturally emerge from doing the syntax course?
Observe that in order to ask a question we cannot invert the verb with the subject:
(2) e *Emerges it naturally from doing the course?
This is because in English full verbs do not invert with subjects. This was expressed
in generalization (47) in the discussion section of the chapter, repeated here as (3):
(3) Verbs that are not auxiliaries do not invert with the subject.
Consider the position of the underlined verb in the following example in the light of
the generalization in (3):
(4) a From behind detail of courses and qualiﬁcations emerges the progressive conviction
that no one can ever learn enough. (Based on Guardian, 14.2.2002,
p. 9, col. 1)
In (4a) the verb emerges precedes the subject. We note that the subject itself is
long and complex (the progressive conviction that no one can ever learn enough).
At ﬁrst sight we might think that example (4a) constitutes counter-evidence to our
generalization in (3). If this were true we would have to re-examine the data and
somehow weaken our generalization.
However, closer examination of (4a) reveals that this example must be treated
as a separate pattern. This can be seen if we compare this sentence with those in
(1) and (2). Recall that in the cases of SAI illustrated above, a subject could be
replaced by a pronoun (1f), (2d). If we try to replace the subject by a pronoun in
(4a) the result is unacceptable:
(4) b *From behind detail of courses and qualiﬁcations emerges it.
To facilitate further comparison with the examples in (1), let us modify example
(4a) slightly. We do this by inserting the auxiliary will in (4a), thus locating the event
in the future. The resulting order is as in (4c) and not that in (4d):
(4) c From behind detail of courses and qualiﬁcations will emerge the progressive
conviction that no one can ever learn enough.
d *From behind detail of courses and qualiﬁcations will the progressive conviction
that no one can ever learn enough emerge.
We see that in this example the subject is preceded by both the auxiliary and the
verb. Recall that this order was not possible with respect to our typical examples of
SAI: (1d) and (1e) were unacceptable.
Exercises 59
TSC01 07/18/2005, 03:12PM59
60 Chapter 1
9
For a description of locative inversion see Emonds (1976: 34–7), Coopmans (1989), Hoekstra
and Mulder (1990), Bresnan (1994), Culicover and Levine (2001).
10
In Chapter 5, Exercise 15 we will see that in French inversion, postverbal pronouns pattern
differently from postverbal NPs. As this is the very last exercise of the book, this is not the
right moment to tackle it!
Table 1 Two inversion patterns
SAI Inversion type in (4a)
Can the subject be a pronoun? yes no
Will the order be auxiliary – subject – verb? yes no
Will the order be auxiliary – verb – subject? no yes
If you compare the inversion patterns in (1), (2), which we identify as instances
of SAI, and the variations associated with example (4), there are a number of
differences to note. In the routine examples of SAI as applied to (1a), the subject
can be a pronoun (1b), and only the auxiliary will precedes the subject (1c). In (4a),
the subject cannot be a pronoun (4b), the subject follows both the auxiliary and the
lexical verb (4c). We summarize the differences in Table 1.
The inversion pattern illustrated in (4a) has a restricted distribution. One typical
manifestation is that illustrated here, in which the ﬁrst component of the sentence
is a locative element, here from behind detail of courses and qualiﬁcations. The
pattern in (4a) is often referred to with the term locative inversion.9
In such
patterns, the subject has to be relatively heavy; as we have seen, a pronominal
subject is not possible.10
The subject is found in a position towards the end of the
sentence, where it is highlighted.
Identify the locative inversion patterns in the following examples. Compare their
properties with those summarized in Table 1. Can the subject be replaced by a
pronoun? For examples without any auxiliary, try inserting one (with appropriate
change of verb form): what is the resulting word order? For examples with an
auxiliary, comment on the relative position of auxiliary, subject, and verb.
(5) On the credit side of South Africa’s balance sheet goes the 8.4 million people
who now have access to clean water, 3.8 million with electricity, and 1.46
million who have new homes. (Guardian, 24.5.2003, p. 11, col. 2)
(6) Behind the celebrations and enthusiasm lies a project marked by controversy.
(Guardian, Life, 18.3.2004, p. 2, col. 1)
(7) From this has stemmed the bad manners and casual crime we see today.
(Independent, 20.8.2004, p. 12, col. 2)
TSC01 07/18/2005, 03:13PM60
Introduction: The Scientiﬁc Study of Language 61
(8) Through the door rushes his estranged brother, Turley, running for his life.
(Guardian, Review, 20.3.2004, p. 13, col. 5)
Exercise 12 Classifying examples: Predicate
inversion (E, L)
In this exercise we examine the classiﬁcation of inversion patterns. The exercise is
similar to Exercise 11 in that it is longer and more discursive than Exercises 1–10.
The idea is again that we carry further the kind of investigations started in the
chapter and see where that would lead us. In particular we will ﬁnd conﬁrmation
that English has more than one type of inversion, a point already shown in Exercise
11, and we will discover that not all patterns in which an inﬂected form of
be appears to the left of the subject are cases of SAI. This exercise ties in with
Exercise 11.
Recall that we have discussed the derivation of word order patterns in which the
subject of a sentence is preceded by an auxiliary. Let us start from (1a). Identify the
subject of the sentence and replace it with a pronoun. Does the example contain
any auxiliaries? Is the auxiliary followed by a full verb?
(1) a Cost is complicating matters.
The subject of (1a) is cost; we replace the subject by a pronoun in (1b):
(1) b It is complicating matters.
Apply SAI to the example in (1a, b) to form a direct question.11
What is the
resulting order? As you can see we have straightforward sequences of auxiliary
– subject:
(1) c Is cost complicating matters?
d Is it complicating matters?
When we apply SAI, we cannot also move the verb complicating in front of the
subject:
(1) d *Is complicating cost matters?
e *Is complicating it matters?
Let us now turn to (2a). How does it differ from (1a)?
(2) a Cost will be complicating matters.
11
See Chapter 1, section 2.3.2.
Exercises 61
TSC01 07/18/2005, 03:13PM61
62 Chapter 1
The difference between (1a) and (2a) is that the state of affairs expressed by the
latter sentence is situated in the future. This is achieved by means of the auxiliary
will. By adding will to (1a) we create a sentence with two auxiliaries: the modal
auxiliary will and the non-inﬂected form of be.12
Identify the subject and replace it
with a pronoun. The subject of the sentence is cost. We can replace the subject of
(2a) with a pronoun.
(2) b It will be complicating matters.
Apply SAI to the examples in (2a, b) to form a direct question.13
What is the
resulting order? As you can see we again have straightforward sequences of auxiliary
– subject:
(2) c Will cost be complicating matters?
d Will it be complicating matters?
Observe that when we apply SAI, we must move only one auxiliary: it is not
possible to move two auxiliaries across the subject (2e), nor is it possible to move
both auxiliaries as well as the full verb (2f):
(2) e *Will be cost complicating matters?
f *Will be complicating cost matters?
Let us formulate this in terms of a general principle:
(3) SAI moves only one auxiliary across the subject.
We can summarize the various patterns as in Table 2 below. Consider the examples in
(4): the underlined subject is preceded by the verb be. Is be an instantiation of copula
be or of auxiliary be? What arguments could you offer in support of your analysis?
(4) a Complicating matters is cost. (Washington Post, 10.12.2002, p. A16, col. 1)
b Helping to run the house were a cook, a housemaid and a manservant.
(Carol Shields, Jane Austin, 2001: 123)
Identify the participles associated with the auxiliaries. We ﬁnd that in the examples
the participle (complicating, helping) precedes its auxiliary. Restore the sentences to
a more neutral word order in which the subject precedes the auxiliary and the verb:
(5) a Cost is complicating matters.
b A cook, a housemaid and a manservant were helping to run the house.
12
We will discuss sentences containing more than one auxiliary in more detail in Chapter 3,
section 4.
13
See Chapter 1, section 2.3.2.
TSC01 07/18/2005, 03:13PM62
Introduction: The Scientiﬁc Study of Language 63
Table 2 SAI
Can the subject be a pronoun? yes
One auxiliary
Order will be auxiliary – subject – verb? yes
Order will be auxiliary – verb – subject? no
Two auxiliaries
Order will be auxiliary – subject – auxiliary – verb? yes
Order will be auxiliary – auxiliary – subject – verb? no
Order will be auxiliary – auxiliary – verb – subject? no
The initial element of the examples in (4) is a string of words containing the
participial form of the verb (complicating, helping). Given that they display the
order auxiliary – subject we might conclude that the examples in (4) are simply
examples of SAI.
However this conclusion would be rash. We ﬁrst examine (4a). Try replacing the
subject in (4a) by a pronoun. The resulting sentence is no longer acceptable.
(5) c *Complicating matters is it.
The examples in (4) and in (5) contain just one auxiliary be. To examine if they
really illustrate SAI we could also try to test our generalization in (3). What would
happen if there was a sequence of auxiliaries, as in our earlier examples in (2)?
Recall that principle (3) summarizes our ﬁnding that SAI moves only one auxiliary
to the left of the subject. Try adding the future maker will to (5a), replacing the
present tense form of be by will be.
(6) a Complicating matters will be cost.
Observe that if we insert a future marker will in example (4a), the resulting order is
that in which both will and be precede the subject cost. If we apply SAI as described in
(3) above, then only the auxiliary will should precede the subject. But in (6), if only
the auxiliary will precedes the subject while be follows it the sentence is unacceptable:
(6) b *Complicating matters will cost be.
If you return to the main properties of SAI as summarized in Table 2, you will see
that the inversion which is illustrated in (4) is quite different from SAI. We summarize
the differences in Table 3. In the discussion above, identify which examples provide
evidence for the various properties. Because of the differences between the two
patterns, we will not consider the examples in (4) as illustrations of SAI. This means
that the term SAI is restrictive: not every sequence in which an auxiliary precedes
Exercises 63
TSC01 07/18/2005, 03:13PM63
64 Chapter 1
Table 3 Two types of inversion
SAI Examples in (4)
Can the subject be a pronoun? yes no
Two auxiliaries
Order is auxiliary – subject – auxiliary? yes no
Order is auxiliary – auxiliary – subject? no yes
the subject is automatically an instantiation of SAI. If you examine the properties of
(4b) along the lines outlined above you will conclude that that example too is not
an instantiation of SAI.
The patterns illustrated in (4) are often referred to as predicate inversion. In
predicate inversion patterns the predicate, the string of words that would follow
be in the neutral order, precedes be, and the subject, which would precede be in
the neutral order, follows be. Examples (7)–(10) also illustrate predicate inversion.
Restore the sentences in (7)–(10) to the more neutral order, in which auxiliary be is
preceded by its subject:
(7) Hurting the industry’s ability to raise fares is the fact that the big airlines are
putting more seats back into the skies to battle the rapid expansion of the
budget airlines. (Wall Street Journal, 29.3.2004, p. A6, col. 6)
(8) Sitting next to her in the remote cabin was 71-year old Elisabeta Sigilyetova,
one of the last living speakers of a rare dialect of Khanty, a regional tongue
nearly overwhelmed by Russia’s Slavic majority. (Wall Street Journal,
26.3.2004, p. A1, col. 4)
(9) Enjoying the festivities at the Sterling Fire Department’s Patton Hall are,
clockwise from above, Robyn, Blocher’s dog, Winston, 5, Melanie Howard
and Cosmo, 2, and, from left, Scott Morrison, Blocher, Cinston, and Blocher’s
father, Bill Blocher. (Washington Post, 25.3.2004, “Loudoun”, p. 1, col. 1)
(10) Tucked away in the back of the booth of Haboldt & Company, a Paris dealer,
is perhaps the best old master drawing here. (New York Times, 8.3.2004, p.
E5, col. 6)
We return to the patterns illustrated in (4) and in (7)–(10) in Exercise 23 of
Chapter 4.14
14
For discussion see Birner (1992), Birner and Ward (1992), Emonds (1976: 34–43), den Dikken
and Næss (1993), Heycock and Kroch (1997).
TSC01 07/18/2005, 03:13PM64
Diagnostics for Syntactic Structure 65
2 Diagnostics for Syntactic Structure
Discussion
Contents
0 Introduction 67
1 Diagnostics for Structure 68
1.1 Structure and meaning 68
1.2 Intuitions about structure 72
1.3 Substitution 74
1.4 Movement 79
1.5 Question formation 82
1.6 Deletion/ellipsis 84
1.7 Focalizing a constituent 85
1.7.1 The cleft sentence 85
1.7.2 The pseudo-cleft sentence 87
1.8 Co-ordination 89
2 The Verb Phrase and the Sentence: A First Exploration 92
2.1 Starting point 92
2.2 The relation between the auxiliary and the VP 93
2.3 Layered structures 95
2.3.1 Complements vs. adjuncts in the VP 95
2.3.2 Direction of adjunction 97
2.3.3 OV orders? 99
2.3.4 Base positions and movement 100
2.4 Deductive approaches 101
2.4.1 Head and projection 101
2.4.2 Binary branching 102
2.5 A short note on binary branching and word structure 105
TSC02 07/18/2005, 03:13PM65
66 Chapter 2
3 Speciﬁers 107
3.1 Noun phrases 107
3.1.1 A binary branching structure for NPs 107
3.1.2 Speciﬁers in the NP 110
3.1.2.1 Speciﬁer–head agreement 110
3.1.2.2 Prenominal genitives 112
3.1.2.3 Questions about the structure of sentences 113
3.2 Adjuncts and NPs 117
3.3 Questions about the verb phrase 118
4 Summary 121
TSC02 07/18/2005, 03:13PM66
Diagnostics for Syntactic Structure 67
0 Introduction
In this chapter we start our systematic analysis of the structure of sentences. We will
ﬁrst elaborate some techniques for discovering sentence structure, so that, whenever
we formulate a hypothesis about a particular structure, we can test this hypothesis
using a well-deﬁned set of analytical tools. We will decompose English sentences
and demarcate their core constituents. The current chapter mainly concentrates on
the position of the verb in the structure of the sentence. In the next chapters, we also
turn our attention to the relation of the subject to the sentence and to the verb.
Once we have determined what the components of the sentence are, we can formulate
a hypothesis concerning the derivation of the sentence, that is, the way sentences
are built up from such components. This will be further worked out in Chapters 3,
4, and 5.
Throughout the discussion, two kinds of data will be used. On the one hand, we
use attested examples. Sometimes, we will manipulate these examples to clarify
structural relations. In order to test our hypotheses we will often also create our
own examples. When using attested data, we are acting like scientists doing ﬁeldwork.
We discover phenomena and we examine them in their natural settings.
When creating our own examples, we are like scientists who conduct experiments
in their laboratories.
On the basis of our observations and of our experiments, we will formulate a
number of general hypotheses about how sentences are internally organized and
how they are put together. These hypotheses will guide us in later sections.
The chapter is organized as follows: Section 1 develops a series of tests for
identifying strings of words inside a sentence as units or constituents. Section 2
focuses on constituents containing a verb and examines two competing hypotheses
about the composition of verbal constituents or, to introduce the technical term,
verb phrases. According to one hypothesis, the verb phrase contains the verb and
the auxiliaries of the sentence; according to the alternative hypothesis, the verb
phrase contains the verb, its complements, and its adjuncts. We will show that the
second hypothesis is both empirically and theoretically preferable.
While discussing the structure of sentences, we will elaborate the far-reaching
theoretical proposal that all syntactic constituents are the result of combining or
merging two constituents, and that constituents are hierarchically organized around
a head. The head ﬁrst combines with its complement. The resulting constituent then
combines with the speciﬁer. The need to postulate a speciﬁer as one essential part
of the constituent will initially be motivated on the basis of our analysis of the
structure of the noun phrase in section 3. Head, complement, and speciﬁer are the
three core components of the phrase. Phrases may be further augmented by means
of adjoined constituents. Section 4 summarizes the chapter.
TSC02 07/18/2005, 03:13PM67
68 Chapter 2
1 Diagnostics for Structure
1.1 Structure and meaning
The discussion of the ﬁrst chapter of this book was based on the hypothesis that
language is two-faced: it unites form and interpretation (“meaning”). This hypothesis
is trivially correct in the sense that words have a form and a meaning, but we
have also seen that the arrangement of words into a sentence in itself is meaningful
in that it also contributes to the interpretation of the sentence. Consider the underlined
string of words in (1). What kind of information (“meaning”) does it contribute
to the sentence? Could you replace the underlined words by just one word?
(1) Mr Straw decided to appoint a panel of independent doctors to examine
General Pinochet on January 5. (Guardian, 13.1.2000, p. 1, col. 3)1
The string of words on January 5 provides information about a date, or, in more
general terms, about ‘time’. The date given, January 5, refers to the time of some
action or event. Which event is this? In fact, upon reading our example carefully,
it seems that there are three possible ways of relating on January 5 to the sentence:
(i) on January 5 may denote the time of Mr Straw deciding, (ii) it may denote the
time of appointing, or (iii) it may denote the time of examining. (1) is three ways
ambiguous: one string of words has three interpretations. The ambiguity does not
reside in a lexical ambiguity of any one of the individual words in the string on
January 5. These words have a constant meaning in this example, whichever the
interpretation chosen. Let us look more closely at each of the three readings.
In the ﬁrst reading, Mr Straw decided to do something, and his decision was
taken on January 5. On January 5, what did Mr Straw decide to do? The answer to
this question is that he decided to appoint a panel of independent doctors to examine
General Pinochet. The string of words to appoint a panel of independent doctors
to examine General Pinochet is a unit that functions as the object of the verb
decided. The string of words on January 5 itself is not part of the answer to the
question as to what Mr Straw decided, it is not part of the object of decide:
(1) a Mr Straw decided
when? – [on January 5]
what? – [to appoint a panel of independent doctors to examine GP]
In the second reading of the sentence, Mr Straw decided to do something and this
activity would take place on January 5. To the question what Mr Straw decided
to do, the answer would now be “to appoint a panel of independent doctors to
1
Cf. sentence (2) in Exercise 3 of Chapter 1.
TSC02 07/18/2005, 03:13PM68
Diagnostics for Syntactic Structure 69
examine General Pinochet on January 5.” In this reading of (1), the string on
January 5 is part of the answer to the question of what Mr Straw decided, in other
words it is part of the object of decide and it speciﬁes the time of appointing. The
time of Mr Straw’s decision-making itself is now not speciﬁed. In this reading, the
object of decide is the string to appoint a panel of independent doctors to examine
General Pinochet on January 5. To the question when he will appoint the panel the
answer is that he will appoint them on January 5. To the question who Mr Straw
will appoint on January 5 the answer is: “a panel of independent doctors to examine
General Pinochet.” For this second interpretation, we can represent the relations
between the elements of the sentence informally as in (1b). In (1b) on January 5 is
part of the object of decided. Recall that in (1a) above, the string on January 5 was
not part of object of decided.
(1) b Mr Straw decided
what?
[to appoint a panel of independent doctors to examine GP on January 5]
to appoint
when? – [on January 5]
who? – [a panel of independent doctors to examine GP]
Finally, in the third reading, the string on January 5 speciﬁes the time of the examining.
We are not told when Mr Straw took his decision, nor are we told when the
appointment will take place, but we are told on which date the appointed panel will
examine General Pinochet. In this reading, Mr Straw again decided to do something.
To the question what Mr Straw decided to do, the answer would again be “to
appoint a panel of independent doctors to examine General Pinochet on January
5.” In the third reading of (1), the time of the decision-making is not speciﬁed. The
direct object of decide is the string to appoint a panel of independent doctors to
examine General Pinochet on January 5. The string on January 5 is part of the
object of decide, but in the third interpretation it does not specify the time of
appointing. On January 5 speciﬁes on which date the panel will examine General
Pinochet. The answer to the question who Mr Straw will appoint is “a panel of
independent doctors to examine General Pinochet on January 5.”
(1) c Mr Straw decided
what?
[to appoint a panel of independent doctors to examine General Pinochet on
January 5]
to appoint
who? – [a panel of independent doctors to examine General Pinochet on
January 5]
to examine
who? – [General Pinochet]
when? – [on January 5]
TSC02 07/18/2005, 03:13PM69
70 Chapter 2
2
See Sperber and Wilson (1986) for an account of this type of disambiguation.
3
This sketch is provisional. We return to the various steps in this chapter and in Chapters 3
and 4.
(1) thus has three interpretations which arise from the three different relations that
the string on January 5 can have with the remainder of the sentence. As mentioned,
the string on January 5 itself does not change its meaning in the three interpretations.
On January 5 denotes a temporal speciﬁcation, the ﬁfth day of the month of
January. What changes is the way this temporal speciﬁcation is integrated into the
sentence. In (1c) on January 5 speciﬁes the timing of examine General Pinochet; in
(1b) it speciﬁes the timing of appoint a panel of independent doctors to examine
General Pinochet; in (1a) on January 5 speciﬁes the timing of decided to appoint a
panel of independent doctors to examine General Pinochet. The different meanings
come about by the way the unit on January 5 is hooked onto the sentence; in other
words, the different meanings come about by the various ways by which the sentence
can be assembled. The three readings of on January 5 are due to the structural
relations in the sentence, its syntax. Ambiguities which arise through different structural
relations are structural ambiguities.
The ambiguity that arises in (1) is not an exceptional phenomenon. This type of
structural ambiguity is fairly frequent in actual usage, even though it rarely leads
to problems of communication. In a particular communicative setting, the reader/
hearer of ambiguous sentences will be able to pick out the appropriate reading
easily.2
Sometimes, though, a writer/speaker may deliberately exploit the potential
for ambiguity created by the syntax. The following extract illustrates this point:
(2) I went to the National Gallery today, but it brought back painful memories of
B., so I went back to Soho and paid two pounds to watch a fat girl with spots
remove her bra and knickers through a peephole. I watched her through a
peephole. She didn’t remove her underclothes through a peephole. Query: are
there night classes in syntax? (Sue Townsend, Adrian Mole: The Wilderness
Years, 1993: 248–9.)
Let us look at (1) once again. Our discussion of this example implies that a sentence
is not put together at one go but that it is assembled step by step from smaller units.
The different readings of (1) can directly be related to the way the sentence is
assembled. In particular, we can relate the ambiguity of (1) to the timing of hooking
the unit on January 5 onto a particular part of the sentence. In (1c), the string on
January 5 belongs with to examine General Pinochet. We could say that when
assembling the sentence, the string on January 5 is hooked onto the string to examine
General Pinochet, creating a unit to examine General Pinochet on January 5.
The resulting unit is then hooked up to a panel of independent doctors. In turn,
the resulting unit a panel of independent doctors to examine General Pinochet on
January 5 is hooked onto appoint, and ﬁnally the result is itself hooked up to the
verb decided and its subject Mr Straw.3
(3a) schematizes the steps of the assembly
process to create the reading in which on January 5 modiﬁes examine.
TSC02 07/18/2005, 03:13PM70
Diagnostics for Syntactic Structure 71
(3) (i) to examine General Pinochet + [unit1 on January 5]
⇒[unit2 to examine General Pinochet [unit1 on January 5]]
(ii) a panel of independent doctors + unit2
⇒[unit3 a panel of independent doctors [unit2 to examine General Pinochet
[unit1 on January 5]]]
(iii) appoint + unit3
⇒ [unit4 appoint [unit3 a panel of independent doctors [unit2 to examine
General Pinochet [unit1 on January 5]]]]
(iv) Mr Straw decided + unit4
⇒[unit5 Mr Straw decided [unit4 to appoint [unit3 a panel of independent
doctors [unit2 to examine General Pinochet [unit1 on January 5]]]]]
In order to achieve the reading in (1a), we assemble the sentence rather differently.
When we link the string to examine General Pinochet with the string a panel of
independent doctors we do not yet integrate on January 5. The temporal speciﬁcation
only comes in later, when we are putting together decide with the remainder of
the sentence.
(4) (i) a panel of independent doctors + [unit1 to examine General Pinochet]
⇒[unit2 a panel of independent doctors [unit1 to examine Greneral Pinochet]]
(ii) appoint + unit2
⇒ [unit3 appoint [unit2 a panel of independent doctors [unit1 to examine
General Pinochet]]]
(iii) Mr Straw decided + unit3
⇒[unit4 Mr Straw decided [unit3 to appoint [unit2 a panel of independent
doctors [unit1 to examine General Pinochet]]]]
(iv) unit4 + [unit5 on January 5]
⇒[[unit4 Mr Straw decided [unit3 to appoint [unit2 a panel of independent
doctors [unit1 to examine General Pinochet]]]] [unit5 on January 5]]
In representation (3), the unit on January 5 is deeply integrated into the sentence; it is
combined early on (in step (i)) with the verb examine. In representation (4), the same
unit is added at the ﬁnal stage of the construction of the sentence (in step (iv)). The
brackets used in the schematic representations above reﬂect the level of integration:
in (3) on January 5 is followed by 5 right-hand brackets; in (4) by only 2 such brackets.
The displays in (3) and in (4) are imprecise. For one thing using labels such as
unit1, unit2 suggests that all these entities are similar in nature, though they make
different contributions to the interpretation of the sentence. Also the representations
are very difﬁcult to read. They are a complex ways of showing the history of how the
sentence is put together and how the interpretations are arrived at. In the remainder of
this chapter we will elaborate a more precise and transparent way for representing the
structure of sentences and we will also provide tools to determine the structural units.
TSC02 07/18/2005, 03:13PM71
72 Chapter 2
1.2 Intuitions about structure
In section 1.1 we talked about structure in a fairly intuitive and loose way. We
appealed to our linguistic awareness as speakers of English to informally represent
some of the structural units that build the sentence with the different interpretations
associated to the sentence. We indicated these units by bracketing, [ . . . ]. Units of
form, i.e. sequences of words, such as on January 5, are taken to correspond to
units of meaning, the string on January 5 is a time speciﬁcation.
Consider a sentence such as (5a). Going by your intuitions as to who does what
and when, how would you identify the major meaningful units in this sentence?
Represent each unit by using square brackets ([unit . . . ]).
(5) a The customer in the corner will order the drinks before the meal.
Probably, you will have bracketed the string as in (5b):
(5) b [unit The customer in the corner] will order [unit the drinks] [unit before the
meal].
Square brackets will from now on be used to demarcate units of structure. We don’t
have to label each set of brackets as “unit”: the very presence of the brackets
means that the string of words contained in the brackets is a unit.
(5) c [The customer in the corner] will order [the drinks] [before the meal].
In (5c) the brackets identify three units or constituents: (i) the customer in the corner,
(ii) the drinks, (iii) before the meal. There is no indication as to how the auxiliary
will and the verb order are integrated into the sentence. In Chapter 1, section 2.2.1,
we saw that auxiliaries tend to associate with a verb. We might propose that the
auxiliary will in (5) forms a unit with the verb order. On this assumption, we could
formulate the hypothesis that the assembly of the sentence is as in (5d).
(5) d Hypothesis A
[The customer in the corner] [will order] [the drinks] [before the meal].
But others might say that the bracketing in (5d) is counter-intuitive because the verb
order should ﬁrst be assembled with the string the drinks, which is the direct object
of the verb and which refers to the entity affected directly by the action expressed
by the verb. If you use the verb order you expect to ﬁnd a direct object: ‘order
what?’ ‘Ordering’ is an activity that implies there will be some entity being ordered.
This relation between order and the drinks is independent of the presence of the
auxiliary will: we can use the string order the drinks also in the absence of an
auxiliary such as will:
(5) a′ The customer in the corner orders the drinks before the meal.
TSC02 07/18/2005, 03:14PM72
Diagnostics for Syntactic Structure 73
The close relationship between order and the drinks is not revealed in the bracketing
in (5d). How could we represent this relationship between the verb and its
object? The bracketing in (5e) is meant to show that the verb order and the unit
the drinks are ﬁrst assembled to form a unit.
(5) e Hypothesis B
[The customer in the corner] [will] [order [the drinks]] [before the meal].
Let us compare (5d) and (5e) and try to make more explicit the claims that the
two alternative bracketings make. Having clariﬁed this issue, we will examine the
predictions of these claims more carefully.
In the diagrams below we use another format to represent how sentences are built
up. This format is called the tree diagram format: like bracketing, it schematizes
how sentences are formed from smaller units. The tree diagrams below correspond
to the bracketed representations above, but they allow us to read off more easily
how the various parts of the sentence are put together.
According to one representation, (5d)/(5d′), when assembling the sentence, we
ﬁrst assemble the auxiliary will and the verb order, and then we combine the resulting
unit with the other constituents: the subject the customer in the corner, the
direct object the drinks, and the time speciﬁcation before the meal. In the tree, a
triangle associated with a constituent is a device to show that the internal make-up
of the unit in question is not relevant to the current discussion.
(5) d′ Sentence
subject auxiliary verb object time speciﬁcation
The customer
in the corner before the mealwill order the drinks
According to the alternative in (5e), the assembly proceeds differently. Here, the
verb order is ﬁrst assembled with its object, the drinks. Then we combine the
resulting unit with the auxiliary will, with the time speciﬁcation before the meal,
and with the subject the customer in the corner.
(5) e′ Sentence
subject auxiliary
verb
order
object
the drinks
time speciﬁcation
The customer
in the corner before the mealwill
TSC02 07/18/2005, 03:14PM73
74 Chapter 2
These structural representations have different implications. For instance, starting
from the root of the structure “Sentence,” to reach the object the drinks in (5d′) you
go directly from the root to the unit the drinks. This means that, starting from
Sentence, the subject the customer in the corner and the object the drinks are both
equally accessible. On the other hand, to reach the object the drinks in (5e′), we ﬁrst
have to enter into the unit order the drinks. In (5e′) the subject the customer in the
corner is more readily accessible from the root “Sentence” than the object. From
Sentence it takes one single step to reach the subject; it takes two steps to reach the
object. The subject is presented as an immediate constituent of the sentence, the
object is presented as an ultimate constituent of the sentence: it is a component of
the sentence by virtue of being a component of the unit order the drinks, which is
itself an immediate component of the sentence. (5e′) introduces a subject-object
asymmetry, it suggests that the relation between the sentence and the subject is
more immediate than that which exists between the sentence and the object. Conversely,
according to (5e′) the object has a closer relationship with the verb than the
subject, since verb and object together form a constituent that excludes the subject.
In what follows we will evaluate the two hypotheses by examining their consequences.
Intuitively, both have some appeal, so we cannot simply rely on intuitions
to discard one or the other. We will investigate whether there are any criteria that
could be used to distinguish between the two ways of integrating the verb into the
sentence. In other words, we are trying to elaborate diagnostics for syntactic structure.
In order to elaborate diagnostics for structure, we will look at attested examples
to see if the language itself perhaps provides any indications that a particular string
of words acts as or is perceived as a unit.
1.3 Substitution
Anaphoric elements, such as, for instance, pronouns, are elements that can be used
to replace strings of words.4
This is illustrated in (6). The pronoun he in (6b) refers
to the subject, the customer in the corner, in (6a), and the pronoun them refers to
the object the drinks. The constituent that is replaced by a pronoun is its antecedent.
(6) a The customer in the corner will order the drinks before the meal.
b He will pay for them later.
The fact that the string of words the customer in the corner can serve as the antecedent
of a pronoun suggests this string is conceived of as a unit; it is a constituent.
At ﬁrst sight, the most important element of this constituent is the noun customer.
The noun customer denotes the entity that we are talking about. A constituent
whose most important element is a noun is called a noun phrase, abbreviated as NP.
Typically, noun phrases can be replaced by pronouns. The string the drinks is also
4
See Chapter 1, Exercises 9 and 10.
TSC02 07/18/2005, 03:14PM74
Diagnostics for Syntactic Structure 75
a constituent: it can be replaced by the pronoun them. Because the most important
element of the constituent is drinks, the plural form of the noun drink, the constituent
the drinks is also a noun phrase.
Before the meal is another constituent, it can be replaced by a word such as then.
The string before the meal contains a noun phrase, the meal, which can be replaced
by it (cf. before it). Before the meal combines a preposition, before, with an NP; it
is a prepositional phrase (PP). Other examples of prepositional phrases are in the
garden, for his brother, after the war, etc.
Depending on their core elements, constituents will be of different types: constituents
belong to categories. The core element of the constituent, which determines its
category, is called the head. An NP contains a noun (N) as its head, an NP is headed
by an N.
Based on this conception of constituency, a verb phrase is a constituent whose
head is a verb. For our test sentence we have elaborated two hypotheses for the
identiﬁcation of the verb phrase (VP): according to (5d) the VP is will order, according
to (5e) the VP is order the drinks. Let us see if substitution of strings containing
a verb can help us choose between these hypotheses. Examine how substitution
affects verbal units in the examples in (7):
(7) a If I had wanted to hurt someone, believe me, I would have done. (Elizabeth
George, Missing Joseph, 1993/1996: 172)
b If Sir Alex wants to sign somebody he can do. (Guardian, 31.12.2002, p. 14,
col. 1)5
In these examples the verb do serves to replace a verb and its object. In (7a) done =
hurt someone; in (7b) do = sign somebody. If a verb and its object can be replaced
together, this suggests that the relevant string of words is a unit, a constituent.
Consider (8). What does the pronoun he refer to? What does the string do so
stand for?
(8) The home secretary is under an obligation to examine any evidence of discriminatory
treatment. He can only do so through assessment, examination
of facts, communication with people and rational arguments and actions.
(Guardian, 9.9.2002, p. 11, col. 3)
In (8) the pronoun he refers to the home secretary, an NP. Do so stands for examine
any evidence of discriminatory treatment, i.e., a verb + its direct object.6
With the examples in (7) and (8) as your models, try to replace a string of words
containing the verb order in our test sentence (6a) above. In (9), we give some
possible results.
5
Note that the substitution illustrated in (7) may not be accepted by all speakers of English. In
particular, speakers of British English accept it more easily than American speakers, and even
among British speakers there is variation.
6
See also Chapter 1, Exercise 9.
TSC02 07/18/2005, 03:14PM75
76 Chapter 2
(9) a The customer in the corner will order the drinks before the meal but in order
to do so before the meal he will ﬁrst need a wine list.
b The customer in the corner will order the drinks before the meal but in order
to do so he needs a wine list now.
c If the customer in the corner wants to order the drinks before the meal he
can do.
Do so in (9a) stands for order the drinks. Let us return to the representations for the
sentence which we are examining. Representation (5d′), based on hypothesis (5d), is
not really compatible with the substitution in (9a). According to (9a) the string
order the drinks acts as one constituent: it can be replaced by do so. But in (5d′) the
verb order does not form a constituent with the object the drinks. The constituent
that contains the verb is the string will order, it is composed of the verb and the
auxiliary. On the other hand, representation (5e′), based on hypothesis (5e), represents
the verb and the object order the drinks as forming a unit. The most important
element in this unit is a verb (order): it tells us what kind of action is going on. A
constituent whose most important element is a verb is a verb phrase or VP.
Compare (9a) with (9b) and (9c). Which constituents are substituted for in
(9b) and in (9c)? In these examples, the substitution process also affects the time
speciﬁcation before the meal. Do so in (9b) and do in (9c) stand for order the
drinks before the meal. Is this type of substitution predicted by hypothesis (5d)
and representation (5d′)? Clearly not, since, as we have just seen, according to (5d)
the auxiliary will and the verb order are taken to form a constituent, but the object
the drinks and the time speciﬁcation before the meal are not represented as being
part of that constituent.
However, hypothesis (5e) and its representation (5e′) also do not predict that the
substitutions in (9b, c) are possible. If substitution identiﬁes constituents, i.e. strings
of words that act as units, then the string order the drinks before the meal must be
a constituent. (5e′) does not offer a basis for this substitution: the time speciﬁcation
is not part of the constituent containing the verb. In order to ensure that the unit
containing the verb, or the VP, contains the time speciﬁcation as well, we should
integrate the time speciﬁcation into the constituent headed by the verb, the VP.
What we want is something like (5f). We have labeled all constituents according to
their category.
(5) f Sentence
NP auxiliary VP
The customer in
the corner will
NP PPV
the drinksorder before the meal
TSC02 07/18/2005, 03:15PM76
Diagnostics for Syntactic Structure 77
In this representation, the string order the drinks before the meal as a whole is
represented as a VP. From the point labeled VP, three lines link down to three
constituents: the verb order, the object the drinks, and the PP before the meal. In the
tree diagram representation, a point in which a number of lines come together is
called a node; nodes are given labels to indicate their category. The lines linking
nodes in a tree to their constituents are called the branches of the tree.
Compare (5f) with (5d′) and with (5e′). Does (5f) have more of the properties of
(5d′) or is it closer to (5e′). To answer this question you should assess the implications
the tree would have for the relations of the various constituents to the sentence.
You will probably conclude that like (5e′), (5f) establishes an asymmetry between
the subject and the object, linking the object more closely to the verb than the
subject, and conversely, relating the subject more directly to the sentence as a whole.
The trees (5e′) and (5f) differ in the way in which they treat the time speciﬁcation,
that is, the PP before the meal. According to (5e′), the time speciﬁcation remains
completely outside the VP (order the drinks); in (5f) the PP before the meal is fully
integrated into the VP. We can ask ourselves what the advantages of (5f) are. And
also: does it have any drawbacks?
Recall that we assume that syntactic structure determines interpretation. Let us
consider the claims made by the different representations above for the interpretation
of the relevant strings. When comparing (5f) and (5e′), we observe that in (5e′)
the verb is assembled with the object, the temporal PP is not part of the resulting
verb-object unit. (5e′) represents the verb and its object as having a closer relationship
than that which holds between the verb and the time speciﬁcation. Such an
asymmetry between the object NP and the temporal PP seems intuitively plausible:
the action described by the sentence is ‘ordering drinks’. The time of that action is
additional information that does not alter the nature of the action: ordering drinks
before a meal or during a meal remains the same kind of activity. (5f) suggests that
the verb is assembled with its object and with the time speciﬁcation at the same
time. Such a representation of the structure of the sentence fails to reﬂect any
asymmetry between the object NP and the time PP. In (5f) the hypothesis seems to
be that the time PP is automatically part of the VP.
What predictions do the representations make for substitution? One prediction of
(5f) is that whenever you replace a constituent containing the verb (= a verb phrase),
this will automatically affect the PP before the meal. Or, phrasing the prediction
differently: it should not be possible to simply replace a unit consisting of the verb
order and its object the drinks. Is this prediction correct? If you turn back to the
examples of substitution in (9), you will conclude that (5f) is incompatible with
(9a), while (5e′) is compatible with this example.
Having already discarded (5d′), we still have a problem to deﬁne what should
be the appropriate representation of the sentence. We ﬁnd ourselves in a sort of
paradox. To account for the substitutions in (9b, c), we would favour (5f).
To account for the substitution in (9a), we would favour (5e′), because the latter
representation captures the closer relation between verb and object. What we need
then is a more articulated representation with a VP that singles out the verb and the
TSC02 07/18/2005, 03:15PM77
78 Chapter 2
object, excluding the PP, and which at the same time allows the verb, the object,
and the PP to be a unit. This can be achieved if we assemble the VP step by step:
ﬁrst we assemble the verb and its object, then we assemble the resulting unit with
the time speciﬁcation:
(5) g Sentence
NP auxiliary VP
will
x PP
V NP
the drinks before the meal
The customer
in the corner order
(5g) combines the good points of (5e′) with those of (5f). Substitution can pick
either the unit x, composed of [V + NP], (as in (5e′)) or the unit labeled VP,
composed of [V + NP + PP], (as in (5f)). (5g) introduces a hierarchy internally to the
VP: the object is “closer” to the verb than the time speciﬁcation.
What kind of label should we give to the constituent order the drinks? Its core
element is a verb, and substitution is by means of do or do so, typically verb phrase
substitutes. This suggests that this constituent is also a VP. The structure we end up
with has what is called a layered VP. First, we construct the core VP, the central
layer containing the verb and its object. Then we extend this layer with a time
speciﬁcation by adding the PP before the meal, creating a larger constituent. If we
assume that the structure of a sentence is related to its meaning (as we have to do in
order to account for structural ambiguity), then the layered VP should feed into
the interpretation. Such a layered structure implies that a time speciﬁcation is less
central to the activity expressed by the verb than the object. This consequence of the
structure mirrors our intuitions about interpretation: a time speciﬁcation such as
before the meal does not deﬁne the kind of activity denoted by the VP, but it merely
provides accessory information on the timing of that activity.
In (5h) we provide both the tree representation and the labeled bracketing
representation for the structure of the sentence. In the latter, the labels in the
left-hand corners identify the category of the constituent. (5h) contains two nodes
labeled VP, one immediately above the other. What is the head of the lower VP?
What is the head of the higher one? For both constituents, the related head is the
verb order. It is not the case that there are two distinct VPs in this structure. Rather,
we have one core VP augmented with an extra constituent, the time speciﬁcation.
The time speciﬁcation is not central to the information conveyed by V. It can be
omitted.
TSC02 07/18/2005, 03:15PM78
Diagnostics for Syntactic Structure 79
(5) h Hypothesis B (revised)
The diagram
S(entence)
NP auxiliary VP
will
VP PP
V NP
the drinks before the meal
The customer
in the corner order
Labeled bracketing
[S [NP The customer in the corner] [will] [VP [VP order [NP the drinks]] [PP before
the meal]]].
The argumentation above introduces two kinds of motivations for adopting the
layered structure with the double VP node. On the one hand, the representation is
motivated on empirical grounds: we have shown the need for introducing the layered
structure by looking at substitution data. The structure is also motivated by the
general hypothesis that the structure of a sentence is related to its interpretation.
The latter is a theoretical hypothesis; we could also say that the second argument is
conceptual, since it follows from the way we have conceived our theory. Both types
of argumentation contribute to the analysis. When formulating an argumentation, it
is important to be able to identify which type of reasoning has been used.
1.4 Movement
In a neutral sentence, the subject typically precedes the verb and the object follows
it. The preverbal position of the subject and the postverbal position of the object are
called their canonical positions. Identify the direct objects in the sentences in (10).
You will observe that objects do not always occupy their canonical positions. Some
objects have been moved to a different position. Try to restore any displaced objects
to their canonical positions.
(10) a Baxter said that he had been using a Sinex liquid decongestant . . . but then
spotted the Vicks inhaler when shopping in Park City, and bought it since
he preferred to use it. “The British one, I have been using since I was about
nine.” (Guardian, 22.3.2002, p. 3, col. 1)
b The news, when it comes, he seems to take well enough. (Guardian, G2,
26.7.2002, p. 2, col. 1)
TSC02 07/18/2005, 03:15PM79
80 Chapter 2
As you can tell, two direct objects have been shifted to the beginning of the sentence.7
If we restore them to their canonical positions we arrive at the following:
(10) a′ I have been using the British one since I was about nine.
b′ When it comes, he seems to take the news well enough.
The underlined strings of words in the primed (10a, b) are constituents. They can be
replaced by pronouns.
(10) a″ I have been using it since I was about nine.
b″ When it comes, he seems to take it well enough.
Typically, constituents can be moved around in the sentence. Consider (11a). What is
the subject related to the verb think? What is the subject of are just a vast conspiracy
to divorce you from ordinary life? Do these subjects occupy their canonical positions?
(11) a A lot of the elements that surround you in the job, you sometimes think
are just a vast conspiracy to divorce you from ordinary life. (Guardian,
26.4.2002, G2, p. 6, col. 4)
In (11a) the subject of the verb think is the pronoun you; it occupies its canonical
preverbal position. The subject of are a vast conspiracy is the noun phrase a lot of
elements that surround you in the job. This subject is not in the expected position,
to the immediate left of are. It has apparently been shifted leftward. We can restore
it to its canonical position as follows:
(11) b You sometimes think a lot of the elements that surround you in the job are
just a vast conspiracy to divorce you from ordinary life.
Identify the displaced constituent in (11c), identify its category, and restore it to its
canonical position:
(11) c Our dustmen arrive too early for me to check, but our ﬁshmonger and his
staff in Petersﬁeld all wear ties (Letters, October 22) and very smart they
look too. (Guardian, 23.10.2002, p. 9, col. 5, Letter to the editor from
David Dew, Horndean, Hants)
In this example the string very smart has been fronted, its canonical position is to
the right of the verb look (they look very smart too). Very smart has as its main
component the adjective smart. A constituent headed by an adjective is an adjective
phrase or AP.
7
For the interpretive effect of preposing constituents see Ward (1988) and the references cited
there. See also Reinhart (1981), Authier (1992), Rizzi (1997).
TSC02 07/18/2005, 03:16PM80
Diagnostics for Syntactic Structure 81
Let us return to our initial example (5a). Is it possible to displace the direct object
or the time speciﬁcation?
(12) a The drinks, the customer in the corner will order before the meal (but the
dessert, they will order later).
b Before the meal, the customer in the corner will order the drinks.
In English, the order in which the subject precedes the verb and the object follows
it is the unmarked word order.8
Fronting of a constituent gives rise to a special or
marked word order, i.e. an order that deviates from the neutral order. We assume
that creating a pattern that deviates from the normal neutral word order is an
additional operation. Recall that we proposed that language is guided by a principle
of economy. In Chapter 1, section 2.2.3, we proposed that units are only inserted if
they have some impact on the interpretation of the sentence. We illustrated this
hypothesis by the discussion of the use of the auxiliary do. We could extend the
application of the principle of economy by proposing that operations that rearrange
constituents also have to be associated with some particular interpretive effect. If
they were not, then, by virtue of the principle of economy, there would be no point
in performing the operation. In other words, a non-neutral or marked order will be
associated with some difference in interpretation. For instance, the fronted object in
(12a) gives rise to some contrasting effect: we contrast the drinks and the dessert.
When we front a time speciﬁcation (12b), we organize the information in the sentence
according to temporal information. In the following examples, the authors are exploring
the possibility of fronting constituents. Identify the fronted constituents. Restore
them to their canonical position. What is the category of the fronted constituents?
(13) a “They must talk about it, and talk about it they must,” he said. Food for
thought, there! It’s a phrase that could add a measure of gravity to any
press conference. “We must do this, and do this we must.” (Simon Hoggart,
Guardian, 29.1.2003, p. 2, col. 5)
b But I was still a long way from ﬁguring out what my goal was. I told the
governor [of the prison] that I wasn’t sure how I was going to manage it –
but manage it I would. (Guardian, G2, 15.5.2003, p. 7, col. 4)
8
The unmarked order is the neutral word order. Marked word orders are less neutral in that
they carry some speciﬁc communicative effect. For instance, the word order in sentence (ia) is
unmarked: the object NP this book follows the verb like. In (ib) fronting of the object NP this
book gives rise to a marked word order. This example could be used, for instance, if the
speaker wants to contrast the book under discussion with another book.
(i) a I didn’t like this book very much.
b This book, I didn’t like very much (but that one I really enjoyed).
For discussion of the concept of markedness and its relation to interpretation see de Hoop,
Haverkot, and van den Noort (2004). For the interpretive effect of preposing constituents see
Ward (1988).
TSC02 07/18/2005, 03:16PM81
82 Chapter 2
In (13a) the author fronts talk about it and do this, and in (13b) manage it is
fronted. Since these strings are constituents whose most important element is the
verb, they are verb phrases:
(13) a′ “They must talk about it, and [VP talk about it] they must,” he said. Food
for thought, there! . . . “We must do this, and [VP do this] we must.”
b′ I told the governor [of the prison] that I wasn’t sure how I was going to
manage it – but [VP manage it] I would.
Let us return to our test example, (5a). On the basis of the structural representation
of the sentence, we should be able to predict how VP fronting will apply. In representation
(5h), there are two nodes labeled VP: the core VP and the augmented VP
including the temporal speciﬁcation. Try fronting either of these. You will ﬁnd that
both operations give an acceptable result:
(14) a Order their drinks before the meal, they will.
b Order their drinks, they will before the meal.
The data in (14) provide empirical support for the structure in (5h). Would
the fronting data in (14) be compatible with the other representations that we
had envisaged, that is (5d′), (5e′), and (5f)? Adopting (5d′) would pose a problem:
in that representation there is no constituent containing the verb and the object
(14a) or the verb, the object, and the time speciﬁcation (14b). (5e′) fares slightly
better in that it allows (14b) but it does not allow (14a). Conversely, (5f) allows
(14a) but not (14b). So (5h) represents a better hypothesis about the structure of
the sentence.
1.5 Question formation
Invent an answer to the questions in (15):
(15) a Who have you invited to the party?
b Who has invited you to this meeting?
c What have you bought?
d An Indian meal or ﬁsh and chips. Which do you prefer?
When you think up answers to the questions above, it is quite possible that you will
come up not just with one word but with a string of words. The relevant string
of words functions as a unit in the communicative exchange: it provides the answer
to the question. An answer to a question will ﬁll in the missing information that is
represented in the question by words such as what, who, which. As discussed in
section 1.3, strings of words that are replaced by one word are constituents. For
instance, take (15a). A possible answer could be:
TSC02 07/18/2005, 03:16PM82
Diagnostics for Syntactic Structure 83
(16) a My friends from college.
Question (15a) implies that ‘you have invited someone’, and it signals that the
speaker doesn’t know who the invitee was. (16a) supplies the missing information:
it supplies a replacement for the interrogative word who. Questions which ask for a
replacement of an interrogative constituent, are called constituent questions: the
answer to such questions supplies the missing constituent. Because most interrogative
words in English begin with wh, such questions are also called wh-questions,
and interrogative constituents such as who, what, etc. are called wh-constituents.9
(16b) inserts the answer to (15a) into the sentence. Compare the form of question
(15a) and the form of the answer (16b):
(16) b I have invited [my friends from college].
(15a) differs from (16b) in a number of ways. (i) In (15a) the direct object is realized
as a wh-constituent, who. (ii) This (interrogative) direct object of the verb invited
does not occupy its canonical position but it takes up an initial position. (iii) There
is an application of subject auxiliary inversion (SAI, see also Chapter 1, section 2.2).
Of particular relevance to the current discussion is the observation that in the
answer the interrogative constituent of the question is replaced by the constituent
(here my friends from college). We can conclude that another technique for identifying
constituents is to examine whether the strings of words that are taken to be
constituents can serve as answers to questions.
Formulate constituent questions to target each of the underlined constituents in
our test sentence:
(17) a The customer in the corner will order the drinks before the meal.
Recall that we also identiﬁed a constituent centered around the verb. How is the VP
questioned in (18)?
(18) a I think that would be the worst thing in the world for him, a family
holiday. What’s he going to do? Sit on the beach? (Based on Guardian, G2,
29.7.2002, p. 4, col. 3)
b What is Sylvia to do? What are we all meant to do? Hang our cars from
the trees? Throw them away? (Guardian, G2, 28.4.2003, p. 9, col. 3)
c “We need fewer people.” “What would you do? Eliminate people?” (Based
on a cartoon in Washington Post, 29.4.2003, p. C12)
As you can see, verb phrases can function as targets for wh-questions. In our test
sentence (5a), repeated in (17a) above, we identiﬁed a core VP (order the drinks)
9
We look more carefully at the fronting process involved in the formation of wh-questions in
Chapter 5.
TSC02 07/18/2005, 03:16PM83
84 Chapter 2
and what we called an augmented VP (order the drinks before the meal). Using the
examples in (18) as a model, try to formulate questions targeting either of these
VPs. Based on (17a) you could form either of the following questions:
(17) b What will the customer in the corner do before the meal?
c What will the customer in the corner do?
Wh-questions conﬁrm that the strings order the drinks and order the drinks before
the meal are constituents: each string can be the answer to a wh-question.
1.6 Deletion/ellipsis
Consider the following fragments. In the second part of each extract some material
has been omitted. The site of the ellipsis is to the right of the underlined words.
Supply the words that have been omitted. On what basis can you recover the
omitted material?10
(19) a It is up to us other teams to take steps to rectify our performance deﬁciency,
and we will. (Guardian, 8.10.2002, p. 15, col. 7)
b When he ﬁrst ran for ofﬁce four years ago, Gov. Gray Davis vowed to save
California’s old-growth forests. He hasn’t, as Moloney sees it. (Los Angeles
Times, 26.11.2002, p. B7, col. 2)
c After all, Francesca’s hardly news any more. We are all trying to forget her.
As if we could. Although we should. I can’t. (Francis Fyﬁeld, Undercurrents,
2001: 50)
d I saw Mr Clark stand up, throw a punch at Mr McAlpine, kick the table
over, jump at him on the ground, and start choking him, before two chefs
came out of the kitchen and pulled them apart . . . We have an open-plan
kitchen, and so my staff jumped in and separated them; I wouldn’t like to
think what would have happened if they hadn’t. (Guardian, 11.11.2002,
p. 9, col. 4)
e All in the name of a pretence that, with just a little bit more time passing, all
obstacles will miraculously recede. They won’t. (Guardian, 6.5.2003, p. 16,
col. 2)
f Only those who were in the room know the absolute truth of this story.
No one else probably ever will. (Washington Post, 25.3.2004, p. D3, col. 5)
g If we could charge more money, we would. (Wall Street Journal, 29.3.2004,
p. A6, col. 6)
h Everyone says you can’t be scientiﬁc and fun, but we think you can. (New
York Times, 8.3.2004, p. C5, col. 2)
10
See also Chapter 1, Exercise 8.
TSC02 07/18/2005, 03:16PM84
Diagnostics for Syntactic Structure 85
In these examples a constituent that is recoverable from the preceding context has
been omitted. For instance, in (19a) we can recover the string take steps to rectify
our performance deﬁciency from the preceding sentence. In each of the examples,
the auxiliary is retained and a constituent centered around the verb is omitted. We
restore the omitted strings in (19′):
(19′) a We will [VP take steps to rectify our performance deﬁciency].
b He hasn’t [VP saved California’s old-growth forests].
c As if we could [VP forget her]. Although we should [VP forget her]. I can’t
[VP forget her].
d I wouldn’t like to think what would have happened if they hadn’t [VP
jumped in and separated them].
e They won’t [VP miraculously recede].
f No one else probably ever will [VP know the absolute truth of this story].
g we would [VP charge more money].
h you can [VP be scientiﬁc and fun].
Let us once again turn to representations (5d′) and (5h). By adopting (5h), we can
straightforwardly describe the processes applying in (19) as an illustration of verb
phrase ellipsis. If we adopt (5d′) we cannot describe the process in (19) as VP
ellipsis: in (5d′) the auxiliary and the verb form a constituent and the object remains
outside this constituent.
Taking (20a) as a basis, how could VP ellipsis be applied to B’s reply?
(20) a Speaker A: The customer in the corner will order the drinks before the
meal.
Speaker B: Actually, I wouldn’t be so sure that he will order the drinks
before the meal.
The application of VP ellipsis to (20a) is given in (20b), where the symbol [VP ∅]
stands for the omitted VP.
(20) b Speaker A: The customer in the corner will order the drinks before the
meal.
Speaker B: Actually, I wouldn’t be so sure that he will [VP ∅].
1.7 Focalizing a constituent
1.7.1 THE CLEFT SENTENCE
In the following examples, a special word order pattern is used which has the effect
that one constituent is promoted to the foreground while the remainder of the
sentence is backgrounded.
TSC02 07/18/2005, 03:16PM85
86 Chapter 2
(21) a It was the prison chaplain’s wife who ﬁrst gave me an inkling that I might
have a talent for writing. (Guardian, G2, 21.4.2003, p. 2, col. 1)
b Ford directed many ﬁlms but it is for Westerns that he will be remembered.
(Guardian, 21.4.2003, p. 310, col. 4)
Both examples contain the pattern it is X who/that Y. In this pattern, the element
in the position of X is highlighted. It is presented as prominent information. The
elements in Y are backgrounded. Consider for instance (21a). We can paraphrase it
with (21a′).
(21) a′ The prison chaplain’s wife ﬁrst gave me an inkling that I might have a
talent for writing.
The word order in (21a′) is neutral. (21a′) does not give the same prominence to the
constituent the prison chaplain’s wife as the original example, (21a). The information
conveyed by the two variants is similar: both sentences communicate that at
some point in the past the prison chaplain’s wife gave the speaker the idea that he
might be a good writer. Sentences (21a) and (21a′) describe the same event. Could
you imagine a situation in which (21a) is true and (21a′) is false? This is not
possible. If (21a) is a true statement, then (21a′) will also be true and vice versa.
The effect of the rewording in (21a) is “presentational”: we reorganize the way the
information is presented. In (21a) the writer highlights that the initial trigger for
the speaker’s writing was the chaplain’s wife.
The wording of (21a) serves to focus on a particular constituent and to background
the remainder of the sentence. The pattern where we focus on a constituent
using the it is X who/that Y pattern is called a cleft sentence. Clefting is a way of
reorganizing the information in a sentence, backgrounding some information and
foregrounding the focal information. As you can see, clefting foregrounds or focuses
on a constituent: in (21a) the NP the prison chaplain’s wife is singled out or focused
on; in (21b) the PP for Westerns is focused on.
Given that clefting promotes one constituent to the foreground, we can use the
pattern to identify constituents. Apply clefting to our test sentence. Again you will
ﬁnd that it serves to isolate constituents: in (22a) the NP the customer in the corner
is focused, in (22b) the PP before the meal is focused on, and in (22c) the NP the
drinks is focused on.
(22) a It is [NP the customer in the corner] who will order the drinks before the meal.
b It is [PP before the meal] that the customer in the corner will order the drinks.
c It is [NP the drinks] that the customer in the corner will order before the meal.
Recall that we have been entertaining two hypotheses for the representation of the
sentence: either the auxiliary and the verb form a constituent (23a), or else the verb
forms a constituent with its object. On the basis of additional evidence, we elaborated
the latter option and proposed that the VP was layered (23b).
TSC02 07/18/2005, 03:17PM86
Diagnostics for Syntactic Structure 87
(23) a Hypothesis A
[S [NP The customer in the corner] [VP will order] [NP the drinks] [PP before
the meal]].
b Hypothesis B (revised)
[S [NP The customer in the corner] [AUX will] [VP [VP order [NP the drinks]]
[PP before the meal]]].
The cleft patterns in (22) do not bear on (23a) and (23b), because clefting of the
type illustrated here does not affect the VP.11
1.7.2 THE PSEUDO-CLEFT SENTENCE
Consider the effect of the rewordings in the paired sentences in (24)–(27).
(24) a I don’t need the equivalent of another car loan.
b What I don’t need is the equivalent of another car loan. (Chicago Tribune,
22.12.2002, section 15, p. 3, col. 1)
(25) a You are seeing the biblical law of reciprocity in Prince George’s Country.
b What you are seeing in Prince George’s Country is the biblical law of
reciprocity. (Washington Post, 29.4.2003, p. A7, col. 2)
(26) a She needed someone to talk to.
b What she needed was someone to talk to. (Washington Post, 29.4.2003,
p. B1, col. 2)
(27) a They will force them underground.
b What they will do is force them underground. (Guardian, 9.7.2002, p. 8,
col. 8)
(28) a Contacting his relatives will cause mayhem in his family.
b What contacting his relatives will do is cause mayhem in his family. (Adapted
from Guardian, G2, 11.4.2003, p. 11, col. 3)
As was the case for the clefting paraphrases discussed in section 1.7.1, the content
of the paired sentences is near identical. Both sentences in (24), for instance, convey
that ‘there is no need for the equivalent of another car loan’. As was also the case
for the clefting paraphrases, the two sentences have the same truth conditions. If
(24a) is a true statement, then (24b) will also be a true statement. The difference is
again one of presentation and focus. (24b) serves to highlight one informational
11
This is not quite correct: in Hiberno English, the variant of English spoken in Ireland, VPs can
be clefted. (Cottell 2002: 111)
(i) Q. What are the women doing?
A. It’s playing backgammon that they are.
TSC02 07/18/2005, 03:17PM87
88 Chapter 2
unit, here the NP, the equivalent of another car loan, and it does so by splitting this
constituent off from the remainder of the sentence using a paraphrase with what.
The pattern displayed here is referred to as pseudo-clefting. Identify the focused
constituents in the (b)-examples (26)–(28). What is their category?
(26) b′ What she needed was someone to talk to.
(27) b′ What they will do is force them underground.
(28) b′ What contacting his relatives will do is cause mayhem in his family.
In (26) the focused constituent is an NP. In (27) and (28) it is a constituent centered
around a verb, a VP (force them underground, cause mayhem in the family). Do
examples (27) and (28) bear on the choice between representations (23a) or (23b)?
What would (23a) predict with respect to pseudo-clefting of a VP? And what would
be the predictions of (23b)?
Using examples (27) and (28) as models, let us also apply pseudo-clefting to our
test sentence. Could (23a) constitute a basis for pseudo-clefting of a verb-centered
constituent? That is, can we use pseudo-clefting to highlight the VP as represented
in (23a)? The answer is that it is not possible to pseudo-cleft the VP as represented
in (23a):
(29) a *What the customer in the corner will do the drinks before the meal is will
order.
On the other hand, (23b) has two constituents labeled VP: order the drinks, the
core VP, and order the drinks before the meal, the augmented VP including the
time speciﬁcation. Try pseudo-clefting either of these. Either VP can be focused by
pseudo-clefting.
(29) b What the customer in the corner will do before the meal is order the drinks.
c What the customer in the corner will do is order the drinks before the meal.
Once again, the layered VP hypothesis of (23b) allows us to predict that pseudoclefting
will affect a constituent centered around the verb, containing the direct object
and possibly the time PP. If we adopt representation (23a), it is hard to account for
the fact that the verb and the constituents to its right can be treated as one constituent
and undergo pseudo-clefting. In (23a), the verb is taken to form a constituent
with the auxiliary and the object and the time PP are separate constituents.
Identify the highlighted constituent in the following examples. Do the examples
bear on the choice between (23a) and (23b)?
(30) a In the Lower 48 states, people consider reindeer as pets, so the last thing
they would do is eat them. (Chicago Tribune, 22.12.2002, section 1, p. 16,
col. 1)
TSC02 07/18/2005, 03:17PM88
Diagnostics for Syntactic Structure 89
b All Pastor Edgar Chacon wanted to do, he says now, was protect the
children. (Los Angeles Times, 26.11.2002, p. B4, col. 4)
c All we can do is do it well. (Guardian, 1.11.2002, G2, p. 9, col. 2)
As before, in (30) we focus on a verb-centered constituent, containing the verb, its
object, and an additional speciﬁcation of manner in (30c). The sentences in (30) are
again best compatible with a representation like (23b).
1.8 Co-ordination
As units of structure, constituents can be manipulated in various ways. So far we
have illustrated that we can replace a constituent by a shorter form (1.3), that we
can move a constituent around (1.4), that a constituent can function as the answer
to a question (1.5), that we can omit a constituent (1.6), and that we can highlight
a constituent by clefting (1.7.1) or by pseudo-clefting (1.7.2). Sometimes, we may
decide to link two constituents together. To do this we co-ordinate them, i.e. we
link them by means of a co-ordinating conjunction (and, or, but).
(31a) contains two sentences. There is some redundancy in this passage. Identify
the overlapping parts between the sentences. Reword the passage to express the
information in (31a) in a more compact way, using only one sentence.
(31) a The customer in the corner will order the drinks before the meal. He will
also order the dessert before the meal.
We can condense the information in (31a) into one sentence by co-ordinating
the object of the ﬁrst sentence, the NP the drinks, and the object of the second
sentence, the NP the dessert, thus turning these two NPs into one constituent:
the drinks and the dessert. To do this, we use the conjunction and. As a result of the
co-ordination, the constituents form one single constituent, as represented by the
outer brackets surrounding the co-ordinated NPs:
(31) b The customer in the corner will order [[NP the dessert] and [NP the drinks]]
before the meal.
What type of evidence could we provide that the string the dessert and the drinks in
(31b) is one constituent? One way of showing that the string the dessert and the
drinks is one constituent is by asking a constituent question targeting just this string
(31c). The string can also be replaced by a pronoun (31d).
(31) c What will the customer in the corner order before the meal?
d The customer in the corner will order them before the meal.
Since the co-ordinated constituent can jointly be replaced by a pronoun, them, this
suggests that the string is an NP, which is the result of co-ordinating two NPs.
TSC02 07/18/2005, 03:17PM89
90 Chapter 2
How could (32a) be expressed more economically by using co-ordinated
structures?
(32) a The customer in the corner will order the drinks before the meal. He will
also order the dessert before the meal. He will also order the coffee before
the meal.
In (32a) there are three constituents that function as the object of order: the
drinks, the dessert, the coffee. They can again be co-ordinated. When we have more
than two elements to co-ordinate there are two options, illustrated in (32b) and
(32c).
(32) b The customer in the corner will order the drinks and the dessert and the
coffee before the meal.
c The customer in the corner will order the drinks, the dessert, and the coffee
before the meal.
Underline all the co-ordinated constituents in the examples in (33), and identify the
category of each of the co-ordinated constituents.
(33) a Det. Insp. Smith told lies in one of his reports and to the enquiry. (Guardian,
15.7.2003, p. 1, col. 7)
b Being in Europe does tend to mean that public transport is functional, that
public health care is not considered a dangerous pipe-dream, and that education
is valued. (Adapted from Guardian, 9.12.2002, p. 12, col. 3)
c Among the larger issues here are why this happened at all, who allowed it
to happen and why the law reinforcement establishment refused to intervene
even after it was clear that a great injustice was occurring. (New York
Times, 28.4.2003, p. A25, col. 1)
d Many parents with children in these schools have felt the impact and seen
the point. (Guardian, 17.7.2002, p. 2, col. 3)
e They are also re-equipping six Iraqi hospitals that were looted and building
a plant in Basra. (New York Times, 28.4.2003, p. A11, col. 6)
f He has made its programs newsworthy and kept the institution aﬂoat.
(New York Times, 28.4.2003, p. A25, col. 4)
g Jones said urban sprawl and heightened environmental concerns were
imposing increased limits on U.S. military activities in Western Europe and
driving up costs. (Washington Post, 29.4.2003, p. A6, col. 3)
h Boyle testiﬁed that she told Malvo four times that he could be silent or see
an attorney but that Malvo continued to talk about the shootings in a
relaxed, almost convivial way, laughing about Buchanan’s slaying and other
shootings. (Washington Post, 29.4.2003, p. B1, cols 3–4)
In (33) we ﬁnd the following illustrations of co-ordination:
TSC02 07/18/2005, 03:17PM90
Diagnostics for Syntactic Structure 91
(33) (i) PP co-ordination
• [PP in one of his reports] + [PP to the enquiry] (a)
(ii) NP co-ordination
• [NP urban sprawl] + [NP heightened environmental concerns] (g)
• [NP Buchanan’s slaying] + [NP other shootings] (h)
(iii) sentence co-ordination
• [S that public transport is functional] + [S that public health care
is not considered a dangerous pipe-dream] + [S that education is
valued] (b)
• [S why this happened at all] + [S who allowed it to happen] + [S why the
law reinforcement establishment refused to intervene even after it was
clear that a great injustice was occurring] (c)
• [S that she told Malvo four times that he could be silent or see an
attorney] + [S that Malvo continued to talk about the shootings in a
relaxed, almost convivial way, laughing about Buchanan’s slaying and
other shootings] (h)
(iv) VP co-ordination
• [VP felt the impact] + [VP seen the point] (d)
• [VP re-equipping six Iraqi hospitals that were looted] + [VP building a
plant in Basra] (e)
• [VP made its programs newsworthy] + [VP kept the institution aﬂoat]
(f)
• [VP imposing increased limits on U.S. military activities in Western
Europe] + [VP driving up costs] (g)
• [VP be silent] + [VP see an attorney] (h)
Do the co-ordination data in (33) bear on the choice between Hypothesis A (23a)
and (revised) Hypothesis B (23b) for the structure of the VP? In particular, would
Hypothesis A in (23a) lead us to expect the patterns of VP co-ordination displayed
in (33)? The answer is negative. In (33d), for instance, co-ordination affects two
constituents consisting of a verb and the object. Crucially, the auxiliary remains
outside the co-ordinated structure.
Recall that the revised hypothesis B in (23b) allows for layering inside the VP: the
verb and the object form a core VP, which then combines with less central material,
the time speciﬁcation in our earlier example. If VPs can co-ordinate, then we predict
that for our test example two types of VP co-ordination are possible, one affecting
the augmented VP, one affecting just the core VP. Discuss the relevance of the
co-ordinations in (34) for this prediction.
(34) a The customer in the corner will order the drinks before the meal and
accompany his guests into the dining room.
b The customer in the corner will order the drinks and choose the dessert
before the meal.
TSC02 07/18/2005, 03:18PM91
92 Chapter 2
In (34a) the co-ordinated strings are order the drinks before the meal and accompany
his guests into the dining room. Once again, the ﬁrst co-ordinated constituent
contains the verb, the object, and the time speciﬁcation. This conﬁrms that the
string order the drinks before the meal is a constituent. (34b) conﬁrms that the
string order the drinks is also a constituent, since it is co-ordinated with choose
the dessert. The data in (34) offer further support for the layered VP in (23b).12
2 The Verb Phrase and the Sentence: A First
Exploration
2.1 Starting point
Recall that in the previous section we were led to choose between two representations
of sentence structure, repeated in (35). The label S abbreviates the label “Sentence.”
(35) a S
NP PPVPNP
The customer in the corner
Aux V
before the mealthe drinkswill order
b S
VPAuxNP
The customer
in the corner
VP PP
will
V
order
NP
the drinks before the meal
While (35a) is “ﬂat,” (35b) introduces hierarchical levels of structure. The structure
in (35b) is more “articulated”: the constituents of the sentence have different
12
Exercises 1A, 3, 4, 5, 6, 8, 9, 19.
TSC02 07/18/2005, 03:18PM92
Diagnostics for Syntactic Structure 93
relations between each other. There are two important differences between the
representations:
(i) (35a) presents the subject NP and the object NP as being on equal footing;
(35b) creates a subject/object asymmetry. From the root S it takes one
single step to reach the subject NP, it takes three steps to reach the object
NP.13
(ii) Conversely, in (35b), the object is contained in the VP, the subject is outside
it. This means that in (35b) the object has a closer relationship with the verb
than the subject.
(iii) (35a) presents the direct object and the temporal PP as being on equal footing.
In (35b) the object and the verb form a separate unit, the core VP; the temporal
PP is added to the periphery of that same VP.
2.2 The relation between the auxiliary and the VP
Representation (35a) follows from the intuition that there is a close relationship
between the auxiliary and the verb. This relationship is not captured in (35b), in
which the auxiliary has a symmetric relation with subject and VP. Consider the
underlined co-ordinating conjunctions in (36). For each conjunction, identify the
co-ordinated constituents. Does either of the structures in (35) allow us to predict
the co-ordinations in (36)?
(36) a Parents of the 53 Dartmouth swimteam members are pressing administrators
and college trustees for the team’s reinstatement, and are asking for
help from Dartmouth alumni and donors. (Adapted from Boston Globe,
4.12.2002, p. A3, col. 3)
b Stop and search isn’t working and won’t work. (Guardian, 9.11.2002,
p. 11, col. 6, letter to the editor from Marc Cohen)
c State ofﬁcials say Massachusetts may eliminate a prescription drug program
that covers 80,000 elderly, and is considering a variety of other cuts and
restrictions. (Based on New York Times, 28.4.2003, p. A21, col. 2)
d Three high-rise towers would wall off the stadium from the skyline and
would drastically shrink the center ﬁeld park. (Based on New York Times,
28.4.2003, p. A22, col. 1)
e We believe that such a proposal could seriously undermine the voucher
program and could potentially harm the millions of low-income people
assisted with housing vouchers. (Washington Post, 29.4.2003, p. A8,
col. 2)
13
In Chapter 5, sections 3.2.3.1, 3.2.3.2, 5.2.2, and 5.3.2.2, we will discuss a number of subject/
object asymmetries in connection with question formation and relative clause formation.
TSC02 07/18/2005, 03:18PM93
94 Chapter 2
The constituents co-ordinated by the underlined conjunctions in the above
examples are isolated in (37):
(37) a [are pressing administrators and college trustees for the team’s reinstatement]
+ [are asking for help from Dartmouth alumni and donors]
b [isn’t working]
+ [won’t work]
c [may eliminate a prescription drug program that covers 80,000 elderly]
+ [is considering a variety of other cuts and restrictions]
d [would wall off the stadium from the skyline]
+ [would drastically shrink the center ﬁeld park]
e [could seriously undermine the voucher program]
+ [could potentially harm the millions of low-income people assisted with
housing vouchers]
The problem arising for both representations in (35) is that the co-ordinated constituents
in (36) correspond to the combination of the auxiliary, the verb and its
complements and other speciﬁcations. But in neither of the representations in (35)
do these strings form a constituent: (35a) does combine the auxiliary with the verb,
but neither the direct object nor the temporal speciﬁcation is part of the VP. In
(35b) the auxiliary and the VP do not form a constituent.
Based on our test sentence we can create sentences such as the following, raising
a similar problem for the representations in (35):
(38) a The customer will order the drinks before the meal but may pay for them
later.
b The customer will order the drinks before the meal and is waiting for the
wine list.
c The customer in the corner has just taken his seat and will order the drinks
before the meal.
According to representation (35a), to be able to operate on a constituent that
contains the auxiliary, the verb, and the additional material, we would have to
take the complete sentence (S). The sequence VP + NP + PP is composed of three
autonomous constituents, but these constituents do not exhaustively form a bigger
constituent. In (35b), the problem is identical: the sequence Aux + VP does not
form a constituent without the subject NP.
The attested data in (36) and the experimental data in (38) both lead to the
conclusion that Aux and the components of the VP in (35b) must also form a
constituent together. Let us adjust the tree accordingly. Rather than having the
ternary structure in (35b), in which three branches start from the S node, we need a
binary branching structure, as in (39): for the time being we do not give a speciﬁc
label to the constituent formed by the auxiliary and the VP. We return to this point
in section 3.1 and also in the next chapter.
TSC02 07/18/2005, 03:18PM94
Diagnostics for Syntactic Structure 95
(39) a [S [NP The customer in the corner] [CONSTITUENT [AUX will] [VP [VP order [NP
the drinks]] [PP before the meal]]]].
b S
CONSTITUENTNP
The customer
in the corner
Aux VP
VP PP
V
order
NP
will the drinks before the meal
This structure splits the sentence into two parts, the subject and a second constituent,
which we could loosely refer to as the “predicate.”14
The subject “speciﬁes” or
delimits the domain of application of the predicate. That is to say: the predicate tells
us what kind of action is taking place, and the subject says to which entity this
action is applied. The role of the auxiliary is to link the VP and the subject. The
auxiliary also qualiﬁes the validation of that link: for instance, the auxiliary may
specify that the link will only be validated in the future (will), or that the validation
is situated in the past (has), etc.
The structures in (39a, b) give rise to an immediate question. Not every English
sentence contains an auxiliary. We will have to devise a way of representing the
structure of sentences such as (39c).
(39) c The customer in the corner ordered the drinks before the meal.
We return to this issue in Chapter 3, section 1.2.
2.3 Layered structures
2.3.1 COMPLEMENTS VS. ADJUNCTS IN THE VP
Recall that we proposed that the structure of the VP is layered. In (40) we repeat the
structure we arrived at based on our experimental sentence.
14
Exercises 2 and 10.
TSC02 07/18/2005, 03:18PM95
96 Chapter 2
(40) a [VP [VP order [NP the drinks]] [PP before the meal]]
b VP
PPVP
V NP
the drinks before the mealorder
Let us assume we start building the sentence from the verb, which denotes the activity
that will be reported in the sentence, here order. If we proceed from the bottom of
the structure to the top, (40b) seems to suggest a certain ordering in the assembly of
a sentence. First we assemble the core VP, containing the verb and its object (order
the drinks); then we augment this core by adding a temporal PP (order the drinks
before the meal). This structure reﬂects the intuition that the temporal PP is
interpretively less central to the activity expressed in the sentence than the direct
object. The PP before the meal does not serve to deﬁne the kind of activity
expressed by the VP; it simply provides accessory information on the timing of that
activity. The PP is more peripheral to the VP: the core VP consists of the verb and
the object. That core VP is assembled ﬁrst.
In traditional terms the verb order is a transitive verb. The verb order selects the
object the drinks, in that it requires that there be a complement and it determines
the type of complement that is required, here an NP. We call the constituent
selected by the verb a complement. The time PP is not a complement of order; the
verb order does not select a time speciﬁcation, it does not require that there be a
time speciﬁcation, nor does it require that a time speciﬁcation in the sentence be of
a particular category. Material that is not selected is attached outside the core VP.
The temporal PP is said to be adjoined; a constituent that is adjoined to another
constituent is also called an adjunct.
(40) c VP
PPVP
V NP
the drinks
complement
before the mealorder
adjunct
The nature of the process that puts together the core VP is slightly different from
that of adjoining the time PP to it. When we assemble the core VP we combine a
TSC02 07/18/2005, 03:18PM96
Diagnostics for Syntactic Structure 97
head, the verb, which is one word, with a phrase, the NP object. The resulting unit
denotes a particular type of action. The head selects the complement. When we
adjoin the adjunct, we assemble two ready-made phrases: the “ﬁnished” core VP
and, in our example, the adjunct PP. The adjunct, though part of VP, is not fully
integrated into its core and it does not serve to differentiate a particular action from
another one. Verbs do not select time speciﬁcations, and if there is going to be a
time speciﬁcation this can be expressed in different ways.
To use a metaphor: we can compare adjoined constituents to, for instance, the
garage that is built next to a house. Garages are “external” to a house. If you say
that you are “in” the house, you will not normally be thought to be in the garage.
On the other hand, if you sell a house you will also sell the garage with it. And
moreover, the presence of a garage will mean that the house is worth more. But
garages remain extras: in order to qualify as a house, a building does not need a
garage; if you demolish the garage that goes with a house, the house itself remains
a house.
The structure in (40c) is elaborated for the VP. We can generalize (40c) to all
constituents, proposing that for each type of head we distinguish its complements,
the constituent selected by the head in question, and which, together with the head,
forms the core constituent, from the adjoined constituents, peripheral elements
which may but need not be added. We use the labels X , Y, and Z as variable labels
that stand for any category type (N, V, P, etc.). In (41), the constituents YP and ZP
will be formed according to the same schema: they will have a head (Y, Z), and may
have a complement and/or an adjoined constituent.
(41) XP
ZPXP
X
head
YP
adjoined constituent
complement
2.3.2 DIRECTION OF ADJUNCTION
Identify the VPs in (42). Do the verbs select a complement? Are there any adjoined
constituents? What is the category of the adjoined constituents? Discuss the difference
between (42a) and (42b). Draw the structure for the VPs.
(42) a One of the most controversial takeovers in British sporting history was
awaiting a government decision last night. (Based on Guardian, 13.3.1999,
p. 1, col. 1)
b One of the most controversial takeovers in British sporting history was last
night awaiting a government decision. (Guardian, 13.3.1999, p. 1, col. 1)
TSC02 07/18/2005, 03:18PM97
98 Chapter 2
In (42a) last night is an adjoined constituent; it follows the verb and it is realized by
an NP. In (42b) the same constituent precedes the verb. To allow for both the order
adjunct – verb and the order verb – . . . – adjunct, we can enrich our structure by
allowing both left-adjunction and right-adjunction. Using our earlier architectural
metaphor, garages may be built to the right of a house or to its left.
(40) d Right-adjunction Left-adjunction
VP
adjunctVP
V complement
Among our earlier examples, (19e) also illustrated left-adjunction. We repeat it for
convenience in (43a). We can reword the example and replace the left-adjoined
adjunct by a right-adjoined one as in (43b).
(43) a All in the name of a pretence that, with just a little bit more time passing,
all obstacles will [VP miraculously [VP recede]].
b All in the name of a pretence that, with just a little bit more time passing,
all obstacles will [VP [VP recede] miraculously].
This example is of interest because the core VP does not contain any complement.
This is so because the verb recede does not select a complement. Recede is an
intransitive verb. Adopting the format in (40d) we assign the structures in (43c) and
(43d) to the VPs in (43a) and (43b). Though recede is just one word, we represent
it as constituting a core VP, because we want to show that the adjunct miraculously
is not part of the core VP but is adjoined to it.
(43) c Left-adjunction d Right-adjunction
VP
VPadjunct
miraculously V
recede
VP
adjunctVP
V miraculously
recede
Identify the co-ordinated VPs in the following example:
(43) e We budget for shows with the expectation that it will increase the gaming
revenue and in turn pay for the show. (New York Times, 28.4.2003, p. B5,
col. 4)
VP
adjunct VP
V complement
TSC02 07/18/2005, 03:19PM98
Diagnostics for Syntactic Structure 99
As you can see, the second co-ordinated VP in turn pay for the show contains a leftadjoined
PP, in turn.
(43) f [VP in turn [VP pay for the show]]
Finally, exploiting our metaphor once again, it is also possible to build a garage
both to the right and one to the left of a house. In the same way, we can also adjoin
constituents both to the right and to the left of a VP. (43g) is a constructed example
with one adjunct to the right and one to the left, both adjuncts can also appear on
the left (43h) or on the right (43i).
(43) g They will carefully analyze the ﬁnancial situation of the company later on.
h They will later on carefully analyze the ﬁnancial situation of the company.
i They will analyze the ﬁnancial situation of the company carefully later on.
Our hypotheses of VP-layering and of adjunction can account for these word
orders. Observe that examples such as those in (43g–i) also show that a VP may be
augmented by more than one adjunct.15
2.3.3 OV ORDERS?
The data we have encountered lead to the hypothesis that adjunction may either be
to the right or to the left of the VP. When the object of a verb was located in the
VP,16
then it was always found to the immediate right of the verb. This follows from
our structure: the complement of a head appears to its right. We may however
wonder whether this is the only option. Having introduced both left-adjunction and
right-adjunction, should we also allow for the complement to occur either to the
right or to the left of the verb in the VP? In other words, do we also have to allow
for the following structure for the core VP?
(40) e Hypothesis: Object-Verb patterns in English
VP
Vcomplement
How do we ﬁnd the evidence for or against this hypothesis? We could look for
examples of this type in which the object remains in the VP but occurs to the
15
Exercise 14. We will return to multiple adjuncts in Chapter 3, section 2.1, and in Exercise 3 of
Chapter 3.
16
The object can be moved out of the VP to an initial position, as shown in section 1.4. This
fronting operation is also used for forming constituent questions that bear on the object.
Clefting (section 1.7.1) and pseudo-clefting (section 1.7.2) may also rearrange the position of
the object with respect to the verb. For a discussion of movement see Chapter 5.
TSC02 07/18/2005, 03:19PM99
100 Chapter 2
immediate left of the verb. Now, suppose we do not ﬁnd any such examples, even
after an intensive search using sophisticated computer technology. Would that deﬁnitely
mean that such structures are to be excluded on principled grounds? Though
the total absence of relevant examples could well be taken to suggest that certain
structures are ill conceived, we cannot be totally sure about this conclusion. It could
also be that the relevant data are rare for other reasons, for instance, because they
are stylistically extremely marked. Not ﬁnding the relevant example is suggestive
but not conclusive. Recall from the discussion in Chapter 1, section 2.3.2 that the
fact that we have never seen a black swan would not mean that they cannot exist.
To check the validity of a hypothesis we can also try to conduct an experiment.
To do this, we create our own examples, corresponding to the pattern in (40e) and
check their status. We either simply create new sentences out of the blue that illustrate
the pattern that we are interested in, or we modify the examples that we are
currently working on. Find the NP complement in (42) and place it to the immediate
left of the verb – is the result good or bad? Try the same using our test sentence.
What do you conclude? It seems to be the case that in English the complement of
the verb cannot be inserted to the immediate left of the verb.
(44) a *One of the most controversial takeovers in British sporting history was
[VP [VP [NP a government decision] awaiting] [NP last night]].
b *One of the most controversial takeovers in British sporting history was
[VP last night [VP [[NP a government decision] awaiting]]].
c *The customer in the corner will [VP [VP [NP the drinks] order] before the
meal].
We decide that such sentences “are not English.” The word orders illustrated here
are not allowed by the structure of the language. The grammar of English does not
allow the structure in which the object occurs to the immediate left of the verb.
When an unacceptable structure is ruled out because the grammatical system of the
language does not allow it then we call it ungrammatical. In the preceding discussions
we have often labeled strings as unacceptable, without saying that they
were ungrammatical. This was mainly because at the point of discussion we were
simply interested in observing that these strings were not correct. When we say that
an unacceptable string is ungrammatical, we suggest an explanation for why it is
unacceptable. In what follows we will use the labels unacceptable and ungrammatical
indiscriminately. This is because we are interested in unacceptable sentences
precisely because they give us an insight into the grammar.
2.3.4 BASE POSITIONS AND MOVEMENT
While adjuncts may apparently be left-adjoined and right-adjoined to the VP, the
complements of V are inserted to the immediate right of V.17
Does this mean that a
17
Exercise 18.
TSC02 07/18/2005, 03:19PM100
Diagnostics for Syntactic Structure 101
direct object of a verb will always be found to the immediate right of the verb?
Recall the examples in (10), repeated here as (45):
(45) a The British one I have been using since I was about nine. (Guardian,
22.3.2002, p. 3, col. 1)
b The news, when it comes, he seems to take well enough. (Guardian, G2,
26.7.2002, p. 2, col. 1)
In these examples the objects the British one and the news do not occupy a position
next to the selecting verb. Restore these “dislocated” objects to their canonical position.
Could the objects appear in a position immediately preceding the verb?
(45) c I have been using the British one since I was about nine.
d When it comes, he seems to take the news well enough.
Once again you would not be able to put the object in a position to the left of the
verb.
(45) e *I have been the British one using since I was about nine.
f *When it comes, he seems to the news take well enough.
We have to allow for constituents such as the direct object to occupy various
positions in the sentence. On the other hand, the object is closely related to the verb.
We proposed that the object must ﬁrst combine with V to form a core VP before
adjoined constituents are added to the VP by the assembly process. One possible
way of thinking about the sentence-initial position of the object illustrated above is
to say that the object is ﬁrst inserted in the position to the right of V but that it
may subsequently be moved away from that position. To allow for the construction
of sentences such as (45a, b), while at the same time excluding patterns such as
(45e, f), we will have to restrain the landing site of a moved constituent: movement
of constituents apparently targets designated positions.18
2.4 Deductive approaches
2.4.1 HEAD AND PROJECTION
So far we have encountered constituents belonging to various categories: NP, VP,
AP, PP. The label of the constituents (NP, VP, etc.) was determined by the category
of the head (N, V, etc.). Let us formalize this procedure by explicitly deﬁning the
category of the constituent in terms of the category of its head:
18
We turn to some aspects of this issue in Chapter 5.
TSC02 07/18/2005, 03:19PM101
102 Chapter 2
(46) a Head hypothesis
The category of a constituent is determined by the category of its head.
(46a) is a general hypothesis about structure: it applies to all the material we have
discussed so far and, being general, it will also apply to new material. If we discover
a new type of head X, then by (46a) we deduce that this head will project a constituent
of type XP:
(46) b XP
. . .X. . .
(46) is thus not simply a summary of our ﬁndings so far. It goes beyond the
observations and generalizes what we have found to all constituents. The effect
of the generalization is to restrict the way we will analyze any new data we may
come across in future; following (46a) we assume that all structure is headed. In
postulating (46a), we draw the bounds of possible representations more narrowly.
Based on (46a) we predict, for instance, that there will never be any structures of
the type in (46c):19
(46) c XP
. . .Y. . .
*
This also means that if we did come across a constituent that seems to require
representation (46c), we should rethink our analysis of that particular constituent
or, alternatively, we would have to rethink our theory.
2.4.2 BINARY BRANCHING
Let us compare some of the tree structures discussed so far, focusing simply on the
geometry of the branches, the lines that compose the trees. For instance, let us
compare the way these lines are organized in (5e′), repeated here as (47a), with the
organization in (39b), repeated here as (47b).
19
There is a huge literature concerning the general format for syntactic structure known as
“X-bar theory.” For ﬁrst proposals see Chomsky (1970), see also Jackendoff (1977) for a ﬁrst
fully elaborated account. For a critical evaluation of various implementations see also Kornai
and Pullum (1990).
TSC02 07/18/2005, 03:19PM102
Diagnostics for Syntactic Structure 103
(47) a S
VP PP
NP
AuxNP
V
will the drinks before the mealorderThe customer in the corner
b S
PP
NP
VP
will the drinks before the mealorderThe customer in the corner
NPV
CONSTITUENT
Aux VP
(47a) and (47b) display different types of combinations. In (47a) the root node S
splits up into four separate branches. S dominates four immediate constituents:
(i) the subject NP, (ii) the auxiliary, (iii) the VP, and (iv) the PP. The same tree
also contains instances of a node dominating just two constituents; for instance,
the VP node dominates the verb and the object NP. If we removed the time speciﬁcation
from the sentence, we would end up with yet another type of composition,
since the root S would then dominate only three constituents. (47a) suggests
that the types of nodes in a tree are unconstrained: we can have binary branching
nodes (V + NP in (47a)), or ternary branching nodes (if we omit the time PP from
(47a)), or even four-way branching nodes (S in (47a)). This wealth of options is in
sharp contrast with the restricted branching pattern in (47b): here all the nodes
dominate just two constituents. (47b) only has binary branching nodes. (47b) is
more restrictive.
Let us think about the branching patterns in more general terms. What we are
concerned with in this book is “syntax,” that is: putting units (here units of language)
together, or assembling units. It is trivially true that in order to assemble
elements, you need at least two of them. In fact, in order to assemble elements, you
do not need more than two of them. If you do have more than two elements to put
together, you can still use a binary combination system to assemble them. Consider
TSC02 07/18/2005, 03:19PM103
104 Chapter 2
(47c). We have ﬁve elements to assemble. We can assemble these ﬁve elements into
a simple structure, using the binary branching system:
(47) c List of elements: A, B, C, D, E
d X
XA
XB
XC
ED
The tools needed to form (“generate” to use a technical term) (47b/d) are more
restrictive than those needed to generate (47a). Or, putting it differently, (47a) is
more permissive. In (47b/d) we only allow combinations of two constituents; in
(47a) we also have admitted combinations of three or of four. (47b/d) is preferable
on theoretical/conceptual grounds, because it requires fewer assumptions about what
is possible. By restricting branching patterns to binary branching as in (47b), we
deﬁne more narrowly what is possible in language, we narrow down the options for
combining constituents to combinations of just two constituents, or binary combinations.
(47b/d) thus once again will correspond better to Einstein’s description
of the goals of scientiﬁc research given in Chapter 1, section 1.2.3:20
The grand aim of all science is to cover the greatest possible number of experimental
facts by logical deduction from the smallest number of hypotheses or axioms. (Einstein
1954, cited in Abraham et al. 1996: 4)
For (47a) we need at least three types of branching. For (47b) we only need binary
branching: each constituent grouping, each syntactic unit, combines two constituents.
Thus a theory including (47b) is more elegant and economical than one that
includes (47a), the latter having more tools available.
(47b) is supported both by empirical evidence (the evidence provided by applying
the constituency tests to data as discussed in this chapter) and by theoretical arguments.
Let us therefore postulate that all linguistic structure is based on binary
branching.21
From now on, when we try to derive structures of sentences we will
20
See also the discussion of “Ockham’s Razor” in Chapter 1, section 1.2.3.
21
Exercise 11.
TSC02 07/18/2005, 03:19PM104
Diagnostics for Syntactic Structure 105
work with the hypothesis in (48).22
Once again, (48) is a theoretical hypothesis
which, though inspired by a set of empirical data, goes beyond these data.
(48) Binary composition hypothesis
All syntactic structure is binary.
From now on, our analysis will be guided by (48), a theoretical principle formulated
to restrict our theory. When theoretical interest guides the way we look at data we
adopt a deductive methodology. In such an approach the hypotheses that constitute
a theory form the basis of the deduction, but again what we deduce will be a general
principle and once again this principle will be tested by observation, or, if possible,
by experiment. Our theory about syntax, about putting words together to form
sentences, proposes that all syntax, all “putting together,” is ruled by binary
combinations; all syntactic branching will be binary.23
(48) says that syntax is the
combination of two elements. A constituent is formed by the fusion or the merger
of two constituents; that newly formed constituent can be merged with yet another
constituent to form another constituent etc. Concretely, the derivation of our
example sentence proceeds as follows:
(49) a assemble core VP = V + NP
b assemble augmented VP = VP + PP
c assemble “constituent” = Aux + VP
d assemble sentence = NP + constituent
Recall that the assembly of the core VP (49a) is different from the assembly of the
augmented VP (49b). In the former case we assemble a head and an element that is
selected by that head to form a phrase. In the second type of assembly we combine
two phrases which are already themselves fully formed. We have seen that we need
both types of procedures in order to differentiate obligatory complements and
optional adjuncts.
2.5 A short note on binary branching and word structure
We have so far only dealt with the assembly of sentences (and only a small sample
of possible sentences), but we might generalize our proposals to say that the binary
branching hypothesis in (48) is a general hypothesis about structure in language.
In other words, the hypothesis would be that all linguistic units are built up by
combining two elements.
22
As brieﬂy discussed in Exercise 7, co-ordinated structures raise a problem for the binary
branching hypothesis. For some sophisticated proposals on how to deal with co-ordination see
Camacho (2003), Goodall (1987), van Oirsouw (1987).
23
In generative linguistics, the binary branching hypothesis was elaborated by Richard Kayne
(1984).
TSC02 07/18/2005, 03:19PM105
106 Chapter 2
Words are smaller units of language than sentences; words are themselves
combined into phrases to form sentences. Let us brieﬂy look at English words.
Does the internal structure of words conform to the binary branching format?
Consider the underlined words in (50). Do they have internal structure? What are
their components? How are the components assembled?
(50) a Anthrax was hard to weaponize. (Guardian, 15.10.2001, p. 5, col. 2)
b [The powder] was not, however, “weaponized” – genetically modiﬁed to
be antibiotic-resistant. (Independent, 22.10.2002, p. 5, col. 3)
c Among those countries are nations that have tested the weaponization of
those chemical and biological agents. (Guardian, 10.10.2001, p. 2, col. 8)
The root common to the words weaponize, weaponized, and weaponization is the
noun weapon. On the basis of the noun weapon we create the verb: weaponize
(50a), meaning “turn something into a weapon,” “use something as a weapon.”
This verb can be inﬂected: in (50b) we ﬁnd its past participle weaponized. In (50c)
we create a noun, weaponization, on the basis of the verb weaponize. It is clear
that in the examples illustrated here, the process of creating new words is based on
binary branching: it is summarized in the structures in (51):
(51) a V
ize
N
weapon
b N
ation
V
izeN
weapon
Consider the underlined words in the following extracts. Describe the process by
which they are formed:
(52) a The current State Department reaction to criticism by the hawkish commentator
Newt Gingrich offers examples of weak and strong vituperation.
(New York Times, 28.4.2003, p. A25, col. 5)
b [He was] conscious of the discomfort at Washington’s growing hawkishness.
(Guardian, 26.4.2003, p. 6, col. 3)
Discuss the interpretation of the adjective undoable in the following examples:
TSC02 07/18/2005, 03:19PM106
Diagnostics for Syntactic Structure 107
(53) a This task is undoable: we don’t have the people and we don’t have the time.
b This knot is undoable; all we need is a ﬁne needle and a lot of patience.
The adjective undoable has two interpretations: ‘which cannot be done’ and ‘which
can be undone’. In the same way that sentences can have different interpretations as
a result of being structurally ambiguous, the ambiguity of undoable can be related
to its internal make-up, its structure:
(53) c [A un [A [V do] able]] vs. [A [V un [V do]] able]
By adopting the binary branching hypothesis for all structural relations we have
elaborated a simpler theory than if we allow for all types of branching. We need
only one type of combination: structure is formed by putting together (“merging”)
two constituents. As mentioned, a theory that uses only binary branching is to be
preferred over a theory that has both binary branching and ternary branching.24
3 Speciﬁers
3.1 Noun phrases
3.1.1 A BINARY BRANCHING STRUCTURE FOR NPS
Consider the NP the teacher of English. We will try to develop a structural representation,
using as our guidelines the theoretical concepts elaborated so far.25
In particular, we assume that all structure is headed and all structure is binary
branching. This means that, for instance, representation (54a) will be excluded on
theoretical grounds.
(54) a NP
N PP
of English
Det
teacherthe
24
Exercises 16, 17.
25
This is a ﬁrst approximation of the structure of NP. The structure would have to be revised in
a number of ways in the light of further developments of the theory. We do not go into this
area but refer the reader to the existing literature. For a ﬁrst introduction see Haegeman and
Guéron (1999: chapter 4). For more detailed and more advanced discussion see also Bernstein
(2001).
TSC02 07/18/2005, 03:20PM107
108 Chapter 2
The label “Det” in (54a) stands for “determiner.” The term is used provisionally as
a cover term for articles such as the, a and for elements that seem to occupy the
same position in the NP, such as demonstratives.26
In principle, either (54b) or (54c) would respect the binary branching hypothesis;
we again use the label constituent as a provisional label:
(54) b cNP
CONSTITUENTDet
N PP
of Englishteacherthe
NP
PPCONSTITUENT
N
teacher
Det
the of English
Which of these should we adopt? In order to choose between the two structures, we
can use theoretical considerations as well as empirical ones. Ideally, the two kinds
of argumentation should converge and lead to the same conclusion. Consider ﬁrst
the noun teacher, which is morphologically related to the verb teach. In the NP the
teacher of English, the relation of the noun teacher to the PP of English is similar to
that between the verb teach and a direct object in the sentence, cf. (54d):
(54) d S
CONSTITUENT
English
NP
teachMy sister
VPAux
NPV
will
Tentatively, we could conclude that this analogy favors structure (54b), in which
the N teacher ﬁrst merges with the PP of English. The structure would be one in
which the determiner the is outside the resulting constituent [N + PP]. Is there any
evidence that the unit teacher of English is a constituent? And if so, what kind of
constituent would it be? Do the data in (55) shed light on this question?
26
Below we will replace this label by a more adequate label for the position occupied by such
elements. For the importance of the determiner in the structure of the NP see Abney (1987),
Bernstein (2001), and the references cited. These publications are advanced and should only be
tackled after you have ﬁnished this book.
TSC02 07/18/2005, 03:20PM108
Diagnostics for Syntactic Structure 109
(55) a This teacher of English has arrived today but that one arrives only tomorrow.
b John’s teacher of English is British and Mary’s is American.
(55a) contrasts two NPs: this teacher of English versus that one. The opposition is
expressed by the demonstratives this and that. The component teacher of English is
not contrasted and in the second part of the co-ordinated structure the string teacher
of English is replaced by the word one. The fact that one word, one, can substitute
for a string of words suggests that the string, teacher of English, is a constituent, a
conclusion compatible with representation (54b), though not with (54c).27
The same conclusion is reached if we examine (55b). Here the non-contrastive
part of the NP is deleted:
(55) c John’s teacher of English is British but Mary’s ∅ is American.
If we assume that ellipsis affects constituents (as argued in section 1.6), then again
this example is evidence for the structure in (54b).
We have argued for the structure in (54b) on the basis of one-substitution and
of ellipsis. If we adopt structure (54b), the question arises as to the nature of the
constituent N + PP. How does it relate to the determiner? If we pursue the
analogy with VP, i.e. the hypothesis that N is the head of the constituent [teacher
of English] in the way that V is the head of [VP teach English], then we ought
to conclude that the constituent N + PP is an NP. On the other hand, it is not an
ordinary NP. Teacher of English does not have the same distribution as other NPs:
it cannot function as an object (56a), nor can it function as a subject (56b).
(56) a *I met teacher of English.
b *Teacher of English arrived late today.
What minimal “correction” would save the unacceptable strings in (56)? To enable
the string teacher of English to freely occupy subject and object positions we
need to add a determiner, such as a, the, this, that. How could we characterize the
interpretive effect of adding these elements? How do the teacher of English and
teacher of English differ in meaning?
(57) a I met a/the/this/that teacher of English.
b A/the/this/that teacher of English arrived late today.
The string teacher of English denotes a property of the entity that we are referring
to. In other words, we attribute to that entity the characterization that he/she ‘teaches
English’. To pick out an entity with the relevant property (to “refer” to such an
entity), we need to use a determiner. By inserting the indeﬁnite article a we signal
27
For discussion of one substitution see also Panagiotidis (2002, 2003).
TSC02 07/18/2005, 03:20PM109
110 Chapter 2
that there is at least one entity that has the property ‘teacher of English’, and that
this entity is being introduced in the discourse. By inserting the deﬁnite article the
we signal that one such entity with the relevant properties is already familiar in the
discourse: speaker and hearer can identify the entity of the type ‘teacher of English’
we are talking about. In other words a and the serve to specify reference: these
elements help us pick out the entity to which we attribute the property ‘teacher of
English’. Observe that we only use one such determiner:
(57) c *a this teacher of English
Determiners specify the reference of the NP. We insert them in a position labeled
speciﬁer, as in (58). Because the combination teacher of English cannot on its own
function as a complete NP, while at the same time it is a constituent headed by an
N, we will label it N′ (N-prime), corresponding roughly to a partial NP.28
(58) NP
N′speciﬁer
the/a/this/that
PPN
of Englishteacher
At this point, the speciﬁer position is new to our structures. So far we have been
operating without it. If we do introduce this concept in the NP, we should show
that it is really required. We should, for instance, show that a speciﬁer is distinct
from a complement and from adjoined constituents. Once we have motivated the
speciﬁer position in the NP, we have to address the question whether there could be
a similar speciﬁer position in other constituents, such as VP and S. If the answer to
that question is positive, we will want to know which elements occupy the speciﬁer
positions. If the answer is negative, we have to determine why speciﬁers are restricted
in their distribution.
3.1.2 SPECIFIERS IN THE NP
3.1.2.1 Speciﬁer–head agreement
The speciﬁer position in the projection of the noun has some properties that set it
apart from other constituents in the NP. First, look at the form of the demonstratives
in the following English examples: how do we account for the different forms?
28
The intermediate level N′ is also sometimes referred to as “N-bar.”
TSC02 07/18/2005, 03:20PM110
Diagnostics for Syntactic Structure 111
(59) a this teacher of English
b these teachers of English
c that teacher of English
d those teachers of English
As you see, the form of the demonstrative depends on the form of the head noun: a
singular N teacher goes with the forms this/that, a plural N teachers with the forms
these/those. The speciﬁer agrees with the head in terms of the number feature. The
same pattern is found in French:
(60) a ce professeur d’anglais
this (sg) professor (sg) of English
b ces professeurs d’anglais
these (pl) professors (pl) of English
French singular nouns also match with the demonstrative for the feature gender, as
shown by the following examples:
(60) c ce cours d’anglais
this (masc) course (masc) of English
‘this English course’
d cette leçon d’anglais
this (fem) lesson (fem) of English
‘this English lesson’
Cours (‘course’) is a masculine singular noun; it is accompanied by the masculine
singular demonstrative ce. Leçon (‘lesson’) is a feminine singular noun; it is accompanied
by the feminine singular demonstrative cette.
French articles also match a singular head noun in gender (60e, f) and number (60g):
(60) e un cours d’anglais vs. une leçon d’anglais
a (masc-sg) course of English a (fem-sg) lesson (fem-sg) of English
‘an English course’ ‘an English lesson’
f le cours d’anglais vs. la leçon d’anglais
the (masc-sg) course of English the (fem-sg) lesson (fem-sg) of English
‘the English course’ ‘the English lesson’
g les cours d’anglais vs. les leçons d’anglais
the (pl) courses (pl) of English the (pl) lessons (pl) of English
‘the English courses’ ‘the English lessons’
TSC02 07/18/2005, 03:21PM111
112 Chapter 2
The constituent in the speciﬁer position of the NP agrees with the head noun. Let us
formulate the hypothesis that agreement relations are realized through speciﬁer–
head relations. Consider the following examples: how is agreement encoded in the
sentence?
(61) a The teacher of English is arriving today.
b The teachers of English are arriving today.
Agreement obtains between the subject NP, the teacher(s) of English, and the
inﬂected auxiliary, is/are. We return to the relevance of this observation in section
3.1.2.3, and also in Chapter 3.
3.1.2.2 Prenominal genitives
Examine the underlined NPs in the following extracts. The head nouns are preceded
by genitive forms of NPs. Can the genitive NP preceding the head N be replaced
by a pronoun? How could we characterize the semantic relationship of the prenominal
genitive NP to the associated N? Can the head nouns of the underlined NP,
which are preceded by the genitive NP, also be preceded by a determiner that
speciﬁes the reference of the head noun (a demonstrative or an article)?
(62) a One of the most controversial takeovers in British sporting history was last
night awaiting a government decision after the Monopolies and Mergers
Commission delivered its verdict to ministers on whether Rupert Murdoch’s
bid for Manchester United Football Club should be allowed. (Adapted
from Guardian, 13.3.1999, p. 1, col. 1)
b Instead, the reminder of the best of Labour came in Estelle Morris’s speech.
(Adapted from Guardian, 4.10.2002, p. 7, col. 1)
The prenominal genitive NPs Rupert Murdoch’s and Estelle Morris’s can be replaced
by pronouns (his, her). We cannot add a determiner to the underlined NPs:
(63) a *the Rupert Murdoch’s bid for Manchester United Football Club
*Rupert Murdoch’s the bid for Manchester United Football Club
b *the Estelle Morris’s speech
*Estelle Morris’s the speech
If we wanted to use a determiner then we would have to remove the prenominal
genitive NP:
(63) c the bid for Manchester United Football Club by Rupert Murdoch
d the speech by Estelle Morris
In our examples, a prenominal genitive NP and the determiner specifying the
reference of the head noun cannot co-occur. Either we use a prenominal genitive
TSC02 07/18/2005, 03:21PM112
Diagnostics for Syntactic Structure 113
or we use a determiner. The prenominal genitive and the determiner are said to be
in complementary distribution. To rule out that a genitive NP preceding a noun
co-occurs with a determiner related to the same noun, we could propose that the
genitive NP and the determiner are inserted in the same position. We have proposed
that determiners occupy a particular position of NP labeled speciﬁer. Let us postulate
that (i) each syntactic position can contain only one constituent and (ii) that
there is just one speciﬁer position in the NP. If prenominal genitive NPs occupy the
speciﬁer position of NP, we correctly predict that an NP contains either a prenominal
genitive or a determiner.29
In the examples above the nouns bid and speech can be related to the verbs bid
and speak; the prenominal genitive NPs relate to the head nouns like subjects relate
to verbs in a sentence:
(64) a Rupert Murdoch will bid for Manchester United.
b Estelle Morris will speak at the conference.
(64c) contains an NP related to our test sentence (5a). Again, the subject NP of the
original test sentence corresponds to the prenominal genitive NP:
(64) c I am waiting for [NP [NP the customer in the corner’s] order of the drinks].
Let us assume then that prenominal genitive NPs occupy the speciﬁer position of
the containing NP. Since possessive pronouns such as his, her, their, etc. replace
prenominal genitives, we assume they also occupy the speciﬁer of NP.
(65) NP
N′speciﬁer
Rupert Murdoch’s
his
this/that
a/the
PPN
for Manchester Unitedbid
Recall also that the speciﬁer position can be a locus for agreement relations. In
English, genitive NPs or possessive pronouns do not agree with the head noun.
French possessive pronouns agree with the head noun, as shown in (66):
29
For data that challenge this proposal see Exercise 20. See Haegeman and Guéron (1999:
chapter 4) for introductory discussion; for advanced discussion see Abney (1987), Bernstein
(2001), and the references cited there.
TSC02 07/18/2005, 03:21PM113
114 Chapter 2
(66) a mon résumé du texte
my (masc-sg) summary (masc-sg) of-the text
‘my summary of the text’
b mes résumés du texte
my (pl) summaries (pl) of-the text
‘my summaries of the text’
c ma description du texte
my (fem-sg) description (fem-sg) of-the text
‘my description of the text’
d mes descriptions du texte
my (pl) descriptions (pl) of-the text
‘my descriptions of the text’
We observe that the speciﬁer of the NP unites two properties: agreement (3.1.2.1)
and subjecthood (3.1.2.2). Both agreement and subject are concepts that we also
operate with at the level of sentences. What we will try to do below is to elaborate
an approach to the sentence in which agreement and subjecthood are tied to a
speciﬁer position.30
3.1.2.3 Questions about the structure of sentences
From the discussions above we conclude that the speciﬁer position of the NP
has some special properties: (i) it has agreement properties, and (ii) it has subject
properties. In French, these two properties coincide in prenominal possessive
pronouns.
We have also mentioned that at the level of sentences, agreement is realized
between the subject NP and the auxiliary. This is illustrated in (67a) and (67b): has
is singular, have is plural.
(67) a The customer has ordered the drinks.
b The customers have ordered the drinks.
It would be most economical to try to characterize agreement properties in a uniform
way across sentences and NPs. If the speciﬁer of NP is the privileged position for
realizing agreement on the determiner, and if it also encodes “subjecthood,” then
ideally when the properties of agreement and subjecthood coincide at the sentence
level, we would also like to associate them with a speciﬁer position. Compare the
structure of the NP as given in (65) with (68), the structure of the sentence elaborated
so far.
30
Exercises 1B, 11. See Chapter 4, section 3.2.2.1, and Exercise 6, for an illustration of different
agreement patterns.
TSC02 07/18/2005, 03:21PM114
Diagnostics for Syntactic Structure 115
(68) S
CONSTITUENTNP
The customer
The customers
VPAux
the drinks
the drinks
ordered
ordered
NPV
has
have
If agreement relations are established in speciﬁer positions, then it would be an
important move forward in our theory if we could say that the subject NP occupies
a speciﬁer position with respect to the auxiliary. This would amount to extending
the projection schema for NPs to the sentence. How could this be done? We would
need to identify a head, X, which takes a phrase as its complement, with which it
combines to form X′, and whose speciﬁer is the subject.
(69) XP = sentence
X′speciﬁer
= subject
YPX
Try to superimpose the general format (69) on the sentence structure in (68). One
way of aligning the two representations is to designate the auxiliary as the head of
the sentence, and to identify the node which we had labeled provisionally as “constituent”
as the intermediate projection, dominating the auxiliary and the VP.
(70) a XP (= AuxP)
X′ (= Aux′)speciﬁer
The customer
VPX (= Aux)
the drinksordered
NPV
has
= S(entence)
= CONSTITUENT
This representation means that we consider the VP as the complement of the auxiliary.
Can we provide any motivation for this view? Recall that verbs select their
TSC02 07/18/2005, 03:21PM115
116 Chapter 2
complements, in that, for instance, certain verbs select an NP, while others select a
PP, etc. This is illustrated by means of the paired examples in (71): expect selects an
NP complement; wait, on the other hand, requires a PP complement.
(71) a Mary is expecting some news.
*Mary is expecting for some news.
b *Mary is waiting some news.
Mary is waiting for some news.
Consider the sets of sentences in (72). In what sense could auxiliaries be said to
select the VP?
(72) a Mary is [VP waiting for some news].
*Mary is [VP wait for some news].
*Mary is [VP waited for some news].
b Mary has [VP waited for some news].
*Mary has [VP waiting for some news].
*Mary has [VP wait for some news].
c Mary will [VP wait for some news].
*Mary will [VP waiting for some news].
*Mary will [VP waited for some news].
We might say that the choice of auxiliary determines the form of the VP: progressive
be selects a VP headed by a present participle (72a), perfect have selects
a VP headed by a past participle (72b), a modal auxiliary selects a VP headed by
an inﬁnitive (72c). In other words, the VP can be seen as being selected by, i.e. as
being the complement of, the auxiliary.
For sentences with auxiliaries we could propose the structure in (70b). The constituent
that combines with Aux′ is by deﬁnition a speciﬁer: this constituent is
inserted in a position deﬁned as the speciﬁer position. This means that we do not
need to add the label “speciﬁer” in the tree. The position of the subject NP in (70b)
corresponds to the speciﬁer position. We only write down the category of the element
that occupies the speciﬁer position, here NP.
(70) b AuxP
Aux′NP
The customer
VPAux
the drinksordered
NPV
has
TSC02 07/18/2005, 03:21PM116
Diagnostics for Syntactic Structure 117
This raises the question as to how to deal with sentences without auxiliaries. We
turn to this issue in the next chapter.31
3.2 Adjuncts and NPs
In addition to the core constituents of the VP, verb and complement, we proposed
that we can also add modiﬁers such as, for instance, temporal PPs. Such modiﬁers
are not selected by the verb, their categories are not determined by the verb heading
the VP. For instance, a temporal speciﬁcation can be expressed in a number of
different ways:
(73) a The customer in the corner will order the drinks [PP before the meal].
b The customer in the corner will order the drinks [S when his guests arrive].
c The customer in the corner will order the drinks [AdvP later].
d The customer in the corner will order the drinks [NP this afternoon].
To integrate modiﬁers into the structure, we have introduced the concept of (left- or
right-) adjunction, creating the augmented VP. If adjunction is available for building
a VP, then the question arises if it is also available for an NP. Consider the underlined
constituent in the following passage. What is its head? What is its category?
Motivate your answer.
(74) a Marks and Sparks advice to its customers, as printed on all their shopping
bags, is: To avoid suffocation, keep away from children. (Adapted from
Guardian, 18.2.2002, p. 13, letter to the editor from Dick Brown, Arnside,
Lancs)
The relevant string is an NP. To show that this is the case you can try using the NP
as subject or object of a sentence. You can also replace the NP by a pronoun:
(74) b Marks and Sparks are reprinting all their shopping bags.
Marks and Sparks are reprinting them.
c All their shopping bags have disappeared.
They have disappeared.
Based on the data in (75), what is the category of their shopping bags?
(75) a Marks and Sparks advice to its customers, as printed on their shopping
bags, is: “To avoid suffocation, keep away from children.”
31
Another question is how to deal with sentences containing more than one auxiliary. See
Exercises 12 and 13, and Chapter 3, section 4. If adjunction is generally available as a way
of augmenting a constituent we would expect to be able to adjoin constituents to a sentence
(= AuxP). See Exercise 15.
TSC02 07/18/2005, 03:21PM117
118 Chapter 2
b Marks and Sparks are reprinting their shopping bags.
Marks and Sparks are reprinting them.
c Their shopping bags have disappeared.
They have disappeared.
We conclude that both all their shopping bags and their shopping bags are NPs. All
is not essential to form an NP, it provides additional information. We could add it
onto the NP structure by left-adjoining it to the “core NP.”32
(76) NP
NPQuantiﬁer
N′speciﬁer
[NP [Q all] [NP their shopping bags]]
[NP [Q all] [NP the shopping bags]]
N
their
the
all
all
shopping bags
shopping bags
3.3 Questions about the verb phrase
Recall that an essential ingredient of scientiﬁc work is doubt.33
We must always
remain aware that our analyses are hypotheses. New insights or new developments
in research may well mean that we must go back on what we think we know and
revise earlier proposals. Having elaborated a hypothesis for the structure of the NP
and applied it to the sentence, we have to reconsider our earlier hypothesis about
the VP.
So far we have elaborated representations for the structure of (i) the VP, (ii) the
NP, (iii) the sentence. We have used our ﬁndings for the structure of one of these
constituents as a guideline to examine another constituent, but in the discussion we
have left some areas unattended. In this section we reconsider the NP and VP and
we will try to see to what extent, if at all, the structures are similar or different.
Recall that we are interested in reducing the differences between the structures to a
minimum. This approach will lead to a simpler theory.34
Our theory is going to be
simpler if we can say that NPs and VPs have the same structure. If they don’t, we
have to introduce two different structures, giving rise to a more permissive theory
32
Exercise 13. For a more careful analysis see Shlonsky (1991).
33
Chapter 1, section 1.2.4.
34
For the role of simplicity see Chapter 1, section 1.2.3, and this chapter, section 2.4.2.
TSC02 07/18/2005, 03:22PM118
Diagnostics for Syntactic Structure 119
and also a more complex theory. In addition, the differences in structure that we
have to postulate have to be explained by relating them to some other principle(s).
In the discussion of the structure of the NP, we identiﬁed four components of
structure: the head, the complement, the speciﬁer, and the adjunct. Generalizing this
format for all syntactic structures, a syntactic constituent would be organized according
to the format in (77):35
a head, X, combines with a complement to form a
partial constituent, X′, and this combines with the speciﬁer to form the complete
constituent (XP). In the literature, the term intermediate projection is used to refer
to what we called a partial constituent, and the term maximal projection is used for
the complete constituent, including the speciﬁer. Adjoined constituents are satellites
to the maximal projection; they are themselves maximal projections. For example,
in our example (5a) the adjunct before the meal is a full PP.
(77) a XP
XPadjunct
X′speciﬁer
complementX
head
phrase, maximal projection
phrase, maximal projection
intermediate projection
However, the structure we arrived at for the VP was different in that we did not
have the level of speciﬁer:
(77) b VP
VPadjunct
V complement
This difference in structure is not really satisfying. A number of questions arise. (i)
Why is the combination of V + complement represented as a maximal projection
(VP), while that of N + complement is represented as an intermediate projection
(N′)? The answer to this question could be that this is because the VP lacks a
speciﬁer. This leads to the next question. (ii) Why does the representation for NP
include a speciﬁer, while that for VP does not?
Because we had empirical grounds to postulate representation (77a) for the NP, it
seems natural to maintain it and to assume that, in principle, constituents can have
35
(77a) replaces the earlier (41). How do the structures differ?
TSC02 07/18/2005, 03:22PM119
120 Chapter 2
a speciﬁer. It would then be desirable to generalize this format to the VP (and
indeed to other categories). What would the structure of the VP be like if we
assumed that it also had a speciﬁer? Based on the structure of the NP in (77a), we
could replace (77b) by (77c), which we apply to our example sentence:
(77) c VP
VP PP
speciﬁer V′
V NP
order the drinks before the meal
??
What kind of constituent could occupy the speciﬁer of VP? So far, we have not
come across any candidates. We cannot consider left-adjoined constituents as
speciﬁers. First of all adjoined constituents, unlike speciﬁers, are freely added to a
complete constituent and they may occupy various positions. Moreover, as shown
in section 2.3.2, a VP may have multiple adjuncts (cf. examples (43g, h, i)), but we
argued in section 3.1.2.2 that there is only one speciﬁer position in the NP. If, on
the other hand, we decided that there is no speciﬁer in the VP, then we would have
to explain why the VP lacks a speciﬁer.
Recall also from section 3.1.2.2 that the speciﬁer position of the NP could be
occupied by a prenominal genitive NP or by a possessive pronoun, both of which
have a subject-like relation to the head N. Let us explore this observation: by
analogy with the NP, the speciﬁer of the VP ought to be the subject NP. However,
in assuming that the speciﬁer of the VP is the subject we create a conﬂict with the
structure of the sentences elaborated above, in which the subject is located outside
the VP. Let us examine why we had concluded that the subject was located outside
the VP. One argument for assuming that the subject is outside the VP was that the
subject was seen to precede the auxiliary, which we considered as a separate unit
that forms a constituent with the VP. Secondly, we have also seen that the subject
precedes any adjuncts that are left-adjoined to the VP.36
The examples in (78)
illustrate both these points:
(78) a [NP The customer in the corner] will [VP order the drinks before the meal].
b [NP The customer in the corner] will [VP deﬁnitely [VP order the drinks before
the meal]].
36
Cf. section 2.2.1.
TSC02 07/18/2005, 03:22PM120
Diagnostics for Syntactic Structure 121
We will return to the relation between the subject and the VP in Chapter 4, but
before doing so we will complete the representation of the structure of the sentence
in Chapter 3.
Another problem that came out of the discussion at the end of sections 2.2 and
3.2 is the question how to deal with sentences without auxiliaries such as (39c),
repeated here as (79). We will tackle this question in the next chapter.
(79) The customer in the corner ordered the drinks before the meal.
4 Summary
In this chapter we have ﬁrst elaborated a number of diagnostic tests for discovering
sentence structure. Units of structure, or constituents, can be detected by manipulating
the sentences in which they occur. We can try to replace strings of words by a smaller
unit (substitution), we can make them the target of a question (question formation),
we can move them around (movement), we can delete them (ellipsis), we can make
them the informational focus of the sentence (clefting and pseudo-clefting), and
ﬁnally, constituents can be co-ordinated. By means of these tools, we have decomposed
English sentences and demarcated their core constituents.
The constituents of a sentence are formed around a core constituent, their head.
For instance, an NP has a noun as its head. We have examined two competing
hypotheses about the structure of the VP. According to one hypothesis, the VP
contains the verb and the auxiliary (or the auxiliaries) of the sentence; according to
the second, the VP contains the verb, its complements, and its adjuncts. The second
hypothesis is both empirically and theoretically preferable. Further examination
of the VP reveals that we need to postulate a more articulated structure in which
we distinguish two hierarchical levels. This allows us to distinguish a core VP, the
combination of the verb with its complement, from the augmented VP, a larger
constituent which combines the core VP with adjuncts.
While discussing the structure of the VP we elaborated the binary branching
hypothesis. This is a theoretical proposal to the effect that all syntactic constituents
are the result of combining or merging two constituents, and that constituents are
hierarchically organized around a head. The head ﬁrst combines with its complement.
The resulting constituent then combines with adjuncts.
In our examination of the structure of the NP we have revealed the need to
identify an additional level of representation to allow for the creation of a speciﬁer
position. A noun head ﬁrst combines with its complement to form N′, an intermediate
projection of an NP. This intermediate projection combines with a
prenominal element in the speciﬁer position to form the fully completed NP, the
maximal projection.
The speciﬁer of the NP hosts a determiner element, a prenominal possessive
pronoun, or a prenominal genitive NP. We assume that each syntactic position can
TSC02 07/18/2005, 03:22PM121
122 Chapter 2
contain only one constituent. We also assume that there is only one speciﬁer position
per NP. This means that an NP will contain either a determiner, or a possessive
pronoun, or a prenominal genitive NP.
The speciﬁer position of the NP is set apart by two distinctive properties: the
constituent contained in it sometimes agrees with the head noun, and it sometimes
seems to have subject-like properties. This observation has led us to the hypothesis
that the subject of the sentence occupies a speciﬁer position. In order to implement
this hypothesis, we have elaborated a representation for sentences with auxiliaries.
According to the proposal, sentences are built around auxiliaries: the auxiliary is
the head of the sentence. The auxiliary takes a VP as its complement, forming an
intermediate projection. This intermediate projection combines with the speciﬁer
position to form the maximal projection. The subject occupies the speciﬁer position.
Given the observation that both the NP and sentence have a speciﬁer position, we
will have to address the question whether the VP also has a speciﬁer position. In
addition, given that not all sentences have an auxiliary, we will have to examine
how to apply the structure elaborated for sentences with auxiliaries to sentences
lacking auxiliaries.
TSC02 07/18/2005, 03:22PM122
Diagnostics for Syntactic Structure 123Exercises 123
Exercises
Contents
Exercise 1 Constituent structure in the NP (T, E) 123
Exercise 2 Bracketing representations (T) 126
Exercise 3 Representations of structure (T) 126
Exercise 4 Evidence for constituent structure (T) 127
Exercise 5 Focalizing a constituent by rightward movement (T, E) 128
Exercise 6 Structural ambiguity and co-ordination (T) 130
Exercise 7 Representing co-ordination (T, E) 131
Exercise 8 Structural ambiguity (T, E) 132
Exercise 9 Co-ordination (T, E) 134
Exercise 10 Realization of subject and object (T, E) 135
Exercise 11 Constituent structure in the NP (T) 137
Exercise 12 Auxiliary strings (T, L) 137
Exercise 13 Auxiliary strings (T, L) 138
Exercise 14 Attachments of adjuncts (T, L) 138
Exercise 15 Adjunction to the sentence (L, E) 141
Exercise 16 Creating new words (T) 145
Exercise 17 The internal structure of deverbal nouns (E) 145
Exercise 18 Comparative linguistics: Deriving OV orders (T, E) 148
Exercise 19 Constituency tests (T) 152
Exercise 20 The speciﬁer of NP (E) 153
Exercise 1 Constituent structure in the NP (T, E)
A Constituency tests (T)
Do the underlined strings in the following examples form a constituent? For each
string, provide at least one argument to motivate your answer.
(1) Have you read our dean’s report of the meeting?
(2) Our new dean’s wife has been invited to give a lecture.
(3) These new students of English are planning a trip to London and those are
going to Paris.
TSC02 07/18/2005, 03:22PM123
124 Chapter 2
(4) We all expected that they would wait for us after the lecture.
(5) I will meet all the students of semantics next week.
COMMENTS
In (1) the string our dean’s is a constituent: it can be replaced by a pronoun (his),
and we can also question it by means of whose. Our dean’s is the genitive of the
NP, our dean. Inside that NP, the possessive pronoun our speciﬁes the reference of
the genitive NP. The structure of the NP in (1) would be as in (6). The NP our dean
occupies the speciﬁer position inside the NP. We use the labels NP1 and NP2 to
distinguish the two NPs: NP1 is the containing NP, NP2 is the prenominal genitive
NP in its speciﬁer. We attach the genitive ending ’s to the N dean.
(6) NP1
N′NP2
N′specifier
N
our
PPN
dean’s report of the meeting
Observe that there are two NPs here: we have two nouns (report, dean), each of
which has its own projection (NP1, NP2).
B Determiners and speciﬁers (E)
In section 3.1.2.2 we accounted for the unacceptability of the examples in (63),
some of which are repeated here as (7), by postulating (i) that a position can be
occupied by only one constituent and (ii) that there is one speciﬁer position in the
NP.1
The unique speciﬁer position in the NP will then be occupied either by a
determiner or by a prenominal genitive NP.
(7) a *the Rupert Murdoch’s bid for Manchester United Football Club
b *Rupert Murdoch’s the bid for Manchester United Football Club
How would this be compatible with the observation that the examples in (8) are
acceptable?
1
For data that complicate this hypothesis see Exercise 20 below. For an introduction to complications
see Haegeman and Guéron (1999: chapter 4). For a theoretical (but more advanced)
discussion see Bernstein (2001).
TSC02 07/18/2005, 03:22PM124
Diagnostics for Syntactic Structure 125
(8) a the teacher’s books
b that teacher’s books
Representation (6) above can help to clarify the issue of the occurrence of determiners
and genitives in (8). In (8a), the determiner the is related to the noun teacher: we
can replace the string the teacher’s by the pronoun his. We can question the string
the teacher’s by means of whose. (9a) provides a structure:
(9) a NP1
N′NP2
N′specifier
N
the
N
teacher’s books
In (8b), similarly, the demonstrative that speciﬁes the reference of teacher: that
teacher’s books roughly corresponds to the books of that teacher. We have seen that
English demonstratives agree with the head noun. Consider representation (9b).
According to the structure in (9b), do we predict that the demonstrative that will
agree with the N teacher or will it agree with the N books? Is this prediction correct?
(9) b NP1
N′NP2
N′specifier
N
that
N
teacher’s books
In (9b), the demonstrative that is the speciﬁer of NP2, whose head is the N teacher.
Thus we correctly predict that the demonstrative agrees with the head N of NP2: that
is singular and so is teacher. The demonstrative in the speciﬁer position of NP2 does
not agree with the head of NP1: in our example the head of NP1, books, is plural.
If we replace the singular N teacher’s in NP2 by its plural counterpart teachers’,
then the demonstrative will also have to be plural.
Exercises 125
TSC02 07/18/2005, 03:22PM125
126 Chapter 2
(10) a those teachers’ books
b *that teachers’ books
Exercise 2 Bracketing representations (T)
Discuss why there are exactly three brackets following meal in (1). For each of the
right-hand brackets identify the label of the matching left-hand bracket.
(1) [S [NP The customer in the corner] [Aux will] [VP [VP order [NP the drinks]] [PP before
the meal]]].
Does the representation in (1) take into account the layered VP analysis? Does it
represent the auxiliary will as forming a constituent with the VP?
Provide a labeled bracketing representation for the following examples:
(2) The customers will order the drinks.
(3) All the customers will order the drinks.
(4) My analysis of this issue will quite probably surprise all the representatives of
the media.
Exercise 3 Representations of structure (T)
Compare the tree diagram in (1) and the bracketed representation in (2). Which is
more detailed? Motivate your answer.
(1) S
CONSTITUENTNP
The customer
in the corner
Aux VP
VP PP
V
order
NP
will the drinks before the meal
TSC02 07/18/2005, 03:22PM126
Diagnostics for Syntactic Structure 127
(2) [NP The customer in the corner] [AUX will] [VP order the drinks before the meal].
By adding the necessary brackets and labels, modify (2) so that it contains all the
information contained in (1).
Exercise 4 Evidence for constituent structure (T)
Consider the underlined strings of words in the extracts below. The extracts
themselves contain a clue to support an analysis according to which the underlined
strings form a constituent. Try to ﬁnd that clue. For each underlined constituent,
identify the grammatical category (NP, VP, etc.). There is no need to draw the tree;
you should merely explain on what basis the sequences could be argued to form a
constituent.
Example
The news, when it comes, he seems to take —— well enough. (Guardian, G2,
26.7.2002, p. 2, col. 1)
• In this example, movement has affected the string the news. The news is
the direct object of take. Its canonical position is to the immediate right of
the verb. In the example it has been moved to the beginning of the
sentence. The fact that it has been moved suggests that the string of
words is a constituent. Moreover, the string the news is substituted for by
the pronoun it.
• The news is an NP.
(1) Is there anything that can prevent Hurricanes? To date, science and technology
have not given us the ability to do so. (Chicago Tribune, 3.1.2004, section 1,
p. 28, col. 1)
(2) There are hardly any small movies that people go to, and some of the more
interesting ones they won’t go to. (Guardian, 1.11.2002, Review, p. v, col. 4)
(3) I think we could adapt to being poor again if we had to. (Guardian, G2,
28.8.2002, p. 4, col. 2)
(4) Great Expectations, I’ve read three times. (Guardian, G2, 1.4.2003, p. 12,
col. 2)
(5) The Bears are not a talented team and have not been in years, not even in
2001. (Chicago Tribune, 3.1.2004, S3, p. 2, col. 5)
Exercises 127
TSC02 07/18/2005, 03:22PM127
128 Chapter 2
(6) But what we do is go to shopping centers and corporate parties downtown.
(Based on Chicago Tribune, 22.12.2002, section 1, p. 16, col. 6)
(7) The towns Paula Radcliffe has lived in (Nantwich, Bedford and Loughborough)
are the epitome of Middle England. And so is she. (Guardian, 17.12.2002,
p. 13, col. 1)
(8) If they must use children’s stories in a vain attempt to make their speech
sparkier, why can’t they choose new ones? (Guardian, 12.2.2003, p. 2,
col. 4)
(9) They are eyeing retirement and just entering the work force. (Atlanta JournalConstitution,
23.11.2003, p. F5, col. 1)
(10) It’s unbelievable how unlucky he’s been, but he’s certainly proved he’s got
tenacity. Whether he’ll get out of this difﬁcult situation or not, only time will
tell. (Guardian, 8.2.2003, p. 2, col. 8)
Exercise 5 Focalizing a constituent by rightward
movement (T, E)
We propose that a direct object is merged directly with the verb to form V′. The
canonical object position is to the immediate right of the verb. In the following
examples some direct object NPs are not found in their canonical positions. Identify
the displaced objects, identify the verb that selects them, and try to restore the
displaced objects to their canonical positions.
(1) “The Independent” is publishing daily each of the 30 Articles of the Universal
Declaration of Human Rights, illustrated by Ralph Steadman, to mark its 50th
anniversary on 10 December. (Independent, 13.11.1998, p. 13, cols 1–2)
(2) The government has delayed until the New Year the introduction of the
controversial Bill to strip the 750 hereditaries of their right to speak and vote in
the second chamber. (Guardian, 25.22.1998, p. 1, col. 1)
(3) We look forward to the opportunity to defend before the Supreme Court
Maryland’s historic ownership and regulation of this important natural resource.
(Washington Post, 29.4.2003, p. A7, col. 1)
(4) Few if any authors on the region have so successfully compressed into 280
pages the basic outlines of Antarctic life and our relationship to its pristine
abundance. (Guardian, Review, 22.11.2003, p. 11, col. 2)
TSC02 07/18/2005, 03:23PM128
Diagnostics for Syntactic Structure 129
(5) The argument has been that members of Congress get paid by the taxpayer
their normal salary while they’re running for ofﬁce. (Los Angeles Times,
26.11.2002, p. A32, col. 3)
In the above examples the objects of the verbs publishing (1), delayed (2), defend
(3), compressed (4), and paid (5), have apparently been shifted rightward. The effect
of shifting the object to a rightward position is that it becomes the focus of the
sentence, it is focalized.
(1′) “The Independent” is publishing —— daily [NP each of the 30 Articles of the
Universal Declaration of Human Rights, illustrated by Ralph Steadman] to
mark its 50th anniversary on 10 December.
(2′) The government has delayed —— until the new year [NP the introduction of
the controversial Bill to strip the 750 hereditaries of their right to speak and
vote in the second chamber].
(3′) We look forward to the opportunity to defend —— before the Supreme Court
[NP Maryland’s historic ownership and regulation of this important natural
resource].
(4′) Few if any authors on the region have so successfully compressed —— into
280 pages [NP the basic outlines of Antarctic life and our relationship to its
pristine abundance].
(5′) The argument has been that members of Congress get paid —— by the
taxpayer [NP their normal salary while they’re running for ofﬁce].
Experiment: Replace each of the shifted objects in (1)–(5) by a personal pronoun.
Are the resulting sentences acceptable? As suggested by the sharp degradation in
acceptability of the examples below, we conclude that not all direct objects can
freely be moved rightward away from the selecting verb.
(6) *The Independent’ is publishing daily them.
(7) *The government has delayed until the New Year it.
(8) *We look forward to the opportunity to defend before the Supreme Court
them.
(9) *Few if any authors on the region have so successfully compressed into 280
pages them.
(10) *The argument has been that members of Congress get paid by the taxpayer it.
Exercises 129
TSC02 07/18/2005, 03:23PM129
130 Chapter 2
The degradations in (6)–(10) can be related to the focalizing role of the rightward
shift. Direct object NPs which are moved rightward tend to be focused. If some
information is worth focusing then probably it will contain something new. Pronouns
typically represent old information and therefore are not good candidates for focusing.
Consider the test sentence (11). Can we shift the object rightward?
(11) The customer in the corner will order the drinks before dinner.
And what about (12)?
(12) The customer in the corner will order all the drinks to go with the entire meal
and the dessert before dinner.
Finally consider (13). How is the complement of the verb argued realized? Does the
complement occupy its base position?
(13) This week proved wrong the doomsayers who argued after the failures in
Seattle in 1996 and The Hague in 1999 that the pace of globalisation had
outstripped the international institutions’ ability to respond, let alone to
manage it. (Adapted from Guardian, 17.11.2001, p. 9, col. 2)
In (13) the complement of argued is a clause: that the pace of globalisation
had outstripped the international institutions’ ability to respond, let alone to
manage it. For other examples in which the complement of the verb is a clause see
Exercise 10.
Exercise 6 Structural ambiguity and
co-ordination (T)
In the following extracts co-ordination with and may give rise to two interpretations.
Discuss the two readings and identify how they have come about. How
could you rephrase the sentences, changing as little as possible, to eliminate the
ambiguity?
(1) He added that the looting, though continuing, is much reduced. “You will see
a guy or two carrying a table or chairs. We tell them to put it down and go
home.” (Guardian, 7.5.2003, p. 5, col. 14)
(2) If you feel threatened in a mini cab, ﬁrmly ask the driver to stop and get out.
(Adapted from Guardian, G2, 7.3.2003, p. 7, col. 2)
TSC02 07/18/2005, 03:23PM130
Diagnostics for Syntactic Structure 131
Exercise 7 Representing co-ordination (T, E)
Identify the co-ordinated constituents in the following examples. What is their
category? How can we show that co-ordinating two constituents itself creates a
constituent? What is the category of the co-ordinated constituent as a whole?
(1) a The customer in the corner and his wife will order the drinks.
KEY AND COMMENTS
In (1a) we co-ordinate two NPs.
(1) b [NP the customer in the corner]
and
[NP his wife]
We can cleft the resulting co-ordinated string, suggesting that it is a constituent.
(1) c It is [[NP the customer in the corner] and [NP his wife]] who will order the drinks.
We can also question the string by means of who.
(1) d Who will order the drinks?
In (1a/b) the co-ordinated string the customer in the corner and his wife is the
subject of the sentence. It can be replaced by a pronoun.
(1) e They will order the drinks.
Typically, pronouns like they replace NPs. So we conclude that the constituent
consisting of two co-ordinated NPs in (1c) is an NP. (1f) is a provisional structure, in
which we label and as a co-ordinator (“co-ord”).
(1) f NP
co-ord NPNP
the customer in the corner his wifeand
Based on the discussion above, how could we represent the structure of the
underlined co-ordinated constituent in (2)?
(2) The customer in the corner will order the drinks and pay for the meal.
Exercises 131
TSC02 07/18/2005, 03:23PM131
132 Chapter 2
NOTE
With respect to the discussion in section 2.4, observe that the representation of coordinated
structures raises a problem for the binary branching hypothesis. We will
not go into this issue here.2
Exercise 8 Structural ambiguity (T, E)
In section 1.8 of the chapter we discussed co-ordination of VPs and we introduced
example (34b), repeated here as (1):
(1) The customer in the corner will order the drinks and choose the dessert before
the meal.
In the text, this example was used to illustrate the co-ordination of the VPs order
the drinks and choose the dessert. First, use bracketing to represent the co-ordination
as intended in the text. What does the temporal speciﬁcation before the meal
modify?
The bracketing intended in the text discussion is that represented in (2a). The VP
order the drinks and the VP choose the dessert are co-ordinated. The PP before the
meal is added to the result of the co-ordination:
(2) a The customer in the corner will
[[VP order the drinks] and [VP choose the dessert]] before the meal.
The unit consisting of the co-ordinated VPs has the same distribution as a simple
VP: in our example it follows the auxiliary will. We can pseudo-cleft the unit
consisting of the co-ordinated VPs, as shown in (2b):
(2) b What the customer in the corner will do before the meal is [order the drinks
and choose the dessert].
Or we can replace the co-ordinated string by do so:
(2) c The customer on the right will do so later.
These observations suggest that in (2a) the co-ordinated string order the drinks and
choose the dessert is a VP. In tree format the co-ordinated VP could be represented
by (2d). We complete the bracketing notation in (2e).
2
For discussion see Camacho (2003), Goodall (1987), van Oirsouw (1987), and the papers in
Blakemore and Carston (2005).
TSC02 07/18/2005, 03:23PM132
Diagnostics for Syntactic Structure 133
(2) d VP
co-ord VPNP
order the drinks choose the dessertand
e The customer in the corner will
[VP [VP order the drinks] and [VP choose the dessert]] before the meal.
To complete the representation of the VP of (2a) we need to integrate the PP before
the meal. We add the PP before the meal as a temporal adjunct of the co-ordinated
VP in (2d/e). This representation means that the PP will bear on both components
of the co-ordinated constituent:
(2) f VP
PPVP
VP VPco-ord
and choose the dessert before the mealorder the drinks
g The customer in the corner will
[VP [VP [VP order the drinks] and [VP choose the dessert]] before the meal].
According to representation (2f/g), both actions of ‘ordering the drinks’ and ‘choosing
the dessert’ will take place before the meal.
Observe that there is a second interpretation possible for (1) and one that was
not intended in the text discussion. The second interpretation is due to the fact that
we can assign the temporal speciﬁcation to a different domain. Can you see the
second interpretation of (1)? In the second reading, (1) can be paraphrased as in
(3a). How would we bracket example (1) to bring out this reading?
(3) a The customer in the corner will choose the dessert before the meal and
order the drinks.
The alternative bracketing for text example (1), which was not intended in the
discussion in the main body of the chapter, is shown in (3b). In this representation
the PP before the meal is adjoined directly to the VP choose the dessert. The PP is
not intended to modify the VP order the drinks. The co-ordinated VPs are order the
drinks and choose the dessert before the meal.
Exercises 133
TSC02 07/18/2005, 03:23PM133
134 Chapter 2
(3) b The customer in the corner will [VP [VP order the drinks] and [VP [VP choose
the dessert] [PP before the meal]]].
In tree format the co-ordinated VPs in (3b) would be represented as in (3c).
(3) c VP
VP
VP PP
co-ordVP
and choose the dessert before the mealorder the drinks
In this representation the PP before the meal combines with the VP choose the
dessert. It does not specify the time of order the drinks. Pseudo-clefting affects the
string order the drinks and choose the dessert before the meal. In this interpretation,
it is not possible to leave out the PP before the meal when the co-ordinated VP is
pseudo-clefted.
(3) d What the customer in the corner will do is [order the drinks and choose the
dessert before the meal].
Using the above examples as a model, discuss the representation of the co-ordinated
VP in (3a).
NOTE
With respect to the discussion in section 2.4 you will conclude that the representation
of co-ordinated structures raises a problem for the binary branching hypothesis. We
will not go into this issue here.
Exercise 9 Co-ordination (T, E)
On the basis of the examples below, discuss whether the following generalization is
valid:
The co-ordinating conjunction and always links constituents of the same category.
(1) Now that women are totally independent and earning their own money they
are less likely to put up with a bad marriage. (Guardian, G2, 16.10.2002,
p. 11, col. 2)
TSC02 07/18/2005, 03:23PM134
Diagnostics for Syntactic Structure 135
(2) We have expressed our disappointment to the Philippine government very
clearly and at high levels. (New York Times, 1.8.2004, p. 3, section 4, col. 4)
(3) Which is true? Thompson demanded, folding his arms and glowering down
at the witness. (Wall Street Journal, 26.3.2004, p. A6, col. 4)
(4) Winners included a day laborer in Florida who had recently been homeless
and sleeping on cardboard boxes. (New York Times, 8.3.2004, p. C6, col. 4)
(5) Two men who survived the attack were being tended by marine medics and
being prepared for evacuation to hospital. (Based on Guardian, 28.3.2003,
p. 3, col. 1)
(6) I can understand that Jack Straw is a busy man, and unlikely to be able to
afford the time for mathematics. (Guardian, 15.3.2003, p. 11, col. 8, letter
to the editor from Greg Callus, University of York)
(7) His black shirt was soggy and hanging out of his pants. (New York Times,
8.3.2004, p. D1, col. 2)
(8) Fifty minutes later Sarah arrived at the hospital, unconscious and in a ﬁt. At
her home near Whitchurch in Cheshire, her mother, Pauline Campbell, was
preparing to go out, having heard nothing from the prison and unaware that
her only daughter was ﬁghting for her life. (Guardian, 5.5.2003, p. 7, col. 1)
(9) If sound comes on both sides at exactly the same time and with the same
amplitude, the mechanism doesn’t move. (Chicago Tribune, 3.1.2004, p. 4,
section 2, col. 6)
(10) We wanted justice and for the truth to be known. (Guardian, 12.11.2002,
p. 6, col. 7)
Some of the examples above show that co-ordinated constituents do not always
have identical categories. Discuss the problems raised by these examples for the
representation of co-ordinate structures elaborated in Exercise 7.3
Exercise 10 Realization of subject and object (T, E)
In the discussion in this chapter we have systematically been using examples in
which both subject and object are realized as NPs. This might give the impression
that only NPs can function as subjects or as objects. Similarly, our examples of
Exercises 135
3
For a proposal on how to handle such types of co-ordinations see Bowers (1993, 2001, 2002).
TSC02 07/18/2005, 03:23PM135
136 Chapter 2
prepositional phrases always consisted of a preposition with an NP complement.
Based on the examples below and using the concepts elaborated in the chapter,
evaluate statements A, B, and C.
You should ﬁrst read the statements carefully, and for each statement examine
whether the examples are relevant and if so, in what way they conﬁrm/disconﬁrm
the statement.4
A The subject of a sentence is always realized by a noun phrase.
B The object of a verb is always realized by a noun phrase.
C The complement of a preposition is always realized by a noun phrase.
(1) Just because the US administration says that it has “no interest” in implementing
the Kyoto protocol to control climate change doesn’t mean it’s
dead. (Guardian, 31.3.2001, p. 9, col. 8, letter to the editor from Ritu Kumar)
(2) She couldn’t stop asking questions about whether they would be OK.
(Guardian, Travel, 9.8.2003, p. xi, col. 1)
(3) The skirt is also available in black, so now might be just the time to go for it.
(Guardian, 24.11.2000, p. 8, col. 8)
(4) That 85% of children spelt Hogwarts correctly is no surprise. (Guardian,
4.9.2002, p. 5, col. 2)
(5) It’s the curve of those thighs that has, surely, guaranteed that she’s sent to
Islamabad, even though above the waist is all we ever get to see on the
television. (Guardian, G2, 25.10.2001, p. 10, col. 2)
(6) She knew how many novels she would write. (Based on Guardian, Review,
2.8.2003, p. 14, col. 4)
(7) I told Simon Kelner I wouldn’t want to do the job on a long-term basis, and
after a general election seems a natural departure time. (Guardian, G2,
16.4.2001, p. 13, col. 4)
(8) Before 11 September seems like an innocent lost paradise. (Guardian, G2,
13.11.2001, p. 11, col. 5)
(9) Many league members accept that some money is better than no money at
all. (Guardian, 29.3.2002, p. 16, col. 1)
4
For PP subjects see Jaworska (1986). For clausal subjects see Koster (1978), Davies and Dubinsky
(1999, 2001a), Miller (2001).
TSC02 07/18/2005, 03:23PM136
Diagnostics for Syntactic Structure 137
(10) You realise he hates for you to call him “Ollie”, don’t you? (Marcia Muller,
Edwin of the Iron Shoes, 1993: 42)
Exercise 11 Constituent structure in the NP (T)
We might provide the structure in (1) for the NP all those students of English.
Discuss theoretical and empirical problems that arise.
(1) NP
determiner PP
of English
NQ
those studentsall
Exercise 12 Auxiliary strings (T, L)
In our discussion so far we have concentrated on sentences containing just one
auxiliary. This is only one of a number of patterns in which auxiliaries can be found.
Identify all the auxiliaries in the following examples. Which of the auxiliaries are ﬁnite?5
(1) He would have stolen her fame.
(2) You should have become an accountant.
(3) He’ll be staying at the local inn.
COMMENTS
As you can see, each example contains more than one auxiliary: (1a) contains a
modal auxiliary would and a perfect auxiliary have. The modal auxiliary is tensed:
past tense would contrasts with present tense will. The auxiliary have is not tensed,
it is an inﬁnitive form and this is not affected by tense modiﬁcations; would is a
ﬁnite auxiliary and have is non-ﬁnite. Which are the ﬁnite/non-ﬁnite auxiliaries in
(2) and in (3)? We note that in each example there is just one ﬁnite auxiliary and
that this precedes the non-ﬁnite auxiliary. Could there be more than one ﬁnite
auxiliary in the string? Could the ﬁnite auxiliary be second or third in the sequence?
Experiment with examples (1)–(3) to answer these questions.
5
See Chapter 1, sections 2.2.2 and 2.4 for the concept “ﬁnite.”
Exercises 137
TSC02 07/18/2005, 03:23PM137
138 Chapter 2
Exercise 13 Auxiliary strings (T, L)
In the following examples the symbol ∅ signals that material has been omitted.
Identify the omitted material. What is the category of the omitted constituent
(VP, etc.)?
(1) He drove her hard, he stole her fame or would have ∅ if he could have ∅.
(Guardian, Review, 24.5.2003, p. 5, col. 3)
(2) “The Hershey chocolate company is about to be sold!” he says, eyes widening.
“Who could have imagined it?” Very few could ∅. (Based on Guardian, G2,
26.8.2002, p. 2, col. 1)
(3) Michael Jackson has, on some occasions in the past, not eaten when he should
∅. (Guardian, Review, 28.5.2003, p. 3, col. 1)
COMMENT
In (1) the omitted material is a VP: stolen her fame. The auxiliary have is retained.
In (2) and (3) the omitted material combines a VP with an auxiliary: have imagined
it, have eaten. We will tentatively conclude that because they can be subject to
ellipsis such strings are constituents. Hence, the non-ﬁnite auxiliary (here have) and
the VP together form a constituent. How do the following examples bear on this
hypothesis?
(4) Some 24% agreed top-up fees would not have mattered, while 35% would
have considered other universities but probably still have chosen the same
university. (Based on Guardian, 20.1.2003, p. 5, col. 2)
(5) But we have been saying for 15 to 20 years that there are too many games
in the top division and done nothing about it. (Guardian, 15.2.2003, Sport,
p. 11, col. 5)
We return to the structure of sentences with multiple auxiliaries in Chapter 3,
section 4.
Exercise 14 Attachments of adjuncts (T, L)
In this exercise we speculate about the application of adjunction. The exercise is
longer and more discursive than some of the other exercises. The idea is that we
TSC02 07/18/2005, 03:24PM138
Diagnostics for Syntactic Structure 139
carry further the type of argumentation elaborated in the chapter and see where
that would lead us.
We will examine the consequences of the hypothesis that syntactic structure
determines interpretation. Consider sentence (1a). Where would we locate the
adjuncts still and actively?
(1) a I am still actively monitoring the reporting of the press coverage of the trial.
(Guardian, 29.11.2003, p. 7, col. 2)
COMMENT
The two adjuncts can both be adjoined to the VP headed by monitoring:
(1) b VP
VPAdvP
still
AdvP VP
monitoring itactively
The different adjunction sites mirror the interpretation: according to (1b) ‘it is
still the case that we are actively monitoring it’. Or to use the technical term:
the adverb still takes scope over the VP actively monitoring it; the adverb actively
takes scope over the VP monitoring it. The structure in (1b) represents the relative
scope of the adjuncts: an adjunct that is adjoined higher in the structure takes
scope over one that is adjoined lower.
Consider sentence (2a), which was also given in Chapter 1, Exercise 4.
(2) a Tony Blair was admitting that he had run into “tough times” in recent
months yesterday. (Adapted from Independent, 5.9.2003, p. 2, col. 1)
We ﬁnd two temporal speciﬁcations, the PP in recent months and the NP yesterday,
one after the other. Can we adjoin them both to the VP headed by run? Obviously,
if we did that, then we would create some redundancy. Since the stretch of time
denoted by in recent months comprises the time speciﬁed by yesterday, the latter
speciﬁcation would be sufﬁcient. On the other hand, if the time stretch referred to
by yesterday had been intended to be distinct from that referred to by in recent
months, and if both adjuncts were associated with the VP headed by run, then the
two adjuncts would locate one and the same event (‘running into tough times’) at
two different times, and we would end up with a contradiction. Adjoining both
Exercises 139
TSC02 07/18/2005, 03:24PM139
140 Chapter 2
adjuncts to the VP headed by run is thus probably not the correct way to go about
it if we think of the intended interpretation of the sentence.
Let us ﬁrst pick out the main ingredients of (2a). Identify all the verbs in (2a).
Which are lexical verbs? Which are auxiliaries? Which lexical verbs or auxiliaries are
ﬁnite? Which are non-ﬁnite? What is the subject of admitting? What is the object
of admitting? What is the subject of had run into tough times?
(2a) contains two lexical verbs admitting and run, and it also contains two ﬁnite
auxiliaries, was and had. (2a) is a complex sentence: the complement of the verb
admitting is itself realized by a sentence introduced by the subordinating conjunction
that: that he had run into tough times in recent months. When a sentence functions
inside another sentence we say it is embedded. We often use the term embedded
clause.6
If you think about the intended interpretation of the temporal speciﬁcations in
recent months and yesterday, you will conclude that in recent months modiﬁes the
VP had run into tough times, while yesterday modiﬁes admitting that he had run
into tough times. In recent times is an adjunct in the embedded clause; yesterday is
an adjunct in the main clause.
Try to isolate the embedded clause by using the diagnostics we have introduced
to identify constituents. For instance, (i) formulate a question that targets the object
of admitting. (ii) Using the pseudo-cleft pattern, try to rephrase the sentence to
bring out the different attachments of the adjuncts. (iii) Move the direct object of
admitting to the beginning of the sentence.
(3) a What was Tony Blair admitting yesterday?
That he had run into tough times in recent months.
b What Tony Blair was admitting yesterday was that he had run into tough
times in recent months.
c That he had run into tough times in recent months, Tony Blair was admitting
yesterday.
Finally, you can also shift the complement of admitting to the right, across yesterday.7
(3) d Tony Blair was admitting yesterday —— that he had run into tough times in
recent months.
These various diagnostics lead us to propose that the adjunct in recent months is
adjoined to the VP headed by run, and that the adjunct yesterday is adjoined to the
VP headed by admitting. (2b) is a representation modeled on representation (70b)
in the discussion section. In (2b) we have added that to the structure with the label
C. We come back to the position of conjunctions like that in Chapter 5.
6
We discuss embedded clauses in more detail in Chapter 5, section 2.3.
7
For rightward shifting of objects see Exercise 5 above.
TSC02 07/18/2005, 03:24PM140
Diagnostics for Syntactic Structure 141
(2) b AuxP
NP Aux′
TB
Aux VP
VP NP
V clause
admitting
C AuxP
NP Aux′
Aux VP
had
VP PP
run
into tough times in recent monthswas that he
yesterday
Exercise 15 Adjunction to the sentence (L, E)
In this exercise we speculate about the application of adjunction. The exercise is
more discursive than some of the other exercises. The idea is that we pursue the
argumentation elaborated in the chapter and see where that would lead us.
In representation (70b) in the chapter, we analyze sentences containing a ﬁnite
auxiliary as projections of that auxiliary: the auxiliary heads the sentence, the sentence
is an AuxP. In the representation we showed how the head of the sentence, Aux,
takes a VP as its complement and has a subject NP as its speciﬁer.
We did not discuss the possibility of adjunction to the sentence. However, if
adjunction exists as a general option available to any structure formed according to
the format elaborated in the chapter, then sentences ought also to be able to have
constituents adjoined to them. Once again, if we were to conclude that adjunction
to sentences is not possible then we would ideally have to explain this ban.
Consider (1a). If we were to left-adjoin the NP this week to the sentence, what
would the resulting sentence be? Using (70b) as a model, draw a tree diagram for
the sentence with the adjunct integrated into it.
Exercises 141
TSC02 07/18/2005, 03:24PM141
142 Chapter 2
(1) a The director will meet the team.
A constituent left-adjoined to the sentence would precede the subject:
(1) b This week, the director will meet the team.
The resulting structure would be as in (1c):
(1) c AuxP
AuxP
the team
NP
willThis week
Aux′NP
NPV
the director
VPAux
meet
Observe that the adjoined constituent, the NP this week, precedes the subject. We
have a structure in which a core AuxP is augmented with an additional element.
Adjunction is the name for the operation which combines fully formed constituents.
So far, we have used this operation to integrate non-essential components into the
structure. For instance, we used adjunction to add speciﬁcations of time or manner to
the VP. These adjoined constituents were optional and they were peripheral to the
VP. Could obligatory components of the sentence be found in an adjoined position?
Consider example (4) in Exercise 4, repeated here as (2a). The subject I is preceded
by a constituent. What type of constituent is it? What would be the unmarked
position for this constituent?
(2) a Great Expectations, I’ve read three times. (Guardian, G2, 1.4.2003, p. 12,
col. 2)
In this example Great Expectations is an NP. It is the name of a book. It is the
object of the verb read. It can be replaced by a pronoun; we can question it with
what. If we restore it to its base position we have (2b).
(2) b I’ve read [NP Great Expectations] three times.
Before turning the page, try to draw a tree diagram for (2b), using diagram (70b) in
the chapter as a model. You can spell out the contracted auxiliary ’ve as have. The
tree should look as follows:
TSC02 07/18/2005, 03:24PM142
Diagnostics for Syntactic Structure 143
(2) c AuxP
Aux′
three times
NP
readI
VPAux
have
NPVP
Great Expectations
NPV
In (2b) the NP Great Expectations is the object of the verb read, and it occupies its
base position in the core VP. In (2a) the NP Great Expectations precedes the
subject, the NP I. However, Great Expectations is still the direct object of read. For
instance, in (2a) we could not add a new object to the sentence:
(2) d *Great Expectations, I’ve read War and Peace three times.
Sentences (2a) and (2b) are paraphrases: if (2a) is true, then inevitably (2b) is also
true and vice versa. (2b) is the neutral pattern, in which the object occupies its canonical
position and follows the verb; in (2a) the object has been fronted to achieve
some presentational effect. Let us take this movement metaphor literally. Let us
assume that in (2a) the object is ﬁrst merged with the verb, but that it is subsequently
moved to occupy a position to the left of the subject. What could the position to
the left of the subject be? One option that comes to mind is that the NP Great
Expectations is adjoined to the sentence. We use an arrow to indicate the movement.8
(2) e AuxP
AuxPNP
readGreat Expectations
Aux′NP
I
VPAux
NPV
have
Exercises 143
8
In Chapter 4 we will introduce another notation to signal movement. We discuss the mechanisms
of movement also in Chapter 5.
TSC02 07/18/2005, 03:24PM143
144 Chapter 2
In (2e) we have ﬁrst inserted Great Expectations as the object of the verb read.
Then we move it to a position adjoined to AuxP.
Example (2a) is important because it shows that adjunction does not apply
exclusively to optional constituents. The direct object Great Expectations in (2) is an
obligatory complement of the verb. When it is taken out of the VP, it may be
moved to an adjunction position. Observe, however, that the operation that creates
the adjoined position could still be said to be “optional” in the sense that we are
not obliged to move an object out of the VP. If we had left the object in its base
position, we would not have had to resort to adjunction. Inserting the object in the
VP is obligatory, moving it out and adjoining it is optional.
In section 1.4. in the chapter we came across examples such as (13) repeated
here in (3). In each of the underlined segments the subject is preceded by a VP.
How could we represent the structure of the underlined segments?
(3) a “They must talk about it, and [VP talk about it] they must”, he said. Food for
thought, there! It’s a phrase that could add a measure of gravity to any
press conference. “We must do this, and [VP do this] we must.” (Simon
Hoggart, Guardian, 29.1.2003, p. 2, col. 5)
b But I was still a long way from ﬁguring out what my goal was. I told the
governor [of the prison] that I wasn’t sure how I was going to manage it –
but [VP manage it] I would. (Guardian, G2, 15.5.2003, p. 7, col. 4)
Assuming that the structure of the sentence is as in (70b), we would have to
provide accommodation for a VP that precedes the subject. Among the tools available
in our structure we have a possibility of adjunction. We might propose to adjoin the
sentence-initial VP to the sentence as a whole, in the same way that we adjoined
the temporal NP this week to AuxP in (1) and the object NP Great Expectations to
AuxP in (2). According to (70b) the VP is the complement of the auxiliary. Again,
we can propose that while the fronted VP is in a position adjoined to AuxP, it
originates inside the VP.
(3) c AuxP
AuxPVP
Talk about it
Do this
Manage it
Aux′NP
they
we
I
VPAux
must
must
would
TSC02 07/18/2005, 03:24PM144
Diagnostics for Syntactic Structure 145
Exercise 16 Creating new words (T)
Comment on the internal make-up of the underlined words in the following extracts.
(1) The Navy hopes the change will make a difference in its rigorous nine-week
sailorization process. (Chicago Tribune, 30.11.2002, section 1, p. 18, col. 6)
(2) The brigades have already taken part in exercises in preparation for Iraq, and
the process of “desertiﬁcation” – ﬁtting special air ﬁlters, to painting the
camouﬂage in desert colours and other changes – is under way. (Guardian,
7.1.2003. p. 1, cols 4–5)
(3) Options are being investigated for the desertisation of the UK’s CR2s (Challenger
2s). The army has bought 386 Challenger2 tanks – adapting all of them for desert
conditions would cost more than £50m. (Guardian, 8.4.2002, p. 7, col. 8)
(4) In a lawyerly way, she worked with her sister to lay down ground rules. (New
York Times, 2.1.2003, p. F8, col. 1)
(5) He gives me his best schoolteacherly look. (Guardian, G2, 20.1.2003, p. 7,
cols 3–7)
Exercise 17 The internal structure of deverbal
nouns (E)
This exercise explores the internal structure of nouns.9
Consider the underlined
nouns in the examples below. Several of them can be related in form to a verb.
Identify these nouns and for each noun give the related verb. Nouns with a clear
morphological relation to a verb are called deverbal nouns.
(1) Kim’s explanation of the events did not satisfy me.
(2) Kim’s accident changed everything.
(3) His transformation into a werewolf was unnerving.
(4) Kim’s version of the events was not satisfactory.
(5) The occurrence of the accident changed everything.
(6) Sue’s exploration of Easter Island was uneventful.
9
This exercise is based on Fu, Roeper, and Borer (2001).
Exercises 145
TSC02 07/18/2005, 03:25PM145
146 Chapter 2
KEY AND COMMENTS
The nouns in Table 1 are deverbal. There is a debate among syntacticians whether
the internal structure of deverbal nouns should contain a verb, i.e. whether representation
(7a) or (7b) is preferable:
(7) a NP
N′Det
N PP
of the problemexplanationKim’s
b NP
N′Det
N VP
of the problemexplainKim’s
V′
V PP
-ation
Table 1 Deverbal nouns and the related verbs
Noun Verb
(1) explanation explain
(3) transformation transform
(5) occurrence occur
(6) exploration explore
TSC02 07/18/2005, 03:25PM146
Diagnostics for Syntactic Structure 147
(7a) does not contain any indications that the noun explanation is related to a
verb. According to (7b), the noun explanation is formed on the basis of the verb
explain. According to this representation, the verb explain unites with the bound
morpheme -ation, which is dominated by N, to form the deverbal N explanation.
To what extent would the data in (8) be relevant for the choice between the two
representations?
(8) a Kim’s explanation of the problem to the tenants thoroughly did not prevent
a riot.
b The occurrence of the accident suddenly disqualiﬁed her.
c His transformation into a werewolf so rapidly was unnerving.
Nouns for which there is no related verb stem (version, accident) would be taken to
have a representation not containing a verb stem (7a). To what extent would the
following data support this analysis?
(9) a *Kim’s version of the accident thoroughly was not a big help.
b *Kim’s accident suddenly on the track disqualiﬁed her.
KEY AND COMMENTS
In (8), the deverbal nouns co-occur with an adverbial modiﬁer (thoroughly, suddenly,
so rapidly), which we normally expect to be adjoined to a VP. Structure (7b), with
an NP dominating a VP, would allow the adverbial to be an adjunct to the VP. No
such modiﬁer is available for the nouns that are not related to a V-root (9), so
representation (7b) would not be appropriate for these and we would opt for (7a).
Example (10) is provided by Fu, Roeper, and Borer (2001). Discuss whether this
example would be compatible with the hypothesis that deverbal nouns have the
internal structure in (7b).
(10) Sue’s exploration of Easter Island was impressive, but Amy’s doing so was a
real surprise.
In (10) doing so relates to exploration of Easter Island. If we assume that do so
substitutes for a VP (see section 1.3), then the relevant VP could be based on the
verbal root (explore) of the N exploration in (10). Representation (7b) will allow us
to account for do so substitution. The substitutions with do so in (11) again relate
to deverbal nouns.10
Identify these deverbal nouns. What are the related verbs?
(11) a The defection of the seven moderates, who knew they were incurring the
wrath of many colleagues in doing so, signaled that it may be harder to
sell the GOP message on the crime bill than it was thought previously.
(Washington Post)
10
(11a, b, c) are also from Fu et al. (2001).
Exercises 147
TSC02 07/18/2005, 03:25PM147
148 Chapter 2
b Even though an Israeli response is justiﬁed, I don’t think it was in their best
interest to do so right now.
c His removal of the garbage in the morning and Sam’s doing so in the
afternoon were surprising.
d Canon Michael Hunter, rector of St James parish church in Grimsby, said it
was a sad day for natural justice and added that her return to the town
would have caused problems but she should have been allowed to do so.
(Guardian, 13.2.2004, pp. 1 and 2, cols 4, 5)
Observe that the empirical data are complex. Not all native speakers accept examples
like (10) and those in (11). But speakers do agree on the compatibility of adverbials
with deverbal nouns in (8) and the contrast with (9).
Exercise 18 Comparative linguistics: Deriving OV
orders (T, E)
In this exercise we speculate about cross-linguistic word order variation. The exercise
is longer and more discursive than some of the other exercises. The idea is that we
carry further the type of argumentation elaborated in the chapter and see where
that leads us.
In section 2.3.3 of the chapter we saw that in English the canonical position of
the object is to the immediate right of V. Though the object can move away from
that position (as shown by Exercises 4 and 15), it cannot occupy a position to the
immediate left of V.
Consider the Old English examples in (1). What conclusions could we draw
concerning the position of the direct object?
(1) a Hwi sceole we oqres mannes niman?
why should we another man’s take
‘Why should we take something belonging to someone else?’
(Ælfric, Lives of Saints, 24, 188) (from Fischer et al. 2000: 49)
b Hwi wolde God swa lytles qinges him forwyrnan?
why would God so small thing him deny
‘Why would God deny him such a small thing?’
(ÆCHom I, 1.14.2) (from Fischer et al. 2000: 49)
In these examples the complement precedes the verb. This is the unmarked order
for Old English. Old English is referred to as an OV language, this in contrast to
languages such as Modern English which are VO languages.
Recall from Chapter 1, section 3.2, that with respect to question formation, Old
English seemed to behave like Modern Dutch and German. Consider the Dutch
TSC02 07/18/2005, 03:25PM148
Diagnostics for Syntactic Structure 149
sentences below, paying attention to the position of the underlined object and the
verb. Is Dutch an OV language or is it a VO language?
(2) a Ik denk dat Jan [NP dat boek] kent.
I think that John that book knows
‘I think that John knows that book.’
b *Ik denk dat Jan kent [NP dat boek].
(3) a Marie wil vanavond [NP een boek] lezen.
Mary wants tonight a book read
‘Mary wants to read a book tonight.’
b *Marie wil vanavond lezen [NP een boek].
Would the structure of the VP as elaborated in the discussion in the chapter allow
us to represent the structure of the Old English sentences in (1) and of the Dutch examples
in (2) and (3)? How could we enrich our theory in order to allow for these data?
There are at least two ways of ensuring that we can also derive OV orders. One
option to derive the Dutch sentences in (2a) and (3a) is to assume that in Dutch the
base position of the complement position is to the left of V.
(4) a bVP
NPV
Modern English
NP
Dutch, Old English
VP
V
We would then have to propose that languages differ with respect to their assembly
techniques. While a verb will select an object to its right in English, it would have to
select an object to its left in Old English and in Dutch.
A second possibility would be to assume that the complement position of the
verb in Old English and in Dutch is the same as that in Modern English, but that in
Old English and in Dutch the direct object cannot remain in its base position in the
VP and that it must move to a position inside the sentence but to the left of the
verb. We could refer to the sentence-internal leftward shifting of the object as
object shift (Holmberg 1986).
(5) a b
V
Modern English
VP
NP NP
VP
V
“object shift”
Dutch, Old English
Exercises 149
TSC02 07/18/2005, 03:25PM149
150 Chapter 2
To complete representation (5b), we would have to specify which position the object
NP moves to. Following on the discussion in Exercise 15 we might think, for instance,
of adjoining the object to the VP.
(5) c
NP
VP
V
VP
NP
Whichever analysis we choose, we would have to explain why Modern English
differs from Old English and Dutch. This means that either we explain why there is
variation in the base position of the object (4) or we explain why Old English and
Dutch require a displacement of the object (5) and why this displacement is not
even available in Modern English.
Consider the Dutch examples in (6). Do they bear on the choice between (4)
and (5)?
(6) a Ik denk dat Marie [NP dat boek] erg goed kent.
I think that Mary that book very well knows
‘I think that Mary knows that book very well.’
b Jan wil [NP die tekst] helemaal begrijpen.
John wants that text completely understand
‘John wants to understand that text completely.’
According to the hypothesis in (4), the base position of the object in Dutch is left
adjacent to the verb, that in English is right adjacent to the verb. In the Dutch
examples in (6), the direct object NPs dat boek (‘that book’) and die tekst (‘that
text’) precede the verb. However, these objects are separated from the verb by the
adjuncts erg goed (‘very well’) and helemaal (‘completely’). This means that even
if we were to adopt (4b) to account for the OV order in Dutch, we would have
to further supplement hypothesis (4) with an operation that moves the object to
the left across the VP-adjoined constituent. In other words, even if we adopt (4)
we need something like object shift. So to represent the sentences in (6) we would
need:
(i) Hypothesis (4b)
(ii) + Object shift (5b)
The combination of the output of these two assumptions is summarized in (4c):
TSC02 07/18/2005, 03:25PM150
Diagnostics for Syntactic Structure 151
(4) c VP
VPNP
AdvP VP
erg goed
VNP
dat boek kent
Hypothesis (5) expresses the cross-linguistic variation between VO languages and
OV languages by means of the operation of object shift, which allows us to move
the object leftward. According to (5) we do not also assume that the internal order
of verb and complements varies. By simply using object shift we can derive the
Dutch word order (6a) as in (5c):
(5) c VP
VPNP
AdvP VP
erg goed
V NP
dat boek kent
Both hypothesis (4) and hypothesis (5) allow us to derive the sentences in which
the verb follows the object in (2), (3), and (6). Both hypotheses need object shift to
derive the examples in (6), in which the object is non-adjacent to V. But hypothesis
(4) also postulates there is variation in the base position of object and verb. Which
hypothesis is theoretically preferred?11
Exercises 151
11
For the relevance of economy in theory building see also Chapter 1, section 1.2.3.
Exercise 16 in Chapter 3 offers another illustration of the application of object shift.
For an accessible discussion of the structure of Germanic SOV languages see Zwart (1997).
TSC02 07/18/2005, 03:25PM151
152 Chapter 2
Exercise 19 Constituency tests (T)
The following sentence is from Hebrew:12
(1) a Kol ha-yeladim zarku }avanim.
all the children threw stones
‘All the children threw stones.’
We may propose that like all the children in the English paraphrase, the string kol
ha-yeladim in (1a) is a constituent. Consider the sentences in (1b–e). Do they
provide evidence for constituency?
(1) b Ze hayu kol ha-yeladim se-zarku }avanim.
it was all the children that-threw stones
‘It was all the children who threw stones.’
c Mi- se zorek }avanim ze kol ha-yeladim.
who-that throws stones it all the children
‘Those that throw stones are all the children.’
d Kol ha-yeladim, }ani batu}ax se-zorkim }avanim.
all the-children, I sure that throw stones
‘All the children, I am sure throw stones.’
e }etmol zarku stei banot ve-kol ha-banim }avanim }al ha-mora.
yesterday threw two girls and all the boys stones on the teacher
‘Yesterday two girls and all the boys threw stones at the teacher.’
Now consider (2). In the ﬁrst sentence the subject is ha-yeladim kul-am (‘the
children all’). On the basis of the examples in (2b–e) decide whether this string is a
constituent:
(2) a Ha-yeladim kul-am zarku }avanim.
the children all-3MPL threw stones
‘The children all threw stones.’
b Ze hayu ha-yeladim kul-am se-zarku }avanim.
it was the children all-3MPL that-threw stones
‘It was all the children who threw stones.’
12
The data in this exercise are based on Shlonsky (1991: 163–4). Thanks to Ur Shlonsky for help
with (1a) and (2a). Exercises 12 and 21 in Chapter 4 pick up the variation between (1a) and
(2a).
TSC02 07/18/2005, 03:25PM152
Diagnostics for Syntactic Structure 153
c Mi- se zorek }avanim ze ha-yeladim kul-am.
who-that throws stones it the children all-3MPL
‘Those that throw stones are all the children.’
d Ha-yeladim kul-am, }ani batu}ax se-zorkim }avanim.
the children all-3MPL, I sure that throw stones
‘All the children, I am sure throw stones.’
e }etmol zarku stei banot ve- ha-banim kul-am }avanim }al ha-mora.
yesterday threw two girls and the boys all-3MPL stones on the teacher
‘Yesterday two girls and all the boys threw stones at the teacher.’
We conclude that both the sequence kol ha-yeladim (‘all the children’) and the
sequence ha-yeladim kul-am (‘the children all’) are constituents in Hebrew. Discuss
the relevance of (3) for this conclusion.
(3) }Ra}iti }et kol ha-banot ve-}et ha-banim kul-am.
saw-1SG ACC all the-girls and-ACC the-boys all-3MPL
‘I saw all the girls and all the boys.’
Exercise 20 The speciﬁer of NP (E)
In the discussion we assumed that each NP has one speciﬁer and that articles, demonstratives,
possessive pronouns, and prenominal genitives occupy the speciﬁer of NP.
This leads to the correct prediction for English that articles, demonstratives, possessive
pronouns, and prenominal genitives are in complementary distribution:
(1) a *the this book
b *the his book
c *the Jane’s book
Discuss the problems raised for this proposal by the following data:13
(2) a afto to vivlio
this the book
‘this book’ (Modern Greek: Horrocks & Stavrou 1987: 86)
13
For introductory discussion see Haegeman and Guéron (1999, chapter 4). For general discussion
of the structure of the noun phrase see also Bernstein (2001). For discussion based
on Greek see Horrocks and Stavrou (1987: 86); on Hungarian see Szabolcsi (1983, 1994). For
a discussion of possessives see also Giorgi and Longobardi (1991), Longobardi (1996), Alexiadou
(2004a).
Exercises 153
TSC02 07/18/2005, 03:25PM153
154 Chapter 2
(3) a la mia amica
the (FSG) my (FSG) friend (FSG)
‘my girlfriend’ (Italian)
b la meva casa
the (FSG) my-(FSG) home (FSG)
‘my house’
(Catalan, example from Longobardi 1996: 29, his (72a))
(4) a tu Chomsky to vivlio
of-the Chomsky the book
‘Chomsky’s book’ (Modern Greek, Horrocks & Stavrou 1987: 86)
b tu vivliu i kritiki
of the book the criticism
‘the criticism of the book’ (Modern Greek, Horrocks & Stavrou 1987: 86)
c a Mari kalap-ja
the Mary-NOMINATIVE hat-3SG
‘Mary’s hat’ (Hungarian, Szabolcsi 1994: 186)
d Mari-nak a kalap-ja
Mary-DATIVE the hat-3SG
‘Mary’s hat’ (Hungarian, Szabolcsi 1983)
(5) a The papers report on every move of the actress.
b *The papers report on the every move of the actress.
c The papers report on the actress’s every move.
d The papers report on her every move.
TSC02 07/18/2005, 03:25PM154
Lexical Projections and Functional Projections 155
3 Lexical Projections and Functional
Projections
Discussion
Contents
0 Introduction: Scope of the Chapter 157
1 The Structure of the Sentence 158
1.1 The head of the sentence 158
1.2 Sentences without auxiliaries 159
1.2.1 The link between subject and VP 159
1.2.2 The subject requirement 160
1.2.3 The relation between VP and inﬂection 161
1.2.3.1 INFL as the head of the sentence 161
1.2.3.2 Inserting do in I 164
1.2.3.3 Associating V and I 166
1.2.3.4 When the inﬂection cannot lower onto V 169
1.2.4 Verb positions and comparative syntax 171
1.2.4.1 French questions 171
1.2.4.2 Verb positions in French and English 174
1.2.4.3 Inversion 177
1.2.4.4 Negative sentences 179
1.3 Non-ﬁnite clauses with to 181
1.3.1 A ﬁrst hypothesis 181
1.3.2 Clause types: Finite clauses vs. non-ﬁnite clauses 183
1.3.3 Some predictions 184
1.3.3.1 Co-ordination with to 184
1.3.3.2 The split inﬁnitive 185
2 Building Structure by Merge and Move 186
2.1 Summary so far 186
2.2 Merge 189
2.3 The operation “Move” 190
TSC03 07/18/2005, 03:26PM155
156 Chapter 3
3 Meaning Relations and Structure 191
3.1 Thematic roles 191
3.2 Linking thematic roles and syntactic positions 193
3.3 Relating the subject to the verb 194
3.4 Auxiliaries vs. lexical verbs 195
3.5 Lexical head vs. functional head 196
4 Multiple Auxiliaries 197
4.1 A ﬁrst hypothesis: Clustering auxiliaries 197
4.2 A second hypothesis: Auxiliaries as heads of VP 199
4.3 Finite auxiliaries 202
4.4 The structure of auxiliary sequences 206
5 Summary 209
TSC03 07/18/2005, 03:26PM156
Lexical Projections and Functional Projections 157
0 Introduction: Scope of the Chapter
In Chapter 2 we ﬁrst established a set of diagnostic tools for identifying constituents
in the sentence. Using these tools we then elaborated a representation of the structure
of sentences with auxiliaries. The analysis led to the proposal that in sentences
with auxiliaries the auxiliary is a pivotal element that creates the link between
subject and VP.
In this chapter we complete the representation of the structure of the sentence. In
particular we address the question whether sentences without auxiliaries can be
argued to be formed along the lines developed in the preceding chapter. One point
that will emerge is that in sentences without auxiliaries the inﬂectional ending of the
verb is the pivotal element to relate subject and verb phrase. Based on this observation
we will propose an analysis that covers both sentences with auxiliaries and
those without. The structure we elaborate is one in which VP, the projection of the
verb, is dominated by IP, a projection of the inﬂection. While the verb is a lexical
head, the inﬂection is a functional head. VP is a lexical projection, IP is a functional
projection.
We will see that the format for sentence structure elaborated for English can be
extended to other languages, and we will provide an account for the difference
between the position of the verb in French and in English.
Having identiﬁed the components of the sentence, our next step will be to take
the sentence apart and examine how the components are put together. That is,
we will examine how we form or “generate” sentences by putting together smaller
units of structure. We will see that when putting together words to form a sentence,
we make use of two basic operations: the operation Merge, which combines two
units, and the operation Move, which selects a constituent in an existing structure
and moves it to another position. When building the sentence we start with a verb
and we progressively add constituents to build a more elaborate structure. Because of
its central role in the make-up of the sentence, we will brieﬂy consider the semantic
relation of the verb to the remainder of the sentence. For instance, when a verb
denotes an action, this will imply that there will be one or more participants to the
action. We will look at how such participants are realized in the sentence. After having
discussed the structure of sentences with no auxiliary as well as those with one
auxiliary, we will ﬁnally examine sentences that contain more than one auxiliary.
The chapter is organized as follows: section 1 elaborates a proposal for the
representation of sentences without auxiliaries. Section 2 is an inventory of the
components of the system elaborated so far. Section 3 concerns the semantic relation
between the verb and some of the components of the sentence. Section 4 deals
with sentences with multiple auxiliaries. Section 5 is a summary.
TSC03 07/18/2005, 03:26PM157
158 Chapter 3
1
See section 3.1.2.3 of Chapter 2.
1 The Structure of the Sentence
1.1 The head of the sentence
Diagram (1a) reproduces the general format for syntactic structure that we elaborated
by looking at the noun phrase. In diagram (70b) of Chapter 2, repeated here as
(1b) we molded our previous ﬁndings for the structure of the sentence according to
the format in (1a).1
We tentatively proposed that the head of the sentence is the ﬁnite
auxiliary. Aux selects a VP as its complement and it takes the subject as its speciﬁer.
(1) a
complement
X′
X
head
XP
speciﬁer
phrase, maximal projection
intermediate projection
b AuxP
Aux′NP
The customer
Aux VP
V NP
orderedhas the drinks
According to representation (1b), a sentence is a projection of an auxiliary. The
auxiliary is the pivotal element of the sentence: it ﬁrst combines with its complement,
a VP, which denotes some action/situation. The combination of an auxiliary
and a VP is not yet a sentence. Following the schema in (1a) we label it Aux′. This
newly formed constituent [auxiliary + VP] combines with the subject. The auxiliary
is represented as a linking device that brings together a situation (an activity, a state
of affairs) and the entity to which the situation applies. In addition, the auxiliary
qualiﬁes the validation of the link in terms of time and modality, that is, it indicates
whether the link is valid in the present, in the past, in the future, or whether the link
is probable, possible, etc.
TSC03 07/18/2005, 03:26PM158
Lexical Projections and Functional Projections 159
If we deﬁne the auxiliary as a linker between the VP and the subject we can also
account for the observation that sentences must have subjects. A sentence lacking a
subject would be “unbalanced” in that the auxiliary could no longer function as a
linker between its VP complement and another constituent. Metaphorically speaking,
we could say that the absence of a subject creates an “imbalance,” which would
cause the sentence to “topple over” and crash. Another way of looking at this idea
is to say that the subject serves to anchor the content of the sentence. Very often the
information provided by the subject determines what the sentence is about. If a
sentence lacked a subject, the information in the sentence would be unanchored and
drift along without being able to be linked to other informational units.
However, in our account one immediate problem remains. If sentences with
auxiliaries are projections of Aux, then how do we deal with sentences lacking an
auxiliary? Are these fundamentally different? We turn to this issue now.
1.2 Sentences without auxiliaries
1.2.1 THE LINK BETWEEN SUBJECT AND VP
Let us compare a variant of the test sentence used in the previous chapter, (2a), with
a variant without an auxiliary, (2b). Bracket (2a) in order to show its structure,
using the hypothesis that sentences are projections of Aux. The result is (2c).
(2) a The customer in the corner will order the drinks.
b The customer in the corner ordered the drinks.
One might assume that the structure of (2b) is as in (2d): a sentence (S) combines an
NP and a VP.
(2) c [AUXP [NP The customer in the corner] [AUX′ [AUX will] [VP order the drinks]]].
d [S?? [NP The customer in the corner] [VP ordered the drinks]].
If, as postulated in this book, the structure of the sentence determines its interpretation,
(2d) suggests that sentences without auxiliaries are different in interpretation
from those with auxiliaries. For sentences with auxiliaries, such as (2a), representation
(2c) shows that the contribution of the auxiliary is to link the VP and the
subject. The auxiliary also qualiﬁes the validation of that link. For instance the
auxiliary may specify that the link is validated in the future (will), or that is situated
in the past (has), etc. According to (2d), when there is no auxiliary, a sentence
would simply be a juxtaposition of an NP and a VP. Crucially, there is no linker to
encode the relation between the two elements. This means in turn that the relation
between subject and VP cannot be modiﬁed with respect to factors such as time or
modality. Compare (2a) and (2b). Do you see a way in which the relation between
the subject and the VP can be said to be modiﬁed in (2b)? At ﬁrst sight, (2b) is not
TSC03 07/18/2005, 03:26PM159
160 Chapter 3
unlike a sentence such as (2e), in which the auxiliary have serves to locate the event
in the past:
(2) e [AUXP [NP The customer in the corner] [AUX′ [AUX has] [VP ordered the drinks]]].
The perfect auxiliary have and the past tense in English have a similar (though not
identical) role to play: both place the event expressed in the sentence in the past
time sphere. But this is not what representation (2d) shows: in this representation
the expression of past time, the past tense ending -ed, is fully integrated as a part of
the VP. To represent the linking role of the past tense, we could adopt a representation
such as (2f) with a past time linker.
(2) f [S [NP The customer in the corner] [past] [VP order the drinks]].
Observe also that what is expressed by an auxiliary in one language may be rendered
by an inﬂection in another and vice versa. For instance, French uses an inﬂectional
ending to create a future tense (3a), while English uses the auxiliary will to express
future time:
(3) a Pierre achètera le journal demain.
Pierre buy-fut-3sg the newspaper tomorrow
‘Pierre will buy the paper tomorrow.’
b Peter will buy the paper tomorrow.
Similarly, the present perfect is expressed by means of an auxiliary in English but by
an afﬁx in Latin (3c). In both cases, English is the language using auxiliaries where
another language uses inﬂection.
(3) c ama-vi
love-pf-1sg
‘I have loved’
Recall that our representations reﬂect how meaning is conveyed through structure.
We would perhaps be missing something if we were to treat the inﬂectional endings
of the verb as quite distinct from auxiliaries, since they convey similar interpretations.
1.2.2 THE SUBJECT REQUIREMENT
We suggested relating the fact that every sentence has a subject to the hypothesis
that the auxiliary is the head of the sentence combined with the assumption that the
auxiliary functions as a link between a VP and a subject. Sentences without auxiliaries
and with tensed verbs also have to have subjects. However, if such sentences are
not constructed around a linker, then, as things stand, we do not have any means to
TSC03 07/18/2005, 03:26PM160
Lexical Projections and Functional Projections 161
predict that these sentences too have to have subjects. We could of course devise a
second explanation to account for the fact that sentences without auxiliaries have to
have a subject. This would mean that we end up with two different explanations for
the same pattern, namely that all ﬁnite sentences need subjects. It would not seem
to be in the spirit of the scientiﬁc enterprise to come up with two explanations for
what seems to be one pattern. One comprehensive explanation that covers both
sentences with auxiliaries and those without would be preferable for reasons of
theoretical economy.2
1.2.3 THE RELATION BETWEEN VP AND INFLECTION
1.2.3.1 INFL as the head of the sentence
In this discussion we are concerned with ﬁnite sentences; that is, sentences containing
a lexical verb or an auxiliary in the present or past tense.3
Let us examine
representation (2d) in more detail, focusing on the content of the VP constituent. A
tree representation for (2d) is (2g). This representation suggests that in sentences
without an auxiliary and with a tensed verb we simply merge an NP with a tensed
VP. What predictions do we make with respect to operations such as movement, or
ellipsis, of the VP? Speciﬁcally, how do we expect such operations to affect the
inﬂection -ed on the verb?
(2) g S
VPNP
The customer in the corner
V NP
ordered the drinks
In (2d/g) the -ed ending of the verb is solidly fused with the V, hence it is part of the
VP. We predict that whatever operation affects the VP should therefore also affect
the ending on the verb, as this is fully integrated in the V and the VP. Let us see if
this prediction is borne out.4
Consider the underlined sentences in (4) paying particular
attention to the location of the tense inﬂection in relation to the verb that
it modiﬁes.
(4) a “But I couldn’t rewind time, I just had to get over it.” And get over it she
did. (Guardian, 6.9.2001, p. 15, col. 8)
2
See also the discussion in Chapter 1, section 1.2.3.
3
Finiteness was discussed brieﬂy in Chapter 1, section 2.2.2.
4
Exercise 1 is a revision of constituency tests.
TSC03 07/18/2005, 03:26PM161
162 Chapter 3
b We were told journalism is a science. It didn’t make sense then nor does it
now. (Adapted from the Washington Post, 29.4.2003, p. A22, col. 4).
c Many astronauts have a ham radio license; so does she. (Washington Post,
29.4.2003, p. A10, col. 3)
In (4a) the VP get over it is moved to an initial position.5
Observe that the VP is not
fronted with an inﬂectional ending; we do not front the string got over it, but rather
we front the string get over it. Where is the tense marker whose interpretation
would have to be associated with get (over it)? The tense ending is seen to be detached
from the verb get itself and it is realized on did, an auxiliary that remains unaffected
by the movement. In (4b) the VP has been omitted. But the third person inﬂection
survives on the auxiliary do. In (4c) so replaces the VP, the third person inﬂection is
realized on do.
The possible separation of the verb and its inﬂection can also be demonstrated on
the basis of (5a), the variant of our test sentence without an auxiliary.
(5) a The customer in the corner ordered the drinks.
b What the customer in the corner did was [VP order the drinks].
c Everyone expected that Bill would order the drinks. And [VP order the drinks],
he did.
d The customer in the corner ordered the drinks and so did his companion.
In the pseudo-cleft pattern in (5b) the VP is separated from the inﬂection. Similarly,
in (5c) the fronted VP does not carry the tense inﬂection along, and in (5d) so
replaces the VP but the tense ending is realized on did. Representation (2g) does not
lead to the prediction that the verb and its inﬂectional ending can be separated. In
order to account for the various patterns illustrated above, we need to ﬁnd a way of
isolating the inﬂectional morpheme associated with the V from the VP. This conclusion
is fully in line with the conclusion based on an informal semantic analysis
of the sentence represented in (2f), repeated here for convenience as (5e), in which
the past tense was isolated as a constituent separate from the VP.
(5) e [S [NP The customer in the corner] [past] [VP order the drinks]].
In a present tense sentence we can similarly set apart the third person singular
inﬂection:
(6) a The customer in the corner orders the drinks before the meal.
b What the customer in the corner does is [VP order the drinks before the
meal].
5
If you would like to know which position the moved VP occupies, turn to Exercise 14 of
Chapter 2 and to Exercise 3 of this chapter.
TSC03 07/18/2005, 03:26PM162
Lexical Projections and Functional Projections 163
c Everyone always expects that Bill will order the drinks before the meal. And
[VP order the drinks before the meal], he invariably does.
d The customer in the corner orders the drinks before the meal and so does his
companion.
e [S [NP The customer in the corner] [present] [VP [VP order the drinks] before
the meal]].
(7a) offers a representation covering both (5e) and (6e): INFL stands for the inﬂection
of the verb.
(7) a [S [NP The customer in the corner] [INFL] [VP [VP order the drinks] before the
meal]].6
For sentences with an auxiliary, we have proposed representation (7b):
(7) b [AUXP [NP The customer in the corner] [AUX′ [AUX will] [VP [VP order the drinks]
before the meal]]].
The two representations suggest that there is still a considerable difference between
the two sentence types, that with and that without an auxiliary, which seems unwarranted
by the observations above. Crucially, in representation (7a) the sentence
splits into three immediate constituents, the subject NP, INFL, and the VP. (7a)
has not been adapted to the binary branching format. Let us try to reduce the two
representations to just one. Observe that in ﬁnite sentences with auxiliaries, the
auxiliary is itself inﬂected for tense. In other words, the auxiliary carries ﬁnite
inﬂection. The following examples illustrate this point:
(8) a The customer in the corner is ordering the drinks now.
b The customer in the corner was ordering the drinks later.
c The customer in the corner has ordered the drinks already.
d The customer in the corner had ordered the drinks already.
All ﬁnite sentences contain some form of inﬂection, but not all ﬁnite sentences
contain an auxiliary. Pursuing this path and taking into account that in the head
position labeled Aux we insert an inﬂected auxiliary, we can combine (7a) and (7b).
Rather than labeling the head position of the sentence Aux, we could label it INFL
(or I). The head position INFL will host either the inﬂected auxiliary or just the
inﬂectional ending of the verb. Based on this hypothesis, the structure for the
sentence is then relabeled as in (7c):
6
An alternative would be to use tense as a cover label.
(i) [NP The customer in the corner] [tense] [VP order the drinks before the meal].
TSC03 07/18/2005, 03:26PM163
164 Chapter 3
(7) c IP
I′NP
The customer
in the corner
I VP
will
had
-ed
-s
order the drinks before the meal
ordered the drinks before the meal
order the drinks before the meal
order the drinks before the meal
The head of the structure is I. I combines with VP to form I′; the intermediate
projection. I′ combines with the subject to form IP.7
The label IP stands for Inﬂection
Projection or, in other words, a phrase headed by the inﬂection. We have
managed to isolate both the auxiliary and the tense inﬂection from the verb and give
them the same pivotal linking role. The resulting representation shows that:8
(i) Sentences are centered around I.
(ii) I is a linker: it links a VP and a subject, both components are obligatory.
(iii) The content of I qualiﬁes the linking, locating it in time and/or assigning some
modal value to the relation.
1.2.3.2 Inserting do in I
As it stands, the structure in (7c) above does not yet allow us to form a sentence
without an auxiliary. An inﬂectional ending of a verb cannot stand all by itself
because it is a bound morpheme; it is an afﬁx that must be hosted by a free morpheme.
The past tense inﬂectional ending, -ed, must be attached to a verb. One way
of saving an unattached -ed ending is by associating it with its own auxiliary. This
is what happens when the auxiliary do is inserted. If we insert do under a head I
with a past tense ending we derive (9):
(9) a [IP [NP The customer in the corner] [I′ [I do + ed] [VP order the drinks before
the meal]]].
b [IP [NP The customer in the corner] [I′ [I did] [VP order the drinks before
the meal]]].
7
In the alternative representation (note 6) in which we represent the core constituent of the
sentences as Tense, a sentence is a projection of T: TP.
(i) [TP [NP The customer in the corner] [T′ [tense] [VP [VP order the drinks] before the meal]]].
8
Exercise 2.
TSC03 07/18/2005, 03:26PM164
Lexical Projections and Functional Projections 165
In (9) do is not essential.9
We can dispense with it if we attach the tense ending to V:
(9) c The customer in the corner ordered the drinks before the meal.
In (10a) do cannot be eliminated, as shown by (10b):
(10) a And [VP get over it] she did. (cf. (4a))
b *And [VP got over it] she.
In (10a) did combines the past tense inﬂection with do. This sentence illustrates a
leftward movement of the VP, get over it. As a result of this movement, the VP no
longer occupies its canonical position to the right of the auxiliary. Apparently,
when we move the VP, insertion of do becomes essential in the sentence. There is
no way that we can realize the tense on the verb itself (10b). If we restore the VP to
its original position, though, we can dispense with do:
(10) c And she did [VP get over it].
d And she got over it.
In our constructed example (9), do is also not essential. If language is guided by a
principle of economy though,10
the insertion of do should have some effect. If do
were completely superﬂuous, by the principle of economy we should not insert it.
To assess the possible impact of inserting do in a context where it could be omitted,
let us look at some attested examples in which do could also have been absent.11
(11) a I don’t remember much of anything she said in the church foyer or what I
uttered back. She had that dazzling effect on me . . . What I do recall is that
she invited me to a holiday party two nights later at the mutual friend’s
place. (Chicago Tribune, 22.12.2003, section 13, p. 9, col. 1)
b Coleman, who describes himself as a “semi-professional punter”, gave evidence
at a trial in Southampton in October 2001 and his statements to the
court then will form the basis of the case against him. It is still not clear if
he will turn up for the 10 a.m. hearing at the club’s headquarters in London,
but the feeling at Portman Square yesterday was that he would indeed appear
to defend himself. . . . If Coleman does appear this morning, the Jockey Club
may also wish to inquire about another part of the evidence. (Guardian,
22.1.2003, p. 14, cols 1 + 2)
c On Tuesday Clarett disputed the contention of university ofﬁcials that he
had failed to ﬁle the proper paperwork that would have allowed him to
attend the funeral . . . Each side is right, Clarett did ﬁll out the papers but
9
See Chapter 1, section 2.2.3 for this use of do.
10
As discussed in Chapter 1, section 2.2.3.
11
For more examples of this use of do see Chapter 1, Exercise 7.
TSC03 07/18/2005, 03:26PM165
166 Chapter 3
ﬁlled them out too late to receive tickets to ﬂy home. (New York Times,
2.1.2003, p. D1, col. 1)
d Jackson is hardly a virgin forest. Like most of the state’s redwood land, it
has been logged intermittently since about the middle of the 19th century
. . . But the forest does have thousands of acres of 80- to 100-year old
redwoods. (San Francisco Chronicle, 28.11.2002, p. A34, col. 1)
If we eliminate the underlined occurrences of do in (11) we have to associate the
inﬂectional ending of do with the verb itself. In (12) we reproduce the crucial parts
of the contexts in which the sentences with do are used. Can you identify a common
contextual factor for all these examples?
(12) a I don’t remember much of anything she said.
b It is still not clear if he will turn up for the 10 a.m. hearing at the club’s
headquarters in London.
c the contention of university ofﬁcials that he had failed to ﬁle the proper
paperwork
d Jackson is hardly a virgin forest.
By inserting do, the writer/speaker forcefully asserts the validity of the linking of
the subject and the related event in a context in which this linking is not obvious.
The contextual elements in (12) provide indications that the linking of the subject
and the VP, which is emphasized by the insertion of do, is not self-evident and that
it merits being highlighted. Inserting do highlights that link. This interpretation of
the role of inserting do is in line with our proposal that the inﬂectional position of
the sentence (I) encodes the linking between the subject and the VP. By inserting
do in that position, we highlight the validation of that link.
1.2.3.3 Associating V and I
In (13a) identify the subject NP, the VP, the verb, and any VP-adjunct(s). Rephrase
the sentence (i) by replacing the past tense form of the verb ﬁrst by a combination
of the auxiliary have + past participle, then (ii) by reorienting the sentence toward
the future, using the auxiliary will with the inﬁnitive.
(13) a The department later modiﬁed the advice.
In the variants of (13a) that you will have created, the adjunct later remains to the
left of the lexical verb, and the auxiliary is adjacent to the subject.
(13) b The department has later modiﬁed the advice.
c The department will later modify the advice.
Adopting the format displayed above, the tree diagram for (13) will be as in (14),
with alternative ﬁllers for I.
TSC03 07/18/2005, 03:27PM166
Lexical Projections and Functional Projections 167
(14) IP
VP
NP
AdvP
later
later
later
the advice (13a)
the advice (13b)
the advice (13c)
modify
modiﬁed
modify
a The department
b The department
c The department
NPV
I′
I VP
-ed
has
will
In example (13a) the tense ending -ed is located on the verb modify. So in representation
(14a) we need to combine tense and verb. Let us examine how this is
achieved. V and I are head positions. They contain heads, i.e. elements that form the
nucleus of a projection. In (14a), the content of the head I is the inﬂectional ending
-ed; the content of V is the lexical verb modify. These two heads need to be combined
because -ed is a bound morpheme that has to attach to a verb. Presumably,
looking at the tree, there are three options. One is that the verb modify is moved
from its position V to the position I, where it combines with -ed (15a).
(15) a IP
VP
NP
AdvP
later the adviceThe department
NPV
I′
I VP
modify + ed
A second one is that the ending -ed is moved onto the verb (15b).
TSC03 07/18/2005, 03:27PM167
168 Chapter 3
(15) b IP
VP
NP
AdvP
later the adviceThe department
NPV
I′
I VP
modify + ed
There is a third alternative, namely that we create a new position Z in the structure,
perhaps between V and I. V and I would be combined in that new position. If the
new position were also a head position, which we could label Z, then we would
actually have to deduce that there is a complete new projection ZP in the structure.
Should we actually pursue this third proposal?
We have postulated the head positions I and V. In the two options laid out above
(15a, b), we need only those two positions plus a mechanism that moves one of the
constituents from its own position to the position of the other. In the third option
outlined above, we also have to move the inﬂection and the verb. Therefore, we
still need this additional mechanism for moving a constituent from one position to
another. But in addition to that, we also need to create a totally new position inside
the structure for the inﬂection and the verb to unite. A second difference between
(15a, b) and the third option is that according to the proposals in (15a) and (15b),
either the verb moves (15a) or the inﬂection moves (15b), whereas in the third option,
both the verb and the inﬂection move. The third option then is less economical: it
requires creating a new head position (and the related projection) inside the sentence
and it requires moving two heads instead of just one. Because it is less economical, we
will therefore only resort to this option if we can show that the more economical alternatives
(15a, b) are both inadequate. In general, if we can arrive at an analysis using
fewer tools, then we prefer to adopt that rather than adding extra tools to the analysis.
Remember that in the spirit of a scientiﬁc approach to syntax, we should aim for
theoretical simplicity.12
An analysis requiring fewer tools is always more highly valued.
So before examining the third alternative in which a new projection is postulated,
we will ﬁrst compare (15a) and (15b). Could we propose that the lexical verb moves
to the inﬂection? Consider tree diagram (15a). What would be the resulting linear
order? If V were to move to I, as proposed, then we ought to obtain (15c):
(15) c *The department modiﬁed later the advice.
12
See Chapter 1, section 1.2.3.
TSC03 07/18/2005, 03:27PM168
Lexical Projections and Functional Projections 169
But (15c) is not acceptable. The verb must follow the adjunct later, which we
assume is adjoined to VP.
Could the inﬂection move onto the verb? What would be the resulting word
order? If I moves to V, then we obtain (15d):
(15) d The department later modiﬁed the advice.
(15d) is acceptable. The verb follows the adjunct later, which we assume is
adjoined to VP.
Let us look at the geometry of the tree and in particular, let us look at the relative
positions of the heads V and I in the tree diagram representation. From a geometrical
point of view, which is higher, V or I?
(15) e IP
VP
NP
AdvP
NPV
I′
I VP
I combines with VP. I and the projection VP are on equal footing, both are immediate
constituents of I′.13
V is buried inside VP. V is an immediate constituent of the core
VP, so V hierarchically is ranked lower than the augmented VP. This means that
the head V is lower than the head I. When I joins V it “goes down” into VP. We
say that in English I lowers onto V. Conversely, I is higher than V. If the verb were
to move to I, then it would have to leave VP and go up in the tree. In that case, we
would say that it raises to I.14
As shown above, in English I lowers onto the verb
and the verb cannot raise to I.
1.2.3.4 When the inﬂection cannot lower onto V
Suppose that we pick the VP in representation (15e) and move it to a position to the
left of the subject. Leftward movement of the VP should not affect the content of
the head I, which is inserted independently of the content of the VP. In (16a, b),
the VP is moved to the left of the subject. The subject is the speciﬁer of IP, the
maximal projection of I. VP is also a maximal projection. Not every English sentence
13
For immediate constituents see Chapter 2, section 2.4.2.
14
We will discuss instances of V raising to I in sections 1.2.4 and 4.3 below. It is possible to
rephrase the theory in a way that avoids lowering. See Chomsky (1995) and Adger (2003) for
an introduction. We will not go into this issue here.
TSC03 07/18/2005, 03:27PM169
170 Chapter 3
contains such a fronted VP. VP-fronting (or VP-preposing) is an optional operation
that combines a VP with a clause. To represent the resulting structure, we propose
that the fronted VP adjoins to IP. (16a) is a tree diagram representation, (16b, c)
translate (16a) in the equivalent labeled bracketing representation.15
(16) a IP
I
has
will
VP IP
NP I′
VP
the department
the department
Modiﬁed the advice
Modify the advice
b [IP [VP Modiﬁed the advice], [IP [NP the department] [I′ has [——]]]].
c [IP [VP Modify the advice], [IP [NP the department] [I′ will [——]]]].
Let us apply VP-fronting to a sentence without an auxiliary. If we front the VP and
there is no auxiliary under I, we will be left with an inﬂection stranded under I. But
the inﬂection is a bound morpheme and has to combine with a verb. Lowering the
inﬂection onto V, as we did before, now raises a problem. After movement of the
VP, there will actually be no verb to lower the inﬂection onto (16d). Left alone,
the inﬂection is also unable to survive in I because it is a bound morpheme. In such
cases, we have to “rescue” the sentence by enabling the inﬂection to survive independently
of the lexical verb in the verb phrase, and we insert the auxiliary do (16e).
(16) d eIP
IPVP
I′NP
VPI
-ed ???
IP
IPVP
I′NP
VPI
Do + ed
15
For some discussion of this issue see also Exercise 15 in Chapter 2. In Chapters 4 and 5 we will
return to the operation that moves constituents from one position to another position in the
sentence.
TSC03 07/18/2005, 03:27PM170
Lexical Projections and Functional Projections 171
In (16a) VP movement does not give rise to do-insertion. Can you explain why this
should be? The answer is that in these examples, insertion of do is not required to
rescue the sentence because the problem of the stranded inﬂection does not arise.
The auxiliary is a free morpheme. The inﬂection is associated with the auxiliary, will
is a present tense form (as opposed to would, which would be the past tense form).
Since there is no need for inserting do we do not insert it. In VP-fronting contexts,
do is inserted to rescue a pattern which would otherwise not survive. We can say
that these patterns illustrate the effect of economy on the structure.16
We only insert
do if we cannot manage without it.
Recall that we examined the formation of questions in Chapter 1, section 2, and
we observed that direct questions involve subject-auxiliary inversion, abbreviated
as SAI. We also saw that we insert do and invert it with the subject whenever the
relevant sentence lacks an auxiliary. Identify SAI patterns in (17):
(17) a What are my borrowing options? How much can I afford? Where do I
begin? (New York Times, 28.4.2003, p. A22, advertisement Fleet)
b The key question is: what do we know about the 96 new cases of SARS,
who are the 96, when did they start, where did they live? (Based on Washington
Post, 29.4.2003, p. 18, col. 1)
We should again ask ourselves why do is inserted in some of the examples, and if
possible we should try to relate the answer to the preceding discussion of the insertion
of do. We saw in Chapter 1 that SAI is related to question formation. For sentences
without an auxiliary, we would again want to end up saying that do is inserted as a
last resort to rescue the structure. We have assumed that the auxiliaries are inserted
in I and move to the left across the subject in cases of subject-auxiliary inversion.
The position they arrive at, or their landing site, will be examined in Chapter 5. We
should ask ourselves, though, which component is crucially targeted by SAI: is it
the auxiliary itself or is it the inﬂectional element that is part of the auxiliary? In
order to answer this question we will ﬁrst extend our database and look at question
formation in French.
1.2.4 VERB POSITIONS AND COMPARATIVE SYNTAX
1.2.4.1 French questions
In Chapter 1 we saw that French also displays SAI:
(18) a A-t-elle toujours acheté le journal?
has she always bought the paper
‘Has she always bought the paper?’
b A-t-elle toujours écrit des romans?
has she always written novels
‘Has she always been a novelist?’
16
See Chapter 1, section 2.2.3.
TSC03 07/18/2005, 03:27PM171
172 Chapter 3
But in contrast with the English pattern, when there is no auxiliary, direct questions
in French can simply be formed by inversion of a tensed lexical verb and the subject.17
We labeled this pattern subject-verb inversion or SVI. Why is there a difference
between French and English?
In order to understand the difference, we need to compare English sentences with
their French counterparts. Unless we are proven wrong, considerations of economy
lead us to assume that French sentences essentially have the same structure as English
sentences, and, in particular, that French verbs also project a VP. The verb would be
merged in the position V. We call the position in which the verb is inserted its base
position.
(19) IP
I′NP
VPI
NP
le journal
the paper
des romans
novels
V
achete
buy
écri
write
-ait
PAST
-t
PRES
a Elle
she
b Elle
she
Consider the examples in (20):
(20) a Achetait-elle toujours le journal?
a′ *Bought she always the paper?
b Ecrit-elle encore des romans?
b′ *Writes she still novels?
In (20a) the verb achetait (‘bought’) cannot be occupying its base position, the head
position of VP. If the verb were in its base position, it should be adjacent to its direct
object le journal (‘the paper’). Similarly, in its base position the verb écrit (‘writes’)
would be adjacent to the direct object des romans (‘novels’) in (20b). In (21a) and in
(21b), which are the acceptable counterparts to (20a′) and (20b′), the English verbs
buy and write occupy their base positions.
(21) a Did she always buy the paper?
b Does she still write novels?
17
See Chapter 1, section 3.1.1.
TSC03 07/18/2005, 03:27PM172
Lexical Projections and Functional Projections 173
Why does English need do insertion in (21) and why does French not need a similar
device in (20)? The difference cannot be due to the fact that SAI itself is restricted to
English questions given the French data in (18), which also display subject-auxiliary
inversion.
A further question arises for the French examples in (20), in which the verb inverts
with the subject. Recall that we proposed as a working hypothesis that language
is driven by economy: we do not insert elements unless they are needed. In the
examples discussed so far, English do-insertion was motivated either by the need to
highlight the link between subject and VP (in I), or, alternatively, because VP fronting
had made the verb inaccessible to the inﬂection in I. In the latter case, there exists
no alternative without do. Given the French data above, though, we can ask ourselves
why English doesn’t simply use the French strategy of inverting the verb with the subject,
SVI, to form questions. Or, conversely, why does French use SVI? Formulating
things in yet another way, how come French can use SVI at all and English cannot?
It could be that the contrast between English and French is to be directly related
to the process of question formation as such, and that therefore we have to explain
the phenomena in relation to questions. But it could also be that the observed
contrast is due to a more general difference with respect to the behavior of the verbs
in the two languages considered. In the latter scenario, we would expect to ﬁnd
other patterns in which English and French differ with respect to the position of the
verb. In English, the verb occupies its base position next to the object in direct
questions while in French the verb moves away from that position. Are there other
situations in which the two languages differ in a parallel way, that is, in which the
English verb remains in its base position while a French verb moves away? In the
next section we will look at some such patterns.
The goal of the current discussion is to arrive at an explanation for the observed
difference between English and French question formation. Ideally, we want to
provide an account for this difference which relates it to some other property of
the languages. We need to ask ourselves: “Could things have been different?” For
instance, could English question formation have been more similar to French question
formation? As we saw in Chapter 1,18
the answer to both these questions is
positive. English used to be different, and question formation in Old English used to
be similar to question formation in present-day French:
(22) a Hwæt sægest qu, yrdling?
what say you, peasant
‘What do you say, farmer?’
(Acoll, 22, Van Kemenade 1987: 138)
b Hu begæst qu work qin?
how beget you work your
‘How do you carry out your work?’
(Acoll, 22, Van Kemenade 1987: 138)
18
Section 3.2.
TSC03 07/18/2005, 03:28PM173
174 Chapter 3
c Hu lomp eow on lade?
how happened you on journey
‘How did you fare on the journey?’
(Beowulf 1987–1988, Pintzuk 1991: 138)
We do not want an account that implies that English could not have been different
from what it is today. We do not want to postulate that there is an absolute divide
between languages such as French, which can form questions by moving a verb
across the subject, and languages like English, which cannot. Whatever explanation
we propose in order to differentiate French and Modern English should ideally also
account for the difference between Modern English and Old English and it should
explain why English has changed.
1.2.4.2 Verb positions in French and English
Recall that we propose that in an English sentence the lexical verb remains under V.
The inﬂected auxiliary sits under I. The verbal inﬂection is inserted under I and
lowers onto V. This proposal correctly derives the observed patterns:
(i) Subject – inﬂected auxiliary – adjunct – V.
(ii) Subject – adjunct – inﬂected V.
Diagram (23) shows the structure.
(23) IP
I′NP
VPI
VPAdvP
will
-ed
a John
b John
= John
NPV
always
always
always bought the paper.
buy
buy
the paper
the paper
Now consider the French equivalents of John always bought the paper. The French
verb is also the semantic nucleus of the VP; it is also ﬁrst merged with its complement
to create a core VP. Identify subject, verb, object, and adjunct in (24a), using
the English glosses provided. We assume that the adjunct toujours (‘always’) is like
TSC03 07/18/2005, 03:28PM174
Lexical Projections and Functional Projections 175
its English counterpart and that it is VP-adjoined. This seems to be a reasonable
assumption since toujours modiﬁes the sentence in a similar way to the English
adverb always. What conclusions do we reach concerning the position of the verb
achetait (‘bought’) in (24a)? Could it be in V?
(24) a Jean achetait toujours le journal.
John bought always the paper
‘John always bought the paper.’
If the French object NP le journal (‘the paper’) is in its base position, then the verb
achetait (‘bought’) cannot be claimed to occupy its base position, V. Let us try to
plot the position of the verb. We assume that the underlying structure of French is as
in English, as shown in (24b). How could we derive the French word order in (24a)?
(24) b IP
I′NP
VPI
VPAdvP
-aitJean
NPV
toujours achete le journal
If English I and V unite by I-lowering onto V, giving the order adjunct-verb, then the
simplest way to account for the observed French order verb-adjunct is to have V
raise to I:
(24) c IP
I′NP
VPI
VPAdvP
achet + -aitJean
NPV
toujours le journal
TSC03 07/18/2005, 03:28PM175
176 Chapter 3
Table 1 Present and past tense paradigms for French and English
French English
Present tense acheter buy
First person singular J’achète I
Second person singular Tu achètes You
Third person singular Il achète He
First person plural Nous achetons We
Second person plural Vous achetez You
Third person plural Ils achètent They
Number of different verb forms: . . . . . .
Past tense
First person singular J’achetais I
Second person singular Tu achetais You
Third person singular Il achetait He
First person plural Nous achetions We
Second person plural Vous achetiez You
Third person plural Ils achetaient They
Number of different verb forms: . . . . . .
19
This is a slight simpliﬁcation as some of the forms are pronounced in the same way in French.
For instance, in the present tense the forms achète, achètes, and achètent used to be pronounced
differently in earlier stages of French but in Modern French they sound the same; in
the past tense achetais, achetait, and achetaient will sound the same. But even taking these
forms as non-distinct, both tenses keep three distinct forms.
This derivation corresponds to derivation (15a), which we discarded for English; it
gives us the correct order for French. But of course, this is not the end of the discussion.
The next question we need to address is why the inﬂection lowers onto V in
English and why the verb raises to I in French. In both languages, the reverse
pattern is not available. V cannot move to I in English (24d) and I cannot lower to
V in French (24e):
(24) d *John bought always the paper.
e *Jean toujours achetait le journal.
How could we explain this cross-linguistic difference? We are dealing with a relation
between V and I. Presumably, there are three possibilities: the explanation of
the difference lies in V, or it lies in I, or it lies in the relation between V and I. Let us
look at the inﬂections in the two languages. We give the inﬂectional paradigm for
the French verb acheter (‘buy’) in the middle column of Table 1. Supply the missing
forms for English in the right-hand column. For each tense, how many different
forms do you count? The inﬂectional patterns of French and of English are markedly
different. In English there are two forms in the present tense and there is just one
past tense form. French has ﬁve forms each for present and for past.19
We could try
TSC03 07/18/2005, 03:28PM176
Lexical Projections and Functional Projections 177
to account for the different behavior of the verb in English and in French by building a
hypothesis that takes into account the differences between the inﬂections. Let us propose
ﬁrst that the different surface forms of the inﬂection signal the relative strength
of the inﬂections: the more distinct forms there are in the paradigm, the stronger the
paradigm is.20
Pursuing this concept of strength, we could say that the inﬂection in
French is strong and, that, by virtue of being strong, it can attract the verb, while
the inﬂection in English is weak and that, because of its weakness, it cannot attract
the verb. Because I cannot attract V in English, I will move down to unite with V.21
1.2.4.3 Inversion
Consider now the form of direct questions in French and in English:
(25) a A-t-elle acheté le journal?
has-she bought the paper
‘Has she bought the paper?’
b Achetait-elle le journal?
bought-she the paper
‘Did she buy the paper?’
c *Bought she the paper?
d Did she buy the paper?
e Has she bought the paper?
As we have seen before, in French the ﬁnite auxiliary inverts with the subject (25a).
If there is no auxiliary, the ﬁnite verb inverts with the subject (25b). We conclude
that it cannot be the case that in order to form a question we actually have to invert
an auxiliary. There is, for instance, no inverted auxiliary in (25b); in this example,
inversion affects the inﬂected verb. On the other hand, it also cannot be the case
that question formation depends on the inversion of a lexical verb, since, for instance,
in (25a) only the auxiliary inverts with the subject. So even though inversion is
deﬁnitely related to the formation of questions, neither lexical verbs nor auxiliaries
are necessary ingredients.
We have formulated the hypothesis that the French verb is attracted to I. In (25b)
the verb moves up to a position to the left of the subject. In (25a) the inﬂected
auxiliary moves to the left of the subject. What is the common factor between the
inverted constituents in (25a) and (25b)? What is common between these inverted
constituents is the inﬂection itself. In both cases, the content of the inﬂectional node
I ends up to the left of the subject: in French the content of I is either a verb or
20
See Pollock (1989, 1997), Vikner (1997), and Rohrbacher (1999). For a critical view see
Bobaljik (2002), Alexiadou and Anagnostopoulou (1998).
21
Exercises 6 and 7, also Exercise 18.
TSC03 07/18/2005, 03:28PM177
178 Chapter 3
22
See section 1.2.3.4.
23
See section 1.2.4.1.
an auxiliary. We can speculate that this common factor, inﬂection, is essential in
question formation. In order to signal interrogative force, we move the content of I
to the left of the subject. Whenever French I contains a lexical verb, moving I will
take along the lexical verb (25b). In English, I never contains a lexical verb, so we
cannot derive (25c). If I contains an auxiliary, moving I to the left will invert the
auxiliary with the subject (25e).
If English I fails to contain an auxiliary and if we were to move just I to the left
of the subject we would end up with (25f):
(25) f [I -ed] [IP she —— [VP buy the paper]]?
*-ed she —— buy the paper?
Why does derivation (25f) not yield a possible sentence? The problem is once again
that the inﬂectional morpheme -ed is a bound morpheme and that it cannot stand
unattached. But if we were to lower it into VP and insert it onto V, then we would
no longer signal interrogative force. In the case of VP fronting, we saw that an
unattached inﬂectional ending in I was rescued by do insertion.22
By applying do
insertion to (25f) we derive the acceptable (25d).
The reason why lexical verbs can precede subjects in French questions is related
to the fact that French lexical verbs can move to I. This in turn can be explained by
our hypothesis concerning the relative strength of the inﬂection of the verb.
Recall that we have seen that Old English23
lexical verbs inverted with subjects in
questions. What would be your predictions concerning the inﬂectional paradigms of
Old English: how would they compare to the patterns in Modern English? According
to our hypothesis, the movement of lexical verbs is made possible by strong
inﬂection. Thus we predict that Old English had a strong inﬂection. If you turn to
any grammar of Old English you will ﬁnd that the paradigms of the verb had indeed
more different forms than is the case in current English. For instance, the verb
singan (‘sing’) had the inﬂection in Table 2.
Table 2 Present tense inﬂectional paradigm
for Old English singan (‘sing’)
Person Singular Plural
1 sing-e sing-en
2 sing-est sing-en
3 sing-eQ sing-en
TSC03 07/18/2005, 03:28PM178
Lexical Projections and Functional Projections 179
24
This is a rough approximation. For an introductory discussion see Haegeman and Guéron
(1999). For more careful analysis see Pollock (1989), Haegeman (1995), Rowlett (1998),
Zanuttini (1997a, b), and the references cited there. For discussion of negation in English see
among others Klima (1964), Pollock (1989), Baker (1991), Ernst (1992), Haegeman (1995),
Cormack and Smith (2002), and the references cited there.
1.2.4.4 Negative sentences
Consider (26):
(26) a Jean ne mangeait pas de chocolat.
John ne eat-past-3sg not chocolate.
b *Jean ne pas mangeait de chocolat.
c *John eats not chocolate.
d *John not eats chocolate.
e John does not eat chocolate.
In French, negation is expressed by means of two elements, ne and pas. Roughly,
pas corresponds to English not. The element ne is not always realized. We will not
discuss ne here and we will provisionally locate it in I. Let us assume that pas is like
any adverbial and may be VP-adjoined.24
Draw the tree diagrams for (26a).
(27) IP
I′NP
VPI
VPAdvP
ne mange-aita Jean
NPV
pas de chocolat
The verb mange is attracted to I -ait and will hence cross the VP-adjoined negation
marker pas.
If English not is also VP-adjoined, we can account for the unacceptability of
(26c). To obtain (26c), eat would have to have moved out of VP but we have seen
that English lexical verbs do not leave VP. But (26d), in which V would remain in
the VP and the ending -s would have lowered onto V, is also ungrammatical (27c):
TSC03 07/18/2005, 03:29PM179
180 Chapter 3
(27) IP
I′NP
VPI
VPAdvP
eat-sb *John
c *John
NPV
not
not
chocolate
chocolateeat-s
Apparently, for the formation of negative sentences in English we have to insert
the auxiliary do (26e).
In our earlier discussion, we came across two other contexts in which do
insertion was obligatory. What were these? Do you remember the factors that
triggered do-insertion? We saw that do insertion is needed when a VP is fronted
(28a) and also in direct questions which do not contain an inﬂected auxiliary
(28b).
(28) a And get over her problems, she did.
b Did she get over her problems?
What these two examples have in common is that do rescues a stranded inﬂectional
morpheme which is not able to lower onto V. In (28a), VP has moved leftward
and the morpheme in I cannot lower onto V inside VP any more. In (28b) I has to
move to a position to the left of the subject to encode interrogative force and it
cannot lower onto V. If I did lower onto V, the resulting sentence would not be
interrogative.
(28) a′ Get over her problems she -ed.
Get over her problems she do + ed.
b′ -ed she get over her problems?
Do + ed she get over her problems?
Extrapolating from these examples we could conclude that in negative sentences too
do is inserted because the inﬂection would otherwise be stranded:
TSC03 07/18/2005, 03:29PM180
Lexical Projections and Functional Projections 181
(27) IP
I′NP
VPI
VPAdvP
-s
do + s
NPV
not
not
chocolate
chocolate
eat
eat
d John
e John
If this is on the right track, then we conclude that in (26d) -s is not allowed to lower
onto the verb eat, even though the VP has not moved and -s is in I. The reason why
-s cannot lower onto V presumably has to do with the sentence being negative.
What is the semantic contribution of negation to a sentence? Sentence negation
crucially bears on the relation between the subject and the VP: a negated sentence
signals that the link subject-VP does not obtain. Indeed, recall that in French the
element ne is associated with the inﬂected verb in I. Let us assume that when not
negates a sentence, the position I must remain ﬁlled because it is precisely the link
between subject and predicate that is targeted by the negation. Let us assume that
when I must be ﬁlled, its content cannot lower onto V. The inﬂection will remain
in I and do will be inserted precisely to allow the stranded inﬂection to remain
in I.25
1.3 Non-ﬁnite clauses with to
1.3.1 A FIRST HYPOTHESIS
In this section we turn brieﬂy to sentences which lack a tensed verb or auxiliary.
Such sentences are non-ﬁnite.26
Here, we will only be looking at inﬁnitival strings
such as that underlined in (29b). In preparation for the analysis, draw the tree
diagram representation for the structure of (29a). Then consider the underlined
25
For problems with this proposal see Exercise 5.
26
Since a non-ﬁnite sentence is typically embedded, the term non-ﬁnite clause would be more
accurate. In practice, though, the terms clause and sentence are often used interchangeably in
current syntactic discussions.
TSC03 07/18/2005, 03:29PM181
182 Chapter 3
string in (29b). Is the underlined string a constituent? What is its function? Try to
match the constituents of (29a) with those of the underlined string of (29b).
(29) a The customers in the corner should order the drinks.
b I expected the customers in the corner to order the drinks.
The underlined string in (29b) is a constituent: it can be questioned; it can be
replaced by a demonstrative.
(29) c What did you expect?
d I did not expect that.
The underlined string in (29b) is very similar to sentence (29a): there is a parallelism
between the constituents of (29a) and those of the underlined section of (29b). In
(29a) the lexical V order heads the VP order the drinks; the subject is the NP the
customers in the corner. The same constituents reappear in (29b). Based on the tree
diagram that you have drawn for (29a), try to ﬁt the constituents of the underlined
string in (29b) into a tree. When doing this, the only problem that arises is the location
of to. What would be the simplest way to insert to into a tree representation?
(29) e IP
I′NP
VPI
NPV
should
to
The customers in the corner
the customers in the corner
order
order
the drinks
the drinks
In (29e) to is inserted under I, suggesting it is a ﬁller for the inﬂection of the verb.27
Is to on a par with auxiliaries or is it more like the inﬂectional endings of the verb?
In other words, does to have to attach to V? Is to a bound morpheme or a preﬁx?
27
If we had used Tense rather than I to denote the head of the sentence (footnotes 6, 7), we could
distinguish ﬁnite clauses from non-ﬁnite ones by means of a system of features. Finite clauses
would have a positive value for tense [+tense], non-ﬁnite clauses would have a negative value
[−tense].
(i) [TP [NP The customer in the corner] [T′ [±tense] [VP order the drinks before the meal]]].
TSC03 07/18/2005, 03:29PM182
Lexical Projections and Functional Projections 183
Discuss how the following examples provide evidence that to should not be treated
as a bound morpheme:
(30) a But Vita could not write the last act, because she did not know how to.
(Victoria Glendinning, Vita, 1984: 118)
b He never asked for his attorneys. If he’d wanted to, he could have. (Washington
Post, 29.4.2003, p. B1, col. 6)
In the examples above, a VP has been ellipted. The ellipsis does not affect to:
(31) a because she did not know how to write the last act.
b If he’d wanted to ask for his attorneys.
In (32a) co-ordination affects two VPs, and to is outside the co-ordinated constituents.
(32) a After the discovery of a £200m budget shortfall the BBC is being forced to
cut costs and lay off staff. (Based on Guardian, 7.10.2002, p. 1, col. 1)
b After the discovery of a £200m budget shortfall the BBC is being forced to
[[VP cut costs] and [VP lay off staff]].
The data above suggest that to is neither inseparable from V nor is it an integral
part of the projection of V, VP. In order to conﬁrm the hypothesis that to is not an
inseparable part of a verb and of the related VP, complete the answer to (33a) using
VP ellipsis. Apply VP-co-ordination to (33b).
(33) a Do you expect the customers in the corner to order their drinks?
No, actually I don’t expect . . .
b We expect the customers in the corner to order their drinks and we expect
them to proceed to their tables.
The resulting sentences show that to is not a preﬁx on the verb. Rather, to is a free
morpheme that seems to occupy the same position as ﬁnite auxiliaries and as the
verb inﬂection. We will adopt the hypothesis that to ﬁlls the inﬂection node I of
inﬁnitival clauses.28
(34) a Do you expect the customers in the corner to order their drinks?
No, actually I don’t expect them to [VP order their drinks].
b We expect the customers in the corner to [[VP order their drinks] and [VP
proceed to their tables]].
1.3.2 CLAUSE TYPES: FINITE CLAUSES VS. NON-FINITE CLAUSES
If inﬁnitival to is the ﬁller for I, this means that to ﬁlls the head of the inﬁnitival
clause. As a consequence, the properties of to should determine the properties of the
28
Exercise 4.
TSC03 07/18/2005, 03:29PM183
184 Chapter 3
complete clause. This is indeed the case: the ﬁller of I signals the contrast between
two types of clauses, namely “ﬁnite clauses” and “non-ﬁnite clauses.”
Replace the subject NP the customers in the corner of the examples in (29) by
a pronoun. You will ﬁnd that the subject of the ﬁnite clause (29a) is replaced by a
pronoun in the nominative form they (35a), that of a non-ﬁnite clause in (29b) by
a pronoun in the accusative form them (35b):
(35) a They should order the drinks.
b I expected them to order the drinks.
Recall that the speciﬁer of a constituent has a special relation with the head of that
constituent.29
For instance, the subject of the sentence is located in the speciﬁer
position of I and the subject of a ﬁnite clause agrees with I. (35) shows how the
nature of the head of the clause (ﬁnite/non-ﬁnite) correlates with the case form of its
subject. A ﬁnite I takes a subject with a nominative form (I, you, he, she, etc.); a
non-ﬁnite I takes a subject with an accusative form (me, you, him, her, etc.).
Some verbs select as their complements a speciﬁc type of clause. For instance,
English want requires an inﬁnitival clause and does not select a ﬁnite clause:
(36) a I want [the customers in the corner to order their drinks now].
b *I want [that the customers in the corner order their drinks now].
The choice of the ﬁller for I, to as opposed to a ﬁnite auxiliary or a ﬁnite inﬂection,
determines the distribution of the whole clause. This is as expected if I is the head of
the clause: the properties of the head INFL percolate up to its projection, IP, and are
visible for an outside selector (e.g., a verb). A ﬁnite clause is a clause whose head is
a ﬁnite inﬂection; a non-ﬁnite clause is a clause whose head is non-ﬁnite.
1.3.3 SOME PREDICTIONS
We have examined an embedded non-ﬁnite clause containing to and we have proposed
that to is inserted in I. Let us now examine the empirical coverage of our
hypothesis. The hypothesis leads to a number of predictions. Below we examine
two of them.
1.3.3.1 Co-ordination with to
We have already examined VP co-ordination and its relation to the morpheme
to. We have noted that VP co-ordination need not affect to. Which types of constituents
are co-ordinated in the ﬁnite clauses in (37)?
(37) a The customers should order their drinks and they can proceed to their tables.
b The customers should order their drinks and can proceed to their tables.
(37a) co-ordinates two constituents of the type IP; (37b) co-ordinates two constituents
of the type I′, i.e. consisting of strings modal auxiliary + VP.
29
See the discussion in Chapter 2, section 3.
TSC03 07/18/2005, 03:30PM184
Lexical Projections and Functional Projections 185
(38) a [IP The customers should order their drinks]
+ [IP they can proceed to their tables]
b [IP The customers [I′ should order their drinks]
+ [I′ will proceed to their tables]]
Let us try to apply I′ co-ordination to a non-ﬁnite clause with to as the ﬁller of I.
Reduce the following co-ordinated clauses by co-ordinating (non-ﬁnite) I′:
(39) a We want the customers to order their drinks and we want them to proceed
to their tables.
If we co-ordinate I′ units this means that we co-ordinate two sequences to + VP.
This is what we see in (39b, c):
(39) b We want
[IP the customers [I′ to order their drinks]
and [I′ to proceed to their tables]]
c = We want the customers to order their drinks and to proceed to their tables.
Our proposal that to occupies I allows us to co-ordinate two inﬁnitival VPs introduced
by to.
1.3.3.2 The split inﬁnitive
In the following examples we illustrate ﬁnite clauses with a modal auxiliary. Draw
a tree diagram for the examples:
(40) a The jury will unanimously condemn the doctor.
b The sisters should only play those tournaments.
We have proposed that in non-ﬁnite clauses the inﬁnitive marker to occupies
the inﬂectional head, I, and that, similarly to ﬁnite auxiliaries, to is not a bound
morpheme. Hence, to will not need to lower onto V. Like the ﬁnite auxiliaries, to
can remain under I. Following this reasoning, we would expect that to can be
separated from V by material that is left-adjoined to VP:
(40) c IP
I′NP
VPI
VPAdv
will
to
The jury unanimously condemn the doctor
TSC03 07/18/2005, 03:30PM185
186 Chapter 3
Is this prediction correct? Do we ﬁnd sentences that display the sequence to +
adjunct + V? We do indeed: the relevant sentences will exemplify the so-called split
inﬁnitive pattern. Though sometimes frowned upon by prescriptive authors and
grammarians, the split inﬁnitive is a well-attested phenomenon in English. Identify
the relevant pattern in the following examples:
(41) a The jury of seven men and ﬁve women took 33 hours and 55 minutes to
unanimously ﬁnd the doctor guilty of all the murder charges and forging
the £386.000 will of his ﬁnal victim. (Guardian, 1.2.2000, p. 1, col. 1)
b No one here is in a position to accurately assess the prior investigation.
(Chicago Tribune, 30.11.2002, section 1, p. 20, col. 5)
c In the past few years, Web-sites allowing students to anonymously review
their professors on their teaching have also been popular. (San Francisco
Chronicle, 28.11.2002, p. A28, col. 2)
2 Building Structure by Merge and Move
2.1 Summary so far
In the preceding discussion we have described sentence structure: we have identiﬁed
the constituents of the sentence and we have described how they are related to each
other. So far we have mainly worked on completed sentences, whose components
we have identiﬁed. In this section we take a sentence apart and we examine how we
can put it back together. (42) is a schematic presentation of our general hypothesis
about linguistic structure:
(42) a XP
X′speciﬁer
complementX = head
Let us break up this structure and examine how it is built up from its constituents.
Syntactic units are formed by putting together or merging two constituents.30
The
construction of (42a) can be analyzed as a two-step procedure. We ﬁrst merge
the head X and its complement, to form a constituent that we label X′ (42b). The
complement itself is a complete constituent.
30
The proposal that structures are built by the combination of the operations Merge and Move
is one of the basic tenets of the Minimalist Program (Chomsky 1995), the current version
of generative syntax. For introductions to the Minimalist Program see Adger (2003), and
Radford (2004).
TSC03 07/18/2005, 03:30PM186
Lexical Projections and Functional Projections 187
(42) b X′
complementX = head
Then we merge the resulting constituent X′ with a speciﬁer to form XP.
(42) c XP
X′speciﬁer
The two applications of Merge lead to the formation of a complete constituent, XP.
Both speciﬁer and complement are core ingredients of a structure.
Once we have completed a constituent such as that labeled XP in (42a) this
ﬁnished product can be selected by another head, say Y. The selecting head Y
will merge with the constituent and XP becomes the complement of the head it
combines with:
(43) a Y′
XPY
When a head merges with a completed constituent, the latter serves to “complete”
the structure of the head, it satisﬁes some selectional requirement of the head. We
call it the complement.
Alternatively, an already completed constituent may serve as the speciﬁer of a
structure, in which case it serves to delimit the application of that constituent:
in (43b) XP is the speciﬁer of a constituent headed by Z.
(43) b ZP
Z′XP
complementZ
Another way of extending a structure is by adjunction. A completed constituent YP
may be adjoined to another completed constituent, say XP. Adjunction is the merger
of one fully formed constituent with another fully formed constituent. In the case of
adjunction, there is an asymmetric relation between the adjunct and the constituent
to which it adjoins. When an adjunct YP adjoins to a constituent of type XP, the
resulting structure is an augmented XP. (44a) illustrates left-adjunction of YP to
XP; (44b) illustrates right-adjunction of YP to XP. Observe that in both cases the
TSC03 07/18/2005, 03:30PM187
188 Chapter 3
output of the adjunction is a constituent of the type XP. The category of YP is no
longer represented at the higher level. (44c), (44d), and (44e) illustrate some possible
cases of multiple adjunction.31
(44) a bXP
XPYP
XP
YPXP
c dXP
XPZP
XPYP
XP
ZPXP
YPXP
e XP
XPZP
YPXP
In (44a) and in (44b) XP merges with the adjunct YP to form an augmented XP.
In (44c, d, e) there are two adjoined constituents, YP and ZP. The adjoined constituents
are like peripheral satellites of the constituent XP.
In (45) it is the constituent XP that is the satellite to the constituent YP.
(45) a bYP
YPXP
YP
XPYP
Recall that adjoined constituents do not complete the structure they associate with:
a constituent is structurally complete even without any adjoined constituents.
So far we have come across the speciﬁer position in NPs32
and in clauses.33
The
question has already been raised34
whether VP also has a speciﬁer and if it does,
how its speciﬁer is realized. This issue will be tackled in Chapter 4.
31
For example of multiple adjunction see Exercise 14 in Chapter 2 and Exercise 3 in this chapter.
32
See Chapter 2, section 3.1.2.
33
See section 1 of this chapter.
34
Cf. Chapter 2, section 3.3.
TSC03 07/18/2005, 03:30PM188
Lexical Projections and Functional Projections 189
2.2 Merge
Let us look at an invented sentence and think about how it is put together:
(46) a The student will probably very carefully examine the data.
The semantic core of the sentence is the verb examine: the verb denotes the kind of
situation that we are dealing with, and it determines which components will be
obligatorily present. If we want to build a sentence containing the verb examine, we
start off with the verb ((46b), step (i)). We add its complement, the direct object NP
((46b, step (ii)). The object is an essential ingredient, it serves to narrow down the
action, telling us what is the entity that is the target of the examining. The verb
selects the category of its complement; some verbs select NPs, others select PPs, etc.
At the next stage ((46b), step (iii)), we add the adjunct of manner very carefully.
This adverbial phrase “ﬁne-tunes” the action expressed by VP: we specify how the
action is carried out. Observe that the resulting constituent is labeled VP. We add
another adjunct probably ((46b), step (iv)). The resulting VP is then selected by the
auxiliary will: will sets the event in the future ((46b), step (v)) and it serves to link
the VP to the subject: the student ((46b), step (vi)). One additional remark is in
order here: when we add constituents such as NP, AdvP, etc. to the structure,
these constituents have themselves been formed by the same Merge operation. We
ignore these operations here, since they use the same mechanisms as those shown
in (46b).
(46) b (i) V = examine
(ii) VP
NP
the data
V
examine
(iii) VP
VP
examine the data
AdvP
very carefully
(iv) VP
VP
very carefully examine the data
AdvP
probably
TSC03 07/18/2005, 03:31PM189
190 Chapter 3
(v) I′
VP
probably very carefully examine the data
I
will
(vi) IP
I′
will probably very carefully examine the data
NP
the student
We build up the structure starting from V and by using the procedure Merge by
which we combine two constituents to form a new constituent.
In section 2.3 below we elaborate in some more detail the idea that the verb is the
semantic core of the sentence.
2.3 The operation “Move”
Once we have built up a certain amount of structure, we can also move around
constituents from within the already existing structure. We call the position in
which a constituent is merged in the structure its base position. The base position of
the object, for instance, is that shown in (46b(ii)). It is right-adjacent to the lexical
verb. The underlined segments of (47) are examples in which a constituent has been
moved away from its base position. Identify the displaced constituent in the underlined
sentences.
(47) a Baxter said that he had been using a Sinex liquid decongestant . . . but then
spotted the Vicks inhaler when shopping in Park City, and bought it since
he preferred to use it. “The British one I have been using since I was about
nine.” (Guardian, 22.3.2002, p. 3, col. 1)
b Everything I did right for 20 years, he burned up in two or three. (Washington
Post, 29.4.2003, p. C1, col. 1)
c That night, I came home from the movies to ﬁnd trails of red candle
wax all over the ﬂoors. “We lost the electricity for a while,” my husband
explained. “The only light I could ﬁnd was a votive candle.” The red
votive candle I haven’t replaced yet. (New York Times, 2.1.2003, p. G4,
cols 1–4)
d Letter writer:
[The cat] will rummage through our closets looking for socks, drag them
to the hallway and then make a strange sound as though she is calling her
imaginary kittens to eat.
TSC03 07/18/2005, 03:31PM190
Lexical Projections and Functional Projections 191
Animal Doctor’s Reply:
Bringing such portable items home, or going through the house to ﬁnd
things to “gift” to you, I interpret as a cat’s way of showing affection.
(Animal Doctor, Washington Post, 29.4.2003, p. C9, cols 1–2)
e Our dustmen arrive too early for me to check, but our ﬁshmonger and his
staff in Petersﬁeld all wear ties (Letters, October 22) and very smart they
look too. (Guardian, 23.10.2002, p. 9, col. 5, letters to the editor from
David Dew, Horndean, Hants)
f “They must talk about it, and talk about it they must,” he said. Food for
thought, there! It’s a phrase that could add a measure of gravity to any
press conference. “We must do this, and do this we must.” (Guardian,
29.1.2003, p. 2, col. 5)
The role of movement to create sentence structure will be the central theme of
Chapter 5.35
3 Meaning Relations and Structure
3.1 Thematic roles
We have proposed that constituents of a sentence are to some extent determined
by the choice of the lexical verb. For instance, if you are going to build a sentence
around the transitive verb examine, you expect to merge the verb with an object.
On the other hand, when using the intransitive verb yawn, you will not need such a
complement. Some verbs select a complement of a certain type; other verbs don’t
require a complement. The selectional requirements of verbs – whether they require
a complement and if so, what kind – are related to their meanings. The verb examine,
for instance, refers to a situation involving two participants: the one who examines
something or someone and the entity that is being examined. It is hard to think of a
situation of ‘examining’ without also thinking of these two entities. For instance, if
you were asked to draw a picture of the activity of ‘examining’, it is likely that your
drawing would represent an examiner and the element that is being examined. On
the other hand, the verb yawn does not express a situation involving two entities. A
picture of the activity ‘yawn’ might well be restricted to a drawing of a yawning
individual.
Some constituents of the sentence are inherently required by the meaning of the
verb. It is as if each verb sets the scene for some kind of situation: the verb requires
35
For some suggestions as to what position the sentence initial constituents in the underlined
sentences occupy we refer to Chapter 2, Exercise 15 and to Chapter 5. Exercise 8 of this
chapter leads on to the discussion in Chapter 5.
TSC03 07/18/2005, 03:31PM191
192 Chapter 3
a number of entities that will be involved in the situation. The participants in the
situation are called the arguments of the verb. We say that a verb has an argument
structure. When a verb is introduced as the nucleus of a VP, its argument structure
is activated. Examine, for instance, is a verb with two arguments. The ﬁrst argument
is the AGENT of the activity. It is realized by the subject in (48a). The second
argument refers to the entity that undergoes the activity, what we will call the
THEME. In (48a) the second argument of examine is realized by the direct object
NP. The distinct participant roles attributed to the arguments of the clause are
also often referred to as thematic roles. A verb such as give in (48c) has three
arguments: the agent, the beneﬁciary, and the entity that is being transferred, the
theme.
(48) a The student examined the project.
b The student yawned.
c The student gave the book to her friend.
(49) is a list of the labels that have been used in the literature to refer to the thematic
roles that are associated with the arguments of a verb.36
(49) a agent/actor: the one who intentionally initiates the action.37
b theme (1): the entity undergoing a change of state (location, possession).
c patient: the entity undergoing the action. The patient undergoes a change
in its internal state (rather than terms of location or possession, both of
which would be external states).
d experiencer: the entity experiencing some (psychological) state.
e benefactive/beneﬁciary: the entity beneﬁting from the action.
f goal: the entity toward which the theme is displaced.
g source: the place/entity from which the theme is moved as a result of the
activity, the place/entity from which the action originates.
h location: the place in which the action or state is situated.
i cause: the entity that unintentionally initiates the event.
j instrument: the means by which an action is performed.
k possessor: the entity that owns something.
Sometimes the roles of patient and theme (1) are grouped into a single role
theme:
(49) l theme (2): the entity affected by the action.
36
For discussion of the relation between thematic roles and syntactic structure see, among others,
Jackendoff (1987), Wilkins (1988), Grimshaw (1990), Levin and Rappaport-Hovav (1995),
Reinhart (2000).
37
Exercises 9, 10, 11.
TSC03 07/18/2005, 03:31PM192
Lexical Projections and Functional Projections 193
The following sentences illustrate the realization of thematic roles:
(50) a Thelma ate the apple.
agent patient/theme (2)
b Thelma handed Louise the text.
agent beneﬁciary theme
c Thelma liked the text.
experiencer cause38
d The wind broke the fence.
cause theme
3.2 Linking thematic roles and syntactic positions
The argument structure of the verb determines to some extent the composition of a
sentence. All the thematic roles associated with an active verb must be realized in
the sentence; i.e. the verb must have a sufﬁcient number of arguments to be able to
assign all its thematic roles. It is often assumed that there is a one-to-one relation
between thematic roles and arguments.39
For instance, a verb cannot assign one
thematic role to two different arguments. In (51a) the verb examine will either
assign a thematic role to the NP the project or to the NP the book. But it cannot
assign the same role twice. This means that one of the two NPs remains without a
thematic role and hence cannot be related to the verb.
(51) a *[NP Thelma] examined [NP the project] [NP the book].
We can make (51a) grammatical by creating a single argument out of the two noun
phrases. In (51b) the constituent the project and the book co-ordinates the noun
phrases the project and the book. The resulting constituent is itself also a noun
phrase; it functions as one argument which realizes the second thematic role of the
verb examine:40
(51) b [NP Thelma] examined [NP [NP the project] and [NP the book]].
One single argument cannot normally be assigned two thematic roles. The verb
invite takes two arguments in (52a). We infer that it assigns two thematic roles
(52b).
38
For a discussion of the argument structure of psychological verbs such as like, fear, frighten,
etc., see Belletti and Rizzi (1988). Though this paper is essentially on Italian, its ﬁndings extend
to English. The paper is advanced and should be tackled only after Chapter 5.
39
See also Baker (1997), Newmeyer (2001).
40
For the structure of co-ordinated constituents see Chapter 2, Exercise 7. For problems with the
label of co-ordinated constituents see Chapter 2, Exercise 9.
TSC03 07/18/2005, 03:32PM193
194 Chapter 3
(52) a Thelma invites Louise.
b invite thematic role 1: Thelma
thematic role 2: Louise
(52c) is ungrammatical. In this example, both thematic roles associated with invite
are meant to be assigned to the same argument, Thelma:
(52) c *Thelma invites.
d *invite thematic role 1: Thelma
thematic role 2: Thelma
If the agent of the invitation and the theme both happen to be associated with the
person referred to as Thelma, the object (though not the subject!) can be expressed
by means of a reﬂexive (herself in (52e)).
(52) e Thelma invites herself.
f invite thematic role 1: Thelma
thematic role 2: herself
We have proposed that I is an inherent linker (see section 1.1): it links a VP to a
speciﬁer. Given the two-faced nature of I, a sentence always has a subject position
available, the speciﬁer of IP. This means that one argument of a verb can always be
realized as the subject of the sentence. In the case of examine, the agent of examine
is realized as the subject and the patient is realized as the direct object, the NP that
is ﬁrst merged with the verb; examine selects one NP complement.
3.3 Relating the subject to the verb
When discussing the internal structure of the VP we proposed that adjuncts were
more peripheral to the VP than complements (such as direct objects). Our motivation
for this proposal was the observation that adjuncts are less central to the meaning
of the VP than complements. Complements are selected by V. We can relate this
intuition in terms of the concepts of argument and thematic role by saying that
because complements are arguments of the verb, they receive a thematic role from
the verb. Hence complements must be close to V and they are part of the core VP.
VP-adjoined elements are outside the core VP: they do not receive a thematic role
from the verb. Adjuncts need not be as close to the verb precisely because they do
not receive a thematic relation from the V.
However, now an internal contradiction arises in the way we have set up the structure
of the sentence. The problem is that the subject NP will usually also receive a
thematic role from the verb. Now in the tree diagrams elaborated so far, the subject
NP occupies the speciﬁer of IP.41
Hence, in the structure we have elaborated, the
41
See Exercise 17 of this chapter, though, for complications.
TSC03 07/18/2005, 03:32PM194
Lexical Projections and Functional Projections 195
subject is radically outside the VP and it is deﬁnitely not as close to V as any of the
VP-adjoined constituents. How is it possible for the subject to receive a thematic role
from V while it is located outside the VP and while it is further away from V than
adjuncts? This state of affairs makes it hard to motivate the distinction between
complements and adjuncts on semantic grounds. Our present representation makes
it appear as if the subject had a weaker semantic relationship to V than VP-adjoined
constituents, even though the subject is thematically related to V and adjuncts are
not. If the tree structure is intended to reﬂect closeness of semantic relations, then we
would expect all constituents that are thematically related to V to occupy a position
inside the core VP. This would imply that the subject must be part of the core VP.
In order to relate the subject to the verb and to the VP we would need to create a
position for the subject inside the VP. Recall that in our representation of the
structure of the VP we have not yet made use of the speciﬁer position. If we were to
introduce a speciﬁer to the VP that position could become the VP-internal position
for the subject. Note that if we do integrate a subject position in the VP we end up
with two subject positions.
In preceding discussions we have seen that to motivate a certain structure we
might use either empirical arguments or theoretical arguments. If you go over the
discussion above, do you think that our motivation for a VP-internal subject position
at this point is mainly empirical or theoretical? In order to answer this question
you should examine the reasoning that we adopt. Is it based on empirical data or is
it based on the way our theory is conceived? Probably, your conclusion will be that
our motivation has been essentially theoretical. Postulating a VP-internal subject
position is based on two strands of reasoning. On the one hand, we want to create
a closer structural relation between the subject and the verb because we want to
have a ﬁt between syntactic structure and interpretation. But this ﬁt is one that we
have set up as a goal of our theory, and which we use as a guideline for postulating
structure. In addition, when elaborating our structures we had come up with the
hypothesis that a projection could have a speciﬁer. We have identiﬁed the ﬁller for
the speciﬁer of NP and for the speciﬁer of the clause. We have not yet identiﬁed any
element that could qualify as the speciﬁer of VP. The absence of a VP speciﬁer
would be a problem for our approach. We can generalize the speciﬁer position to
VP if we relate the subject NP to that position: the resulting VP will then also have
a speciﬁer and will therefore be fully in line with the structures proposed for the NP
and for the clause. We will return to this point in detail in Chapter 4, where we will
see that there also exist a number of empirical motivations for postulating that there
is a position for the subject in the VP.42
3.4 Auxiliaries vs. lexical verbs
We have already come across a number of patterns in which auxiliaries and lexical
verbs pattern differently. Before reading on, try to make a list of the differences
42
See Exercises 17 and 18 for some empirical evidence supporting this idea.
TSC03 07/18/2005, 03:32PM195
196 Chapter 3
between auxiliaries and lexical verbs. You will probably recall that sentences with
auxiliaries do not require do-insertion in questions or in negative sentences with
not. We can explain this if we assume that English auxiliaries can be inserted in
I, while English lexical verbs do not move to I. Consider the sentences in (53).
How many thematic roles would we associate with the verb neglect? How are
these thematic roles realized in (53)? Assuming that each constituent can only carry
one thematic role, do you think it is plausible that auxiliaries can assign thematic
roles?
(53) a The student neglected his studies.
b The student had neglected his studies.
c The student will neglect his studies.
d The student is neglecting his studies.
We deduce from the examples in (53) that auxiliaries cannot assign thematic roles.
If they did, then we would expect that, compared to (53a), (53b–d) should contain
at least one additional argument to realize the thematic role associated with the
auxiliary, contrary to fact. So we conclude that auxiliaries, though by their nature
verbs, do not assign any thematic roles. Lexical verbs are potential thematic role
assigners.43
3.5 Lexical head vs. functional head
If we compare (54a) in the present tense and (54b) in the past tense, there is no
difference in their thematic structure. Inﬂectional morphemes are not thematic role
assigners. Their function is to place the event expressed by the verb with respect to
time. This means then that the prototypical ﬁller of I, inﬂection, is not a thematic
role assigner. While V is a lexical head, a head such as I, which itself does not assign
thematic roles, is said to be a functional head.44
The lexical verb heads a lexical
projection; I heads a functional projection.45
(54) a John always eats chocolate.
b John always ate chocolate.
Note that this does not mean that I can never contain a lexical head that assigns
thematic roles. We have seen that in French (54c), the lexical verb raises to I. As a
43
The fact that lexical verbs do and auxiliaries do not assign thematic roles could explain why
the former are sometimes called “full verbs.” (See Chapter 1, section 2.4.)
44
See also Chapter 1, section 2.4 for the concept “lexical.”
45
Exercise 13 raises the question if projections of the N also are integrated into a functional
projection.
TSC03 07/18/2005, 03:32PM196
Lexical Projections and Functional Projections 197
result I will contain V, a thematic role assigner. But crucially, the verb is not directly
merged in I. The verb is merged as the head of VP, and it has only ended up in I
because it has been attracted there by the content of I, inﬂection (here the ending
-ait). And once again the inﬂection itself is not a thematic role assigner.
(54) c [IP Jean [I′ [I aim-ait] [VP —— [NP du chocolat]]]].
4 Multiple Auxiliaries
4.1 A ﬁrst hypothesis: Clustering auxiliaries
Identify the auxiliaries in the sentences in (55). Which is the ﬁnite auxiliary? How
do you know? Form a direct question on the basis of these examples. Using the
concepts developed so far, describe the processes used to form a question.
(55) a The inspector will be staying at the pub.
b The press have been writing terrible things about him.
c Without this concession, he would have pulled the trigger.
d He could have been staying at the pub.
Each of the examples in (55) contains more than one auxiliary. In each of them, the
ﬁrst auxiliary is ﬁnite, witness the fact that it is tensed (present vs. past tense), and
the ﬁnite auxiliary inverts with the subject to form a direct question. Below we have
underlined the auxiliaries, we have indicated the tense form of the ﬁnite auxiliary,
and we also illustrate SAI in direct questions:
(56) a The inspector will (present, vs. would) be staying at the pub.
Will the inspector be staying at the pub?
b The press have (present, vs. had) been writing terrible things about him.
Have the press been writing terrible things about him?
c Without this concession, he would (past, vs. will) have pulled the trigger.
Would he have pulled the trigger?
d He could (past, vs. can) have been staying at the pub.
Could he have been staying at the pub?
The question arises how multiple auxiliaries get inserted into the structure of a
sentence. In our earlier discussions, we inserted the inﬂected auxiliary under I.
When we have more than one auxiliary we might think that we should insert all the
auxiliaries as a kind of auxiliary cluster under I. This is illustrated in (57a), for
example (56d). Let us evaluate this proposal.
TSC03 07/18/2005, 03:32PM197
198 Chapter 3
(57) a IP
I′NP
I VP
have. . . . . .beencould
The ﬁrst question that arises is whether (57a) is compatible with the theory we have
been elaborating. If you look at the tree diagrams with which we have been operating
so far, and if you recall our proposals for structure,46
you will decide that (57a)
is not compatible with our theory of structure. (57a) violates the binary branching
hypothesis. In (57a) I is ternary branching: there are three auxiliaries hanging directly
from I.
Needless to say, this particular problem can only arise with sentences with more
than two auxiliaries, such as (56d). When there are only two auxiliaries, the head I
would obviously be binary branching even if the auxiliaries formed a cluster.
We might decide that the binary branching hypothesis is wrong and that it should
be discarded, and we might re-introduce multiple branching into our theory. However,
this rather rash step would increase the tools of our theory dramatically.47
To
maintain that auxiliaries form a cluster, while at the same time preserving the binary
branching hypothesis, we could introduce binary branching inside I. (57b) and
(57c) are two ways of achieving this.48
The labels Perfect and Prog(ressive) identify
which auxiliary is going to be inserted in the relevant position.
(57) b cIP
I′NP
I VP
Aux
Prog
M
Perfect
have beencould. . . . . .
IP
I′NP
I VP
ProgAux
Perfect
havecould
M
been
46
From Chapter 2, section 2.4.2.
47
Recall the discussion of the concept of economy in Chapter 1, section 1.2.3.
48
A version of (57b) was proposed at one time (Chomsky (1975, 1972)).
TSC03 07/18/2005, 03:32PM198
Lexical Projections and Functional Projections 199
Having solved a theoretical objection to (57a), let us now try to think of empirical
evidence that could be invoked to support one of the structures in (57b, c). Alternatively,
what kind of empirical evidence could be invoked to challenge one or the
other?
Do the sentences with SAI in (56) bear on (that is conﬁrm or challenge) the structures
in (57b, c)? In fact, these examples are neutral with respect to the structures in
(57b, c). SAI moves the ﬁnite auxiliary to the initial position, which means that the
ﬁnite auxiliary is a constituent. This is the case in each of the representations in (57),
in which the ﬁnite auxiliary forms the node I in combination with other auxiliaries.
It is in principle possible to move a constituent from a larger constituent: for instance
we can move an object NP out of a VP. In French, we have seen that V can leave VP
and move to I.
To show that either (57b) or (57c) is appropriate we would have to ﬁnd a construction
pattern in which we move not just the ﬁnite auxiliary but, for instance, a
cluster of two or three auxiliaries. This would provide empirical evidence for the
clustering of the auxiliaries in I. Try moving a cluster of auxiliaries in the examples
above. Are the resulting sentences grammatical? Clearly not, as shown in (58):
(58) a *Would be the inspector staying at the pub?
b *Had been the press saying that things were wrong?
c *Will have he pulled the trigger?
d *Could have been he staying at the pub?
*Could have he been staying at the pub?
That the examples in (58) are unacceptable could be due to the fact that the auxiliaries
do not in fact form a cluster. But this explanation is only one possibility. It is
conceivable that the unacceptability of (58) might also be due to some other factor(s).
Perhaps SAI can only move the ﬁnite auxiliary. We do conclude, though, that SAI
does not at the moment provide any direct support in favor of either (57b) or (57c).
Let us examine another prediction of the structures in (57b, c). Assuming (57b,
c), any syntactic operation affecting auxiliaries should always target one individual
auxiliary, or it should target a cluster of auxiliaries. Any operation should affect all
the auxiliaries in I, or it should affect the two auxiliaries under Aux in (57b) or
(57c). A syntactic operation could also target the VP. According to these representations,
no operation should be able to affect, for instance, the rightmost auxiliary
in the cluster under I as well as the VP. If we did ﬁnd evidence that there is such an
operation then that would mean that the auxiliaries do not form a cluster and we
would have to devise an alternative structure. If possible, we would also try to offer
empirical support for the new structure. This is what we will do in the next sections.
4.2 A second hypothesis: Auxiliaries as heads of VP
The attested sentences in (59) contain multiple auxiliaries. Read the sentences carefully,
identify ﬁnite and non-ﬁnite auxiliaries, and decide to what extent, if at all,
TSC03 07/18/2005, 03:32PM199
200 Chapter 3
these data can be made to bear on the question raised in the preceding section concerning
the structure of sentences containing multiple auxiliaries.
(59) a He drove her hard, he stole her fame or would have if he could have.
(Guardian, Review, 24.5.2003, p. 5, col. 3)
b You should have been an accountant and made some money so you could
take care of your parents. (Washington Post, 29.4.2003, p. A23, col. 4)
c I should have gone on the stage or been a diplomat’s wife or an international
spy. (Mary McCarthy, The Company She Keeps, 1989: 3)
d I asked him where this Detective Inspector Thomas Lynley was from . . . He’s
from New Scotland Yard! Staying right here at the inn, he’ll be. He booked
a room hisself not three hours past. (Elizabeth George, Missing Joseph,
1993/1996: 144)
(59a) illustrates ellipsis: the VP associated with have has been omitted.
(59) a′ (he) would have stolen her fame if he could have stolen her fame.
The application of ellipsis in (59a) is compatible with the representations in (57b, c)
in that it shows that the VP must be able to be treated as a constituent. Once again,
this example does not provide any support for the clustering of the auxiliaries
(would have). Similarly, in (59b) and in (59c), VPs are co-ordinated:
(59) b′ You should have [VP been an accountant] and [VP made some money].
c′ I should have [VP gone on the stage] or [VP been a diplomat’s wife or an
international spy].
Again, these examples simply conﬁrm the constituent status attributed to the verb
and its complements, the VP, but they do not bear on the grouping of the auxiliaries
(should have). In (59d) the VP is fronted. Again this example conﬁrms that VP is a
constituent but does not bear on the clustering of auxiliaries (’ll be).
(59) d′ Staying right here at the inn, he’ll be.
Now examine the examples in (60), which also contain multiple auxiliaries.
Do these provide any arguments for determining the structural relations between
auxiliaries? Where possible, try to group related examples.
(60) a “The Hershey chocolate company is about to be sold!” he says, eyes widening.
“Who could have imagined it?” Very few could. (Based on Guardian,
G2, 26.8.2002, p. 2, col. 1)
b Some 24% agreed top-up fees would not have mattered, while 35% would
have considered other universities but probably still have chosen the same
university. (Based on Guardian, 20.1.2003, p. 5, col. 2)
TSC03 07/18/2005, 03:32PM200
Lexical Projections and Functional Projections 201
c Michael Jackson has, on some occasions in the past, not eaten when he
should. (Guardian, Review, 28.5.2003, p. 3, col. 1)
d But we have been saying for 15 to 20 years that there are too many games
in the top division and done nothing about it. (Guardian, 15.2.2003, Sport,
p. 11, col. 5)
The examples above do bear on the structures in (57b, c). Indeed, they illustrate the
patterns that the structures in (57b, c) are predicted to exclude. In (60) an operation
(movement, ellipsis, co-ordination) applies to a combination of one auxiliary of the
cluster and the VP, excluding the other auxiliaries in the cluster from the operation.
That such an operation should be possible is not predicted by either variant (57b) or
(57c). Let us examine some of the examples in (60) in detail.
(60b) and (60d) illustrate co-ordination. In (60b) co-ordination affects two strings
consisting of a non-ﬁnite auxiliary have and a VP (61a). In (60d) co-ordination
affects a constituent containing the auxiliary been and the VP on the one hand, and
a VP on the other hand, (61b).
(61) a 35% would [have considered other universities]
but [probably still have chosen the same university].
b we have [been saying for 15 to 20 years that . . . in the top division]
and [done nothing about it].
If co-ordination affects constituents (as we have been assuming so far),49
it is hard
to see how the auxiliaries could form a cluster of the type proposed in (57b, c). If
auxiliaries did indeed form a cluster, one would not expect that one particular
auxiliary could be affected by a syntactic operation which also affects the VP, while
the other auxiliaries of the cluster remain unaffected. The data suggest that we need
a structure in which we can operate on a (non-ﬁnite) auxiliary and the VP to its
right without at the same time affecting the ﬁnite auxiliary. In schematic terms we
need to be able to isolate AUX2 and the VP without affecting AUX1:
(62) a subject AUX1 AUX2 VP
Put differently, AUX2 and VP should be able to form a constituent. (62b) gives a
partial bracketing representation:
(62) b [IP subject AUX1 [constituent AUX2 [VP VP]]]
If you try to translate this representation into a binary branching tree diagram structure,
you end up with something like (62c) as a ﬁrst approximation. The provisional
label constituent simply means that the auxiliary and the VP are grouped as one
entity.
49
See Chapter 2, section 1.8.
TSC03 07/18/2005, 03:32PM201
202 Chapter 3
(62) c IP
I′subject
I CONSTITUENT
VPAUX2
In (60a) and in (60c) ellipsis affects a combination of one auxiliary (have) and the
VP. These patterns are compatible with the structure proposed in (62c): ellipsis
applies to the newly identiﬁed unit, provisionally labeled constituent.
(61) c Very few could [constituent have imagined it].
d when he should [constituent have eaten].
On the basis of the argumentation above, we need to enrich our structure so that we
can insert all the auxiliaries, both ﬁnite and non-ﬁnite. We have postulated that
there is a unit formed of an auxiliary and the VP. The auxiliary is a head, the VP
can be seen as its complement. Various questions come to mind: What are the heads
in which the auxiliaries are merged? If auxiliaries are heads, they will give rise to a
full projection of their own. What kind of a projection is headed by the auxiliary?
Assuming the projection headed by the auxiliary also has a speciﬁer, the question
arises what, if anything, do we merge in that speciﬁer position? Before answering
this question let us ﬁrst look at another issue concerning ﬁnite auxiliaries.
4.3 Finite auxiliaries
Throughout the discussion, we have inserted the ﬁnite auxiliary directly under I.
Tree diagram (63) is based on the discussion in section 1 of this chapter.
(63) IP
I′NP
I VP
VPAdvP
V NP
modify
modify
modified
modifying
the advice
the advice
the advice
the advice
later
later
later
later
will
-ed
had
were
a The department
b The department
c The department
d The department
TSC03 07/18/2005, 03:33PM202
Lexical Projections and Functional Projections 203
Under I we ﬁnd the present tense form of the modal auxiliary will (a), the past tense
ending of the verb -ed (b), the past tense form of the auxiliary have (c), and the past
tense form of the auxiliary be (d).
There is an inconsistency in our treatment of the tense inﬂection. In the case of an
inﬂected lexical verb like modify we split up the verb and its tense or agreement
inﬂection, inserting the latter in I. In the case of the modal auxiliary will and the
auxiliaries have or be, we don’t apply a split. In other words, we treat inﬂectional
endings differently depending on the type of verb they associate with. We have not
motivated this differential treatment. It would be simpler (and hence preferable) if
all inﬂectional endings of the verb could be treated in the same way.50
Would this be
possible?
Let us ﬁrst look at the auxiliaries have and be. First of all, we note that though
these elements are often followed by a VP, they need not be. Have and be can be the
only verb element in the sentence.
(64) a John had lunch at the club.
b John is the captain.
So, without going into the details of the structures in (64), it seems reasonable to
allow ourselves to treat be and have as verbs in their own right. In their independent
uses, illustrated in (64), have and be are not necessarily ﬁnite: they need not be
associated with a tense form; they may also be non-ﬁnite:
(64) c John has had lunch at the club.
John is having lunch at the club.
John wants to have lunch at the club.
d John has been the captain.
John is being difﬁcult.
John seems to be hungry.
We conclude that have and be are verbs with ﬁnite and non-ﬁnite forms.
In examples with multiple auxiliaries, we also ﬁnd the auxiliaries have and be in
their non-ﬁnite forms. This means rather uncontroversially that in their auxiliary
use too, these verbs are not inherently ﬁnite. In the following examples we pair nonﬁnite
be and the non-ﬁnite lexical verb keep.
(65) a The committee will probably be modifying the advice.
The committee will probably keep modifying the advice.
b The committee have probably been modifying the advice.
The committee have probably kept modifying the advice.
50
A further inconsistency is that we insert participles as one element, disregarding that they too
contain an inﬂectional morpheme. We do not dwell on participial morphology here, as this
would lead us too far. See for discussion Belletti (1990) and Friedemann and Siloni (1993). See
also Exercise 14 of this chapter.
TSC03 07/18/2005, 03:33PM203
204 Chapter 3
Recall that we assume that a lexical verb is merged as the head of the VP and that
it subsequently combines with the tense inﬂection. Concretely, we propose that in
English the inﬂection lowers onto the verb.
(66) a IP
I′NP
I VP
VPAdvP
V NP
modify the advicelater-edThe committee
For the sake of internal consistency, ought we not to try to propose the same analysis
for the auxiliaries have and be? We might propose that have/be are merged in the
structure as verbs and that they subsequently combine with their inﬂection. In fact,
if we did decide against this move, we would actually have to explain why we do
not treat have/be as verbs here, since they seem to be able to act as a verb elsewhere.
Let us treat auxiliaries be and have as verbs. In their use as auxiliaries, be and
have have the speciﬁc property of selecting a VP as their complement. The auxiliary
be selects a VP complement with a present participle as its head; have selects a
VP complement with a past participle as its head. The inﬂection of the auxiliary is
inserted under I, just like any other verb inﬂection. Let us translate this in terms of
the structure.
(66) b IP
I′NP
I VP
VPV
V NP
modifying
modiﬁed
the advice
the advice
be
have
-ed
-ed
The committee
The committee
TSC03 07/18/2005, 03:33PM204
Lexical Projections and Functional Projections 205
In other words, after having merged the components of VP (modifying the advice,
modiﬁed the advice), we merge VP with the auxiliary (be, have) and we merge the
resulting structure with the inﬂection. Hence, we no longer make any special provisions
for the auxiliaries be and have, we treat them as verbs. Observe that in (66b)
there are two VPs: one VP is headed by a lexical verb (modifying/modiﬁed), and one
VP is headed by be/have.
We also need to make sure that auxiliaries and their inﬂectional endings are united.
How can we do this? Given a structure like (66b), two procedures can be invoked.51
We have discovered that English lexical verbs unite with their inﬂection as a result
of the lowering of the inﬂection. On the other hand, French lexical verbs move up
to the inﬂection. Since we initially provided evidence that the auxiliaries be and
have occupy I, it seems reasonable to assume that they unite with their inﬂection by
moving to I. This means that we preserve the essence of our hypothesis in Chapter
2 according to which a ﬁnite auxiliary occupies I. In addition, we align the auxiliaries
be and have with other verbs by merging them as Vs taking a VP complement. When
be and have take a VP complement we call them auxiliaries. As auxiliaries, have
and be distinguish themselves from English lexical verbs in that they move to I.52
The fact that auxiliaries can move to I in English needs an explanation. Recall that
auxiliaries such as have and be are members of a closed class.53
They serve to signal
temporal and aspectual relations. The perfective meaning signaled by English have
would in fact be encoded by an inﬂection in Latin: ‘I have loved’ translates as amavi,
the perfect form of the verb amare. Amavi consists of the verb ama- root with an
inﬂectional ending -vi. We will not develop this point in detail. A full account for
the movement of have and be to I could explore the semantic similarity to inﬂectional
elements.54
In our previous discussions, modal auxiliaries were taken to be inserted directly
under I. It is not obvious that we need to change that hypothesis. Consider the
auxiliary can. It can be used to denote an ability, a capacity ascribed to the subject.
However, the auxiliary can only be used in the forms can or could. In most varieties
of English, we cannot, for instance, refer to a future capacity of the subject by using
can in combination with the auxiliary of the future, will. Similarly, we cannot form
a perfect tense with can:
(67) a *He will can ﬁnish the book.
b *He has could ﬁnish the book.55
51
See section 1.2.3.3 of this chapter.
52
This also shows that the operation of a verb raising to I is not incompatible with the grammar
of Modern English. What is not possible is for a lexical verb to move to I.
53
See Chapter 1, section 2.4.
54
For discussion see Emonds (1978), Pollock (1989), Chomsky (1995), and Roberts (1985). For
an alternative approach which merges the auxiliaries in I-type heads see Cinque (1999).
55
There are English dialects, both in the United Kingdom and in the United States, which do
allow double modals. For Scots see Brown (1992) and Miller (1993). For American varieties
see for instance Battistella (1995). We leave these aside here. For the diachronic development
of modals see Lightfoot (1979) and Roberts (1993).
TSC03 07/18/2005, 03:34PM205
206 Chapter 3
Observe that it is not the intended meaning of (67) that causes the problem here.
We may well want to refer to a future capacity of a person (67c), or we may want
to refer to a past capacity with present relevance and use a perfect form (67d), but
in order to do that we use a different construction.
(67) c He will be able to ﬁnish the book.
d He has been able to ﬁnish the book.
Our conclusion is that modal auxiliaries never come without an associated tense.
We can say that in standard Modern English, modal auxiliaries are inherently tensed;
it is as if the tense and the base of the modal are welded together and can no longer
be separated. So we will continue to insert them directly under I.56
(67) e IP
I′NP
I VP
NPV
the advice
the book
modify
ﬁnish
will
can
The committee
He
4.4 The structure of auxiliary sequences
Having concluded that the auxiliaries have and be ought to be treated as verbs, with
the particular property that they move to I to unite with their ﬁnite inﬂection, we
can return to sentences with multiple auxiliaries and complete their structures. Consider
for instance, (68a) and the partial structure in (68b), which we had arrived
at in the preceding discussion. This structure has been improved by re-labelling the
position “Aux” as V and by making this V the head of a VP. The revised representation
(66b) is repeated in (68c):
56
Of course in the dialects that allow double modals (note 55) the analysis will be different. In
earlier stages of English modals did have non-ﬁnite forms (Lightfoot, 1979).
TSC03 07/18/2005, 03:34PM206
Lexical Projections and Functional Projections 207
(68) a These students would [have considered other universities].
b IP
I′NP
I XP
VPAUX2
c IP
I′NP
I VP
VPV
V NP
considered other universitieshavewouldThese students
There are two projections of V: one is headed by the lexical verb considered, the
other is headed by the auxiliary have.
What about (69a), in which I is occupied by the past tense auxiliary had?
(69) a We had been talking about this problem.
To represent the structure we combine the proposals elaborated in this section.
(i) On the one hand, have and be are treated as verbs, heading their own VP and
selecting a VP complement. (ii) On the other hand, these auxiliaries are inserted
independently of the ﬁnite inﬂection – had is decomposed into two morphemes: the
verb root, which we represent as have, and the past tense inﬂection. The auxiliary is
inserted in V and its past tense is inserted in I. (iii) Have moves to I to pick up the
tense inﬂection.57
57
Exercise 12 shows that the same structure can be used for passive sentences. Exercise 15
introduces some empirical problems.
TSC03 07/18/2005, 03:34PM207
208 Chapter 3
(69) b IP
I′NP
I VP
VPV
V VP
talking
talking about this problem.
about this problemhave-ed
have + ed = had
had
We
= We
V PP
been
been
If we compare the tree diagram in (69b) with a representation such as that in (7c),
(69b) might seem to be more complicated and therefore perhaps less economical
(and less highly valued). After all there seems to be more structure in the tree
diagram in (69b).
However the impression that (69b) leads to a more complex or less economical
theory is deceptive. Let us examine the relevant components needed to form the tree
in (7c) and that in (69b). We are concerned with sentences containing auxiliaries
have or be. (7c) presents inﬂected auxiliaries like had and was as unanalyzable
elements, they are inﬂected forms of the auxiliaries. In other words, (7c) has to
assume the ﬁnite forms of the auxiliaries have and be are separate entities, and it
also has to allow for the non-ﬁnite forms of have and be, which we have shown to
be verbs and which project a VP. (69b), on the other hand, decomposes the ﬁnite
forms of the auxiliaries have and be into more elementary units. It separates out the
verbal root and the tense inﬂection. So according to (69b) auxiliaries are simply
verbs that project a VP.
(69b) does require us to postulate that certain verbs in English, namely auxiliaries,
raise to I, and that lexical verbs don’t raise to I. However, that is not a complication
to our theory since we need to allow for V-raising to I independently to account
for the word order patterns in French. So once again, we do not require a new
mechanism, we exploit an existing mechanism.58
We conclude that none of the elements required to replace representation (7c) by
representation (69b) is new to the theory. We only make use of devices that are
58
As mentioned in note 50, the assumption that participles are inserted with their inﬂection is not
consistent with the current discussion and would have to be rethought. See also Exercise 14.
TSC03 07/18/2005, 03:34PM208
Lexical Projections and Functional Projections 209
already needed independently. Hence assuming (69b) for sentences with ﬁnite
auxiliaries is not more complex or less economical.59
5 Summary
The ﬁrst part of this chapter completes the representation of the structure of the
sentence. Pursuing the representation elaborated for sentences with auxiliaries in
Chapter 2, we elaborate a representation that covers both sentences with auxiliaries
and those without. The proposal starts from the observation that the inﬂection of
the verb is a pivotal element in the sentence that links VP and subject. The sentence
is seen as a projection of the inﬂection I. I selects a VP as its complement, and takes
the subject as its speciﬁer. I thus serves to relate the situation denoted by the VP to
the entity denoted by the subject. I and V are united by a movement operation.
French lexical verbs raise to the ﬁnite inﬂection in I; English I lowers onto lexical
verbs. The difference between the two patterns is related to the strength of the
inﬂection as reﬂected in the number of distinct forms in the verbal paradigm.
We conclude that our syntactic representations make use of two basic operations:
the operation Merge which combines two units, and the operation Move which
selects a constituent in an existing structure and moves it to another position. When
building the sentence we start with a verb and we progressively add constituents to
build a more elaborate structure.
The verb plays a central role in the semantic make-up of the sentence. Lexical
verbs denote situations involving one or more participants, the arguments of the
verb. These arguments have speciﬁc semantic relations with respect to the situation
denoted by the sentence. The semantic relations between the verb and its arguments
are referred to as thematic roles. Thematic roles are assigned to arguments on a oneto-one
basis. In contrast with lexical verbs, auxiliaries do not assign thematic roles.
Auxiliaries are functional elements.
In the ﬁnal part of the chapter it is proposed that the auxiliaries be and have be
treated as verbs selecting a VP complement. An auxiliary is inserted as a V head that
merges with a VP complement. In ﬁnite clause, the auxiliaries have and be move to
I to combine with the inﬂection. Modal auxiliaries, on the other hand, are inherently
tensed, they lack non-ﬁnite forms and they are inserted directly under I.
Both the NP and the sentence have a speciﬁer position. We still have to address the
question whether VPs also have a speciﬁer position. Because verbs have a thematic
relation with their subjects we might wish to propose that the subject of the sentence
should be seen as the speciﬁer ofthe VP.However,so far wehave been assuming that the
canonical subject position is in the speciﬁer of IP, in other words it is outside the VP.
If there is a subject position inside the VP then we end up with two subject positions:
the canonical subject position, the speciﬁer of IP, and a VP-internal subject position.60
59
For some complications to the analysis, though, see Roberts (1990).
60
Exercise 17 introduces some empirical support for the hypothesis that there is more than one
subject position.
TSC03 07/18/2005, 03:34PM209
210 Chapter 3
Exercises
Contents
Exercise 1 Constituent structure (T) 210
Exercise 2 The structure of the sentence (T) 211
Exercise 3 Adjuncts, interpretation, and word order (T) 212
Exercise 4 Modals and to (T) 215
Exercise 5 Sentential negation (E) 216
Exercise 6 Comparing languages (T, E) 217
Exercise 7 Comparing languages (T, E) 218
Exercise 8 Moved constituents (T, L) 220
Exercise 9 Thematic roles: The expression of AGENT (T) 222
Exercise 10 Thematic roles: The expression of AGENT (T, E) 224
Exercise 11 Thematic roles: The implied AGENT (T, E, presupposes
Exercise 10) 226
Exercise 12 Be + passive VP (T) 227
Exercise 13 Lexical heads and functional structure (E) 229
Exercise 14 Non-ﬁnite inﬂections (E) 230
Exercise 15 The structure of sentences: Extending the data (E) 231
Exercise 16 Object shift and verb movement in Icelandic (E) 232
Exercise 17 Variations in subject positions (L) 234
Exercise 18 Modern Greek word order (T, L) 235
Exercise 1 Constituent structure (T)
Discuss how each of the examples below can be argued to contain some support
for the hypothesis that the underlined strings of words are constituents:
Example
The news, when it comes, he seems to take —— well enough. (Guardian, G2,
26.7.2002, p. 2, col. 1)
• In this example, movement has affected the string the news. The news is
the direct object of take. Its canonical position is to the immediate right of
the verb. In the example it has been moved to the beginning of the sentence.
The fact that it has been moved suggests the string of words is a constituent.
Moreover, the string the news is substituted for by the pronoun it.
• The news is an NP.
TSC03 07/18/2005, 03:34PM210
Lexical Projections and Functional Projections 211
(1) Two decades of ﬁnancial squeeze has eroded academic standards and seriously
damaged common-room morale. (Guardian, 26.10.2002, p. 13, col. 1)
(2) Lawyers who’ve handled arbitration claims for years . . . are getting very busy;
lawyers who never have are joining the fray. (Washington Post, 29.4.2003,
p. C2, col. 1)
(3) Boyle testiﬁed that she told Malvo four times that he could be silent or see an
attorney but that Malvo continued to talk about the shootings in a relaxed,
almost convivial way. (Washington Post, 29.4.2003, p. B1, cols 3–4)
(4) You could study this pattern for years and still not wholly understand it. (Ian
Rankin, The Falls, 2001: 240)
Using the structural representation developed in this chapter, discuss the structural
ambiguity of the following examples:1
(5) He added that the looting, though continuing, is much reduced. “You will see
a guy or two carrying a table or chairs. We tell them to put it down and go
home.” (Guardian, 7.5.2003, p. 5, col. 14)
(6) If you feel threatened in a mini cab, ﬁrmly ask the driver to stop and get out.
(Guardian, G2, 7.3.2003, p. 7, col. 2)
Exercise 2 The structure of the sentence (T)
Identify the category of the co-ordinated constituents in the attested example
below. Is this co-ordination compatible with the structure we have elaborated for
the sentence?
(1) Mr Duncan Smith stepped out of central ofﬁce into the autumn sunshine
to declare angrily that he both welcomed the contest and would win it.
(Guardian, 29.10.2003, p. 17, col. 1)
COMMENT
In this example the sequence both . . . and co-ordinates the units (i) [welcomed the
contest] and (ii) [would win it]. In terms of our analysis these units can be taken to
correspond to the constituent labeled I′. Observe, though, that co-ordination in
1
We already introduced these examples in Chapter 2, Exercise 6. Since we have made some
progress in our discussion, it is a good idea to try to reﬁne the answer you might have given
when doing the exercise earlier on.
Exercises 211
TSC03 07/18/2005, 03:34PM211
212 Chapter 3
these examples affects only a partial projection, since the full projection would
comprise the IP layer, that is, it would include the subject.
Exercise 3 Adjuncts, interpretation, and word order
(T)
In this exercise we speculate about the application of adjunction. The exercise is
longer and more discursive than some of the other exercises. The goal of the exercise
is to further explore the consequences of the argumentation elaborated in the
chapter and see where that would lead us.
Identify all the VP-adjuncts in the following examples. Represent adjunction by
means of labeled bracketing.
(1) The student will examine the text very carefully.
(2) The students will very carefully examine the text.
(3) One of the most controversial takeovers in British sporting history was last
night awaiting a government decision. (Based on Guardian, 13.3.1999, p. 1,
col. 1)
(4) It is two-faced of the mayor to one day attack the private sector and the next
day outbid them. (Guardian, 9.12.2000, p. 5, col. 7)
(5) The new rule does not end judicial discretion but it rightly seriously curtails it.
(based on Guardian, 27.3.2001, p. 9, col. 2)
You should end up with the following representations.
(1′) The student will [VP [VP examine the text] [ADVP very carefully]].
(2′) The student will [VP [ADVP very carefully] [VP examine the text]].
(3′) One of the most controversial takeovers in British sporting history was [VP [NP
last night] [VP awaiting a government decision]].
(4′) It is two faced of the mayor to [VP [NP one day] [VP attack the private sector]]
and [VP [NP the next day] [VP outbid them]].
(5′) The new rule does not end judicial discretion but it [VP [ADVP rightly] [VP [ADVP
seriously] [VP curtails it]]].
Example (5′) has two adverb phrases in mid position, rightly and seriously; we can
replace each of the one-word adverbs by more elaborate constituents.
TSC03 07/18/2005, 03:34PM212
Lexical Projections and Functional Projections 213
(6) The new rule does not end judicial discretion but it [VP quite rightly [VP rather
seriously [VP curtails it]]].
The fact that we can replace adverbs that function as adjuncts on their own by a
phrase of which these adverbs are the heads (quite rightly, rather seriously) conﬁrms
the proposal in section 2.1 that “adjunction is the merger of one fully formed
constituent with another fully formed constituent”: in general constituents adjoined
to maximal projections are themselves maximal projections.
(i) Working hypothesis
Adjuncts to projections are themselves projections.
Could the adverbial adjuncts in (5) appear in a different order?
(7) ??The new rule does not end judicial discretion but it [VP seriously [VP rightly
[VP curtails it]]]. (Based on Guardian, 27.3.2001, p. 9, col. 2)
Though (7) is perhaps not completely unacceptable, it is hard to give it an adequate
reading. (5) can be paraphrased by (8):
(8) It is right that the ruling seriously curtails it.
No such obvious paraphrase can be worked out for (7). The reordering of the
adverbials has led to a serious degradation.
The left to right ordering restrictions of the adjuncts seems to reﬂect the semantic
contributions of the adjuncts. We paraphrase (5) as in (8), where the adjective phrase
right modiﬁes the seriousness of the degree of curtailment.
For the interpretation of the structure in (5), the meaning of seriously is ﬁrst
added to the interpretation of the core VP curtail it to give the meaning of the
augmented VP seriously curtail it. Then the resulting unit of meaning is further
modiﬁed by rightly. Below are two representations for a VP with two adjuncts. In
(9a) the adjuncts are both attached to one node; in (9b) they are stacked, with the
lower adjunctY having a closer relation to VP than the higher adjunctX. Which one
of these structures would be preferable and why? When answering the question try
to determine whether your answer addresses the theory as such or the empirical
data that it purports to capture.
(9) a bVP
VPadjunctyadjunctx
VP
VPadjunctX
VPadjunctY
Exercises 213
TSC03 07/18/2005, 03:35PM213
214 Chapter 3
(9b) is preferable since it respects the binary branching hypothesis. Furthermore the
build-up of the structure matches its interpretation in that (9b) suggests that
a higher adjunct (to the left) is added to and modiﬁes the constituent formed by
the VP and the lower adjunct. We say that the higher adjunctX takes scope over the
lower adjunctY. Observe that in (9b) there is only a single VP, headed by the same
verb. The VP consists of a core VP augmented by means of adjuncts. (Think of one
house being extended by a veranda, a garage, a terrace, etc.)
Experiment: Could constituency tests be used to support representation (9b)? Try
to front the VP containing just one adjunct. Also try to pseudo-cleft the VP with
just one adjunct. You will create examples such as (10) and (11).
(10) The new rule does not end judicial discretion but [VP seriously [VP curtail it]],
it [VP rightly] does.
(11) What it rightly does do is seriously curtail it.
Let us experiment with the data. Try inserting the adjunct probably in (1) and in (2)
in a position between the modal auxiliary and the verb. Could probably be inserted
to the right of very carefully in (2)?
(12) a The student will probably examine the text very carefully.
b The student will probably very carefully examine the text.
c *The student will very carefully probably examine the text.
How would you represent the structure of (12a)?
As shown by (12b), the natural position for probably is to the left of very
carefully. Draw a representation for (12b) modeled on the preceding discussion
and using (9b) as your model. Do the following constructed sentences bear on the
structure?
(13) a Speaker A: The student will probably very carefully examine the text.
Speaker B: He will indeed.
Will he really?
No he won’t!
b What the student will probably do is very carefully examine the text.
c *What the student will very carefully do is probably examine the text.
Consider the examples below. Does the left-right sequencing of the underlined
adjuncts correspond to their relative scope? Provide paraphrases to bring out the
relative scope of the adjuncts.
(14) The Hutton inquiry disclosed how the prime minister had earlier personally
ordered Dr Kelly to face a second round of questioning from senior MoD
ofﬁcials about his contacts with the BBC reporter, Andrew Gilligan. (Guardian,
15.8.2003, p. 1, cols 1–2)
TSC03 07/18/2005, 03:35PM214
Lexical Projections and Functional Projections 215
(15) Since the extended saga of the England captain’s move from Manchester to
Madrid bridged the seasons, it could be said that football never really went
away. (Guardian, 9.8.2003, p. 19, col. 1)
(16) In Britain plant-based remedies have been largely ignored, especially since
1945 and the introduction of the NHS, which has traditionally actively discouraged
herbal remedies. (Guardian, G2, 2.9.2003, p. 9, col. 2)
(17) The judges on the state’s highest court yesterday closely questioned both
sides over the constitutionality and legality of a recount. (Guardian, 8.12.2000,
p. 2, col. 3)
(18) A study published last December found that only one of the top 10 medical
schools in the US has clear regulations forbidding researchers from having a
ﬁnancial involvement in the companies they are supposedly impartially testing.
(Guardian, 3.5.2001, p. 8, col. 6)
Recall that in the discussion of the structure of the VP,2
we introduced two layers,
which we kept relatively distinct. There was the core VP with the verb and its complement
(for instance, the direct object) and then there was the augmented VP which
introduced additional information, the adjuncts. This layering was meant to reﬂect
the interpretative relation of the constituents to the verb: complements have a
closer relation to the verb than adjuncts. For one thing, complements receive a
thematic role. The hierarchical relation of adjoined adjuncts also reﬂects their semantic
relation to the sentence: inner adjuncts are more closely connected to V than outer
adjuncts. The analysis is in line with the hypothesis that syntactic structure maps
into meaning.
Recall also from section 3.3 of the current chapter that this mapping hypothesis
leads to a problem with respect to how we treat the subject since the subject has a
closer semantic relation to V than, say, adjuncts. We turn to this in Chapter 4.
Exercise 4 Modals and to (T)
Consider the following examples in the light of the assumption that modal auxiliaries
have the same distribution as the inﬁnitival marker to.
(1) a Do you think John will object to your proposal?
b I don’t expect him to.
(2) a I expected that Mary would have understood the situation by now.
b I expected Mary to have understood the situation by now.
2
Chapter 2, section 2.3.
Exercises 215
TSC03 07/18/2005, 03:35PM215
216 Chapter 3
(3) a I will not give in to such pressure.
b To not give in to such pressure will be hard.
(4) a *I not will give in to such pressure.
b Not to give in to such pressure will be hard.
Exercise 5 Sentential negation (E)
English sentences can be negated in a number of different ways. Identify the
constituent that negates the sentence in the examples below:
(1) I have not actually seen anyone at that shop.
(2) I saw no one at that shop.
(3) I have never actually seen anyone at that shop.
(4) No English students attended the party.
Among the means of expressing negation we ﬁnd two elements in mid position:
not (1) and never (3). In the discussion we proposed that not is like an adverb and
is adjoined to a maximal projection. This means that we should probably propose
a similar analysis for never. Discuss the problems raised by the following examples
for equating the status of not and never:
(5) a He never liked this book.
b *He not liked this book.
c He didn’t like this book.
(6) a Never have I seen such horrible behavior.
b *Not have I seen such horrible behavior.
(7) a I never will accept those conditions.
b *I not will accept those conditions.
c He asked me to accept those conditions but I never will.
d *He asked me to accept those conditions but I not will.
e *He asked me to accept those conditions but I would never.
f He asked me to accept those conditions but I wouldn’t.
(8) a Do you talk to him? Never.
b Do you talk to him? *Not.
TSC03 07/18/2005, 03:35PM216
Lexical Projections and Functional Projections 217
3
Section 1.2.4.4.
4
For discussion of the status of not see also Klima (1964), Pollock (1989), Baker (1991), Ernst
(1992), Haegeman (1995). For an accessible introduction see Haegeman and Guéron (1999).
Table 1 Inﬂectional patterns in Danish and in Icelandic
Danish Icelandic
Present Past Present Past
1sg hører hørte heyri heyrDi
2sg hører hørte heyrir heyrDir
3sg hører hørte heyrir heyrDi
1pl hører hørte heyrum heyrDum
2pl hører hørte heyriD heyrDuD
3pl hører hørte heyra heyrDu
COMMENTS
The data suggest that not and never should be differentiated. Consider the examples
in (5). In the chapter we proposed that the ﬁnite inﬂection lowers onto the lexical
verb in English. We also saw that this lowering is blocked by an intervening not.3
Does the adjunct never also block the lowering of the inﬂection onto V? We must
conclude that the syntactic properties of not and of never as the expressions of
sentential negation are different. In particular, not is special in that it blocks the
lowering of the inﬂection onto V.4
Exercise 6 Comparing languages (T, E)
In Table 1 we give the paradigms for the verb hear in Danish and in Icelandic
(Vikner, 1997). Based on these paradigms and on the conditions we have elaborated
for the movement of the ﬁnite verb to I, which of the sentences in (1) for Danish
and in (2) for Icelandic would you predict to be grammatical?
(1) a Da at Johan ofte spiser tomater.
that John often eats tomatoes
‘that John often eats tomatoes.’
b Da at Johan spiser ofte tomater.
Exercises 217
TSC03 07/18/2005, 03:35PM217
218 Chapter 3
(2) a Ic ad Jón oft bordar tómata.
that Jón often eats tomatoes
‘that John often eats tomatoes.’
b Ic ad Jón bordar oft tómata.
COMMENTS
Danish ﬁnite verbs are invariant, both in the present tense and in the past tense. The
inﬂection is weak and the prediction is that the lexical verb ought not to move to I.
(1a) ought to be grammatical and (1b) ought to be ungrammatical. These predictions
are correct.
In Icelandic the ﬁnite paradigms display a number of different forms, both in the
present tense and in the past tense. So the inﬂection is strong and the prediction is
that the lexical verb ought to move to I. (2a) ought to be ungrammatical and (2b)
ought to be grammatical. These predictions are also correct.5
For more detailed and sophisticated discussion of these data see Vikner (1997)
and the references cited. Vikner provides a careful analysis of precisely what kinds
of verb inﬂections are needed to trigger verb movement and he compares an array
of Scandinavian languages. He also offers a good survey of the relevant literature.6
Exercise 7 Comparing languages (T, E)
We have observed that English grammar has undergone a change through its history.
In Old English, paradigms of verb conjugation displayed a number of different
forms and the lexical verb was mobile. For instance, the inﬂectional system was
stronger in earlier stages of the language7
and this allowed the lexical verb to invert
with the subject. With the loss of its distinct inﬂectional endings, the English lexical
verb has become less mobile.
We might predict an analogous development for languages which today still have
strong inﬂections if, for some reason, their inﬂections were to be eroded through time.
Haitian Creole is a language whose lexicon is strongly based on French but whose word
order is quite different from French, as illustrated in (1)–(4). The inﬂectional system
is given in (5)–(6). Can we account for the difference in the positions of the ﬁnite
verb in the Creole examples in (1)–(4) and their French counterparts in (7)–(10)?
(1) a Bouki deja pase rad yo.
Bouki already irons laundry their
‘Bouki has already ironed their laundry.’
b *Bouki pase deja rad yo.
5
Exercise 16 is also concerned with the distribution of the verb in Icelandic.
6
For additional discussion of Germanic languages see also Rohrbacher (1999). For a critical view
see Bobaljik (2002), Alexiadou and Anagnostopoulou (1998).
7
Cf. Chapter 1, section 3.2, and Chapter 3, section 1.2.4.
TSC03 07/18/2005, 03:35PM218
Lexical Projections and Functional Projections 219
(2) a Mwen toujou ekri manman mwen.
I always write mother my
‘I always write to my mother.’
b *Mwen ekri toujou manman mwen.
(3) a Elèv la byen etidye leson an.
student the well study lesson the
‘The student has studied the lesson well.’
b *Elèv la etidye byen/mal leson an.
(4) a Jak pa janm di bonjou.
Jack NEG never says hello
‘Jack never says hello.’
b *Jak di pa janm bonjou.
(5) mwen /ou /li /nou /yo renmen Boukinèt
I /you /he/she /we/you /they ‘love’ Bouquinette
(6) a Boukinèt ta renmen Bouki (PAST)
b Boukinèt ap renmen Bouki (FUTURE)
(7) a *Jeanne déjà repasse le linge.
Jeanne already irons the laundry
b Jeanne repasse déjà le linge.
‘Jeanne is already ironing the laundry.’
(8) a *Je toujours écris à maman.
I always write to mummy
b J’écris toujours à maman.
‘I always write to mummy.’
(9) a *L’élève bien étudie la leçon.
the pupil well studies the lesson
b L’élève étudie bien la leçon.
‘The pupil studies the lesson well.’
(10) a *Jacques ne jamais dit bonjour.
Jacques NEG never says hello
b Jacques ne dit jamais bonjour.
‘Jack never says hello.’
Exercises 219
TSC03 07/18/2005, 03:35PM219
220 Chapter 3
COMMENTS
When we consider the inﬂectional paradigm in (5) we see that there is no variation
for person and number in Haitian Creole. Nor does the language have a distinct
ending for past or for future, as shown by (6). Given an invariant verb form we
deduce that the inﬂection is weak and hence we correctly predict the lexical verb to
remain in V and to follow adjuncts. In French the ﬁnite inﬂection is strong and
hence we expect the verb to move to I and to precede adjuncts. For more discussion
of the syntax of the Haitian Creole verb see DeGraff (1997), which also contains a
range of references. For an overview of the problems of Creole languages see also
the discussion in DeGraff (1999) and the references cited there.
Exercise 8 Moved constituents (T, L)
We have elaborated a blueprint for the structure of the sentence in which we use
two operations for assembling a sentence. These operations, discussed in section 2
of this chapter, are Merge and Move. So far we have concentrated mainly on the
operation Merge, and we have only invoked the concept Move to account for the
displacement of the verb. Consider the examples below. What is the category of
the underlined constituents? Do you think these constituents occupy their base
positions? If not, try to restore them to their base positions.
(1) Our ﬁshmonger and his staff in Petersﬁeld all wear ties (Letters, October 22)
and very smart they look too. (Guardian, 23.10.2002, p. 9, col. 5, letters to
the editor from David Dew, Horndean, Hants)
(2) Everything I did right for 20 years, he burned up in two or three. (Washington
Post, 29.4.2003, p. C1, col. 1)
(3) The payback burden varies according to earnings later in life, to about £60 a
month, for example, for a civil servant, lower than that for a voluntary sector
worker. The paybacks I don’t think are unreasonable. (Guardian, 20.1.2003,
p. 3, col. 3)
(4) People retain a mystic faith in old exam results or in the snap judgements of
school reports, dogged by lazy character assassinations for the rest of their
days. That is why education should always try to praise a child’s talents, not
brand it with failure. These things she knew well, and it made her a humane
politician and a good education minister. (Guardian, 25.10.2002, p. 7, col. 2)
(5) There are hardly any small movies that people go to, and some of the more
interesting ones they won’t go to. (Guardian, Review, 1.11.2002, p. v, col. 4)
TSC03 07/18/2005, 03:35PM220
Lexical Projections and Functional Projections 221
Examine the underlined sentence-initial constituents in the following examples.
What is their base position? Why do you think that they are displaced?
(6) How can we stop this? (Chicago Tribune, 3.1.2004, S1, p. 33, col. 6)
(7) What kind of delusional rock is Smith living under? (Chicago Tribune, 3.1.2004,
S3, p. 2, col. 6)
(8) I don’t know whom I will be voting for then, but it won’t be Bush. (Chicago
Tribune, 3.1.2004, S1, p. 27, col. 3, letter to the editor)
(9) How good will he become? (Chicago Tribune, 3.1.2004, S3, p. 10, col. 5)
(10) What are they afraid of? (New York Times, 8.3.2004, p. C6, col. 5)
What examples (6)–(10) have in common is that the underlined constituent
originates in a position to the right and has been fronted to a peripheral position in
the sentence. The common factor shared by all these fronted elements is that the
constituent is interrogative and encodes the scope of the question.
(6′) How can we stop this ——?
(7′) What kind of delusional rock is Smith living under ——?
(8′) I don’t know whom I will be voting for —— then.
(9′) How good will he become ——?
(10′) What are they afraid of ——?
In (11)–(15) the underlined constituent has also been displaced. What is the
motivation for the movement? Locate the base position of the moved constituent.
Consider the distance between the moved constituent and its base position.
(11) Who do you think is the more moderate politician? (Guardian, 25.9.2003,
p. 9, col. 6)
(12) What did they think they were making with those girls in there? Animated
cartoons? (Guardian, 13.9.2003, p. 14, col. 2)
(13) At the end of the day everybody eats meat. What do you think your cat or
dog eats? Where do you think that meat comes from? Where do you think
Pedigree Chum comes from? (Observer Magazine, 7.9.2003, p. 41, col. 1)
Exercises 221
TSC03 07/18/2005, 03:35PM221
222 Chapter 3
(14) What has the Chancellor been doing and where does he think the party is
going? (Guardian, 27.9.2003, p. 4, headline)
(15) Some are born great, some achieve greatness and some have greatness thrust
upon them . . . Which do you think you were? (Guardian, G2, 27.10.2003,
p. 6, col. 2)
COMMENT
In (11)–(15) again, the underlined interrogative constituent has been fronted to a
peripheral position in the sentence. What distinguishes these examples from the
earlier ones is that in (11)–(15) the fronted interrogative constituent has been lifted
out of the clause in which it originates. We signal the clause from which the
constituent has been extracted by square brackets in the simpliﬁed representations
below. When a movement operation lifts a constituent out of a clause and into a
higher clause we often talk about long movement. Examples (6)–(10) illustrate
short movement.
(11′) Who do you think [IP —— is the more moderate politician]?
(12′) What did they think [IP they were making —— with those girls in there]?
(13′) What do you think [IP your cat or dog eats ——]?
Where do you think [IP that meat comes from ——]?
Where do you think [IP Pedigree Chum comes from ——]?
(14′) Where does he think [IP the party is going ——]?
(15′) Which do you think [IP you were ——]?
Return to the examples (1)–(5). Identify the example in which the fronted constituent
has undergone long movement.8
Exercise 9 Thematic roles: The expression of AGENT
(T)
In (49) in section 3.1, we deﬁned the thematic role AGENT as follows:
(1) AGENT/ACTOR: the one who intentionally initiates the action.
8
The derivation of examples such as (6)–(15) will be discussed in Chapter 5, section 3.1.
TSC03 07/18/2005, 03:35PM222
Lexical Projections and Functional Projections 223
Consider the generalizations (i) and (ii):
(i) The AGENT thematic role is always realized overtly, and it must be expressed
by the subject of a sentence.
(ii) The subject of a sentence always realizes the AGENT thematic role.
On the basis of the three examples below discuss the validity of generalizations (i)
and (ii).
(1) The news was announced by the director of the company.
(2) The thief was arrested later that night.
(3) John’s analysis of the data surprised everyone.
COMMENT
(1) and (2) illustrate what are called passive sentences. The underlined verbs are
the passive forms of the verbs; they are associated with the passive morpheme -ed.
Typically a situation involving two participants can be presented either by means of
an active sentence or by means of a passive one. Discuss how the thematic roles
of the verb announce are realized in (1a). What about (1b)?
(1) a The director of the company announced the news.
b The news was announced by the director of the company.
It appears that the passive morphology of the verb announced triggers a rearrangement
in the arguments of the verb. Starting from the active verb, we build up the
sentences in the way described in this chapter. Notably the AGENT of the action
expressed by the active verb is realized by the subject NP, the director of the
company, and as such it becomes the anchoring point for the information given in
the sentence. (1a) gives information about the entity ‘the director of the company’.
In the passive version of the sentence, (1b), the AGENT of the activity is not expressed
as a subject, rather it is made explicit by means of a PP introduced by the preposition
by. The NP that denotes the entity affected by the activity, the THEME, now occupies
the canonical subject position and has become the anchor of the information given
in the sentence. (1b) is about the entity ‘the news’.
(2) is also a passive sentence. Is the AGENT expressed? Can it be inferred? We
conclude that in passive sentences the by-phrase expressing the AGENT need not be
present. However, even when not overtly expressed, the AGENT of arrest can be
inferred. For more discussion of such inferred AGENTS see Exercise 11.
In (3) the verb surprise does not take an AGENT argument: the full NP John’s
analysis of the data does not denote ‘the one who intentionally initiates the action’.
The subject denotes an event, events do not have intentions, so events cannot
Exercises 223
TSC03 07/18/2005, 03:36PM223
224 Chapter 3
‘intentionally initiate the action’. The subject NP John’s analysis of the data expresses
a CAUSE. In the passive counterpart of (3) the CAUSE will also be expressed in a
by-phrase:9
(4) Everyone was surprised by John’s analysis of the data.
Now let us turn to the NP John’s analysis of the data. This NP contains as its head
the noun analysis. In terms of its form, this noun is related to the verb analyze. A
noun whose form is related to a verb is called a deverbal noun.10
The verb analyze
is associated with two thematic roles: the AGENT, the person doing the analysis, and
the THEME, the entity that is being analyzed.
(5) a John analyzed the data.
b analyze thematic role 1: John
thematic role 2: the data
The thematic roles of the verb analyze are retained in the deverbal noun analysis.
Turning to the NP John’s analysis of the data we see that the AGENT of analysis is
realized by the prenominal genitive NP John’s and its THEME is realized by the NP the
data, contained in a PP of the data.11
(6) a John’s analysis of the data
b analysis thematic role 1: John’s
thematic role 2: (of) the data
Exercise 10 Thematic roles: The expression of AGENT
(T, E)
In (49), in section 3.1, we deﬁned the thematic role AGENT as follows:
(1) AGENT/ACTOR: the one who intentionally initiates the action.
Comment on the realization of the AGENT role associated with the underlined verbs
in the following examples.
(2) Dawson was invited by Sven-Goran Eriksson to the England get-together in
November. (Guardian, 7.1.2003, p. 14, col. 2)
9
For the realization of arguments of such psychological verbs see also Belletti and Rizzi (1988).
Though based on Italian data the conclusions of this paper are relevant for English.
10
For some discussion of the formation of deverbal nouns see Chapter 2, Exercise 17.
11
For a thorough discussion of the thematic structure of deverbal nouns see, among others,
Williams (1981), Grimshaw (1990), Alexiadou (2001).
TSC03 07/18/2005, 03:36PM224
Lexical Projections and Functional Projections 225
(3) The center is at a public facility, Huilongguan Hospital, and is being funded
quite willingly by Beijing’s city government. (San Francisco Chronicle,
28.11.2002, p. F6, col. 1)
(4) The Which? Web Trader scheme, which has been running since July 1999, is
being closed down by the Consumers’ Association because it is too expensive
to maintain. (Guardian, 7.1.2003, p. 12, col. 5)
(5) A spokesman for Mr Bing said the money would be paid into a trust. (Guardian,
18.12.2002, p. 3, col. 4)
(6) The culture committee’s motion was drafted by members of prime minister
Silvio Berlusconi’s Forza Italia party. (Guardian, 18.12.2002, p. 6, col. 3)
COMMENT
(2)–(6) illustrate what are called passive sentences. The underlined verbs are the
passive forms of the verbs; they are associated with the passive morpheme -ed.
Typically, a situation involving two participants can be presented either by means of
an active sentence or by means of a passive one: (7) would be another illustration:
(7) a The director of the company announced the news.
b The news was announced by the director of the company.
As we discussed in Exercise 9, the passive morphology of the verb gives rise to a
rearrangement in the arguments of the verb. In an active sentence, the AGENT of
the action denoted by the verb is realized by the subject NP, the director of the
company in (7a), and it is the anchoring point for the information given in the
sentence. (7a) gives information about the entity ‘the director of the company’. In
the passive version, (7b), the AGENT of the activity is not realized as a subject; rather
it is made explicit by means of a PP introduced by the preposition by. The NP that
denotes the entity affected by the activity, the THEME, now occupies the subject
position and is the anchor of the information given in the sentence.
Consider the underlined passive verb in (8). Who is the AGENT? Is the AGENT
expressed in the example?
(8) A student was arrested on suspicion of murdering special branch ofﬁcer Stephen
Oake in Manchester. (Guardian, 16.1.2003, p. 1, col. 4)
As already mentioned with respect to example (2) in Exercise 9, a passive verb does
not require the presence of a by phrase. We propose that in the passive sentence,
the by phrase, though it realizes a thematic role, is an adjunct. In (8) the by phrase
is not realized because we can infer from the choice of verb, arrest, that the AGENT
was the police.
Exercises 225
TSC03 07/18/2005, 03:36PM225
226 Chapter 3
Consider the underlined passive verbs in the following sentences. Is the AGENT of
the action denoted by these verbs overtly expressed? Which component of the
sentence do the bold faced parts of the sentence modify?
(9) She was deliberately killed. (Guardian, 16.9.2003, p. 8, col. 7)
(10) There is some technical study till mid-morning teabreak, then lunch at 1 p.m.,
which is cooked together but bought from a kitty to which all the men
contribute. (Guardian, 7.9.2002, p. 7, col. 7)
(11) The victims were shot using a high powered riﬂe. (Guardian, 10.10.2002,
p. 2, col. 5)
(12) Using plastic gloves from the Harrods food hall, the contents were then
inspected and shown to Mr Fayed. (Guardian, 14.8.2002, p. 5, col. 7)
(13) Mr. Mansﬁeld said Blakely’s behaviour was intentionally directed at Ellis, and
given the history of the relationship it amounted to provocation. (Guardian,
17.9.2003, p. 6, col. 4)
Sentences (9)–(13) suggest that even when the AGENT of a passive verb is not
expressed by a by-phrase, it still remains accessible for modiﬁers. In (9), for instance,
the adverb deliberately modiﬁes the attitude of the AGENT of kill, even though that
AGENT is not overtly expressed. Similarly, in (10) the intended interpretation is that a
number of people cook lunch together. Again together bears on the understood
AGENT of cook. In (11) using a high powered riﬂe is a non-ﬁnite string which lacks an
overt subject. The verb using is the present participle. Using a high powered riﬂe is
a non-ﬁnite clause, in particular it is a participial clause. The subject of using is
implicit. We interpret its subject as being identical to the implied AGENT of shot:
those who shot the victims were using a high powered riﬂe.
We conclude that the understood AGENT in a passive sentence remains accessible
to modiﬁcation.
Exercise 11 Thematic roles: The implied AGENT
(T, E, presupposes Exercise 10)
In Exercise 10, we discovered that even when implicit, the AGENT of an activity might
be accessible for syntactic processes such as modiﬁcation by an adverbial or by a
participial clause. However, care must be taken when we state that an implicit
element is syntactically accessible. In particular, not every element that we can infer
from the meaning of a sentence is “accessible.” Compare the expression of the
AGENT role in the following examples.
TSC03 07/18/2005, 03:36PM226
Lexical Projections and Functional Projections 227
(1) a We ﬁnally opened the door with the headmaster’s key.
b The door was ﬁnally opened by the junior staff member.
c The headmaster’s key ﬁnally opened the door.
In (1a) and in (1b) the AGENT of open is overtly expressed. In (1a) it is the subject of
the sentence, we; in (1b) the AGENT is expressed by means of an adjunct by phrase;
in (1c) the subject, the headmaster’s key, realizes the INSTRUMENT role. From the fact
that an INSTRUMENT is referred to we infer that an AGENT must be involved in the
action. The AGENT is the one who uses the instrument. So we can say that in (1c) an
AGENT can be inferred.
These data might lead us to expect that in the same way that the understood
AGENT is accessible for modiﬁcation in a passive sentence, the inferred AGENT in (1c),
which contains a reference to an instrument, will be accessible for modiﬁcation.
Comment on this expectation using the following data.12
(2) a Expecting the worst, we ﬁnally opened the door with the headmaster’s key.
b Expecting the worst, the door was ﬁnally opened by the junior staff member.
c *Expecting the worst, the headmaster’s key ﬁnally opened the door.
COMMENT
There is a difference between the way an AGENT may be implicit in a passive
sentence, in which it remains accessible for purposes of modiﬁcation (1b), and the
way the existence of an AGENT is inferred in a sentence such as (1c), in which the
subject realizes an INSTRUMENT. Even under the assumption, which is possible though
not necessary, that the headmaster is the AGENT in (1c), (1c) becomes unacceptable
when we insert an adjunct modifying the AGENT (2c). Passivization has a way of
preserving the AGENT in the sentence.13
Exercise 12 Be +++++ passive VP (T)
Identify the constituents that are co-ordinated by the underlined conjunction and in
the examples below.
(1) The scheme will start in London and be extended later to other parts of
England. (Guardian, 3.10.2002, p. 3, col. 2)
(2) Mitrokhin has been interviewed on television and co-written a book with
Professor Christopher Andrew. (Observer, Review, 16.9.2001, p. 15, col. 3)
(3) It is understood that one individual who sent an email containing footballers’
names has been warned he could face a libel action and been asked to hand
12
See also Reinhart (2000) for discussion of the relation between INSTRUMENT and AGENT.
13
For a discussion of passivization see among others Baker, Johnson, and Roberts (1989).
Exercises 227
TSC03 07/18/2005, 03:36PM227
228 Chapter 3
over the names of the people he sent the message to. (Guardian, 2.10.2003,
p. 3, col. 2)
KEY AND COMMENTS
The conjoined constituents are all VPs:
(1′) [VP start in London]
+ [VP be extended later to other parts of England]
(2′) [VP been interviewed on television]
+ [VP co-written a book with Professor Christopher Andrew]
(3′) [VP been warned he could face a libel action]
+ [VP been asked to hand over the names of the people he sent the message to]
Two of the VPs are simply projections of a lexical verb:
(4) a [VP start in London]
b [VP co-written a book with Professor Christopher Andrew]
The other VPs contain the auxiliary be and a passive verb form.
(5) a [VP be extended later to other parts of England]
b [VP been interviewed on television]
c [VP been warned he could face a libel action]
d [VP been asked to hand over the names of the people he sent the message to]
Following the discussion in the chapter, we assume that an auxiliary is a verb and
heads its own VP. In (5) the auxiliary be selects a passivized VP, with which it forms
a constituent:
(6) VP
VPV
V . . .
extended
interviewed
warned
asked
be
been
been
been
to other parts of England
on television
he could face a libel action
to hand over the names
of the people he sent the message to
TSC03 07/18/2005, 03:36PM228
Lexical Projections and Functional Projections 229
Discuss how VP ellipsis in the following extract provides further evidence for the
hypothesis that passive be forms a constituent together with the lexical VP headed
by the passive participle:
(7) More than in any recent presidential election, the critical economic issue this
year boils down to whether middle-income people think they are being
squeezed. President Bush passionately argues they are not. (Adapted from
New York Times, 1.8.2004, section 3, col. 2)
Exercise 13 Lexical heads and functional structure
(E)
The merger operation that assembles the clause starts from a lexical element, a
verb. A lexical head V projects a structure, VP, which is selected by a functional
head. Merger of VP and I integrates the VP projection into a functional projection,
IP. It would be interesting to examine if this pattern of a functional projection
dominating a lexical projection is general. Do projections of other lexical elements
also require a dominating functional projection? The noun is a case in point. The
noun analysis in (1a) is the head of an NP. In (1b) we provide the type of structure
proposed for the NP. Does the lexical projection NP merge with a functional structure?
Try to rephrase the NP in (1a) by means of a sentence. To which sentential constituent
would the genitive NP the student’s correspond? And, in a sentence, which constituent
would match the PP of the problem?
(1) a the student’s analysis of the problem
b NP
N′NP
N PP
analysis of the problemthe student’s
(1a) is closely similar in composition to the sentence (2a), whose structure is given
in (2b):
(2) a The student has analyzed the problem.
b [IP The student [I′ [I has] [VP analyzed the problem]]].
We might try to generalize the hypothesis that a lexical projection is inserted as the
complement of a functional head and also merge the NP with a functional head.
Exercises 229
TSC03 07/18/2005, 03:36PM229
230 Chapter 3
We could propose something along the lines of (1c), for instance, in which we
associate the genitive marking with a functional head, which we label F. Draw the
tree diagram that would correspond to (1c).
(1) c [FP the student [F′ [F’s] [NP analysis of the problem]]]
Based on the discussion of the structure of the NP in the chapter, list the constituents
that were shown to alternate with the genitive in (1a). Are these constituents all
phrasal constituents?14
Exercise 14 Non-ﬁnite inﬂections (E)
Consider representation (69b) from the chapter, repeated here as (1):
(1) IP
I′NP
I VP
VPV
V VP
talking
talking about this problem
about this problemhavepast
have + ed = had
had
We
= We
V PP
been
been
One point that emerges from the discussion in the chapter is that the ﬁnite inﬂection
of the verb should be represented separately from V, and that this separation
14
The hypothesis that NPs must also be associated with functional structure was elaborated by
Abney (1987). For introductory discussion of the structural build-up of the nominal projection
see Haegeman and Guéron (1999) and Bernstein (2001). For more advanced discussion see,
among others, Szabolcsi (1994), Giorgi and Longobardi (1991), Gavruseva (2000), and the
references cited there.
TSC03 07/18/2005, 03:37PM230
Lexical Projections and Functional Projections 231
applies both to lexical verbs and to auxiliaries. In sections 4.3. and 4.4 we made the
point that if we represented inﬂectional endings separately for lexical verbs and
if we failed to do so for inﬂections of auxiliaries, we would have to provide an
account for this different treatment. The simpler theory is one in which all ﬁnite
inﬂections are represented as separate. The one exception we did allow for is modal
auxiliaries as these are inherently tensed and they lack non-ﬁnite forms.
However, careful readers will have detected another inconsistency. If ﬁnite inﬂections
are to be represented separately from the associated verb stem, should the
same not apply to non-ﬁnite inﬂections? Should we not treat all verbal inﬂections as
separate from the stems they are associated with? What would the consequences
be for the structure in (1)?15
Exercise 15 The structure of sentences: Extending
the data (E)
Discuss the problems raised for the representation of sentence structure elaborated
so far by the position of the underlined elements in the following examples:16
(1) Michigan coach Lloyd Carr says sorry, but he simply can’t go against the
coaching association’s policy and vote for Southern California as college
football’s No 1 team. (Chicago Tribune, 3.1.2004, S3, p. 1, col. 1)17
(2) Jarvis today will wipe away some of the immediate gloom surrounding the
support services group by announcing a £300 million contract to build 3,000
rooms for the University of Lancaster. (Guardian, 17.10.2003, p. 13, col. 1)
(3) A year ago, these streets at 4 a.m. of a Friday night were raucous and reeling
with drunks. (Based on Atlanta Journal-Constitution, 23.11.2003, p. F1,
col. 5)18
(4) I never could make out what those damned dots meant. (Independent,
14.4.2001, p. 20, col. 6)19
(5) I expect house prices over the year to rise by 4 per cent. (Times, Times 2,
27.12.2000, p. 7, col. 1)
15
For some discussion see Belletti (1990) and Friedemann and Siloni (1993).
16
For some discussion see Pollock (1997). We return to some of the examples in Exercise 22 of
Chapter 4.
17
Cf. Chapter 4, Exercise 22 (7).
18
For the position of a year ago see also Chapter 2, Exercise 15.
19
Cf. Chapter 4, Exercise 22 (8).
Exercises 231
TSC03 07/18/2005, 03:37PM231
232 Chapter 3
(6) American ofﬁcials also have cited a possible business dispute as a reason for
the disappearance of the Angola jet. (Chicago Tribune, 3.1.2004, section 1,
p. 6, cols 1–2)
(7) We expect actors in real life to be like the characters they play, and of
course, they’re not. (Adapted from Sunday Times, News Review, 25.1.2001,
p. 3, col. 6)
(8) Today Gulliver still can barely stand to be among the Yahoos. (Guardian, G2,
31.8.2004, p. 12, col. 1)20
(9) The AP trophy probably will be presented to USC in Los Angeles, not New
Orleans. (Chicago Tribune, 3.1.2004, S3, p. 2, col. 3)
(10) Parents and students probably wouldn’t do that. (New York Times, Education,
1.8.2004, p. 16, col. 2)
Exercise 16 Object shift and verb movement in
Icelandic (E)
Consider the Icelandic example (1a). How could we explain the position of the verb
to the left of the negation marker ekki (‘not’)?
(1) a Jón las ekki bækurnar.
John read not books
‘John didn’t read the books.’ (Collins and Thráinsson, 1993: 132, their (2b))
Draw a tree diagram for the sentence. Discuss how (1b) differs from (1a).
(1) b Jón las bækurnar ekki.
John read books not
‘John didn’t read the books.’ (Collins and Thráinsson, 1993: 132, their (2a))
KEY AND COMMENTS
In (1a) the lexical verb las (‘read’) precedes the negation marker ekki. We saw
in Exercise 6 that Icelandic verbs have a strong inﬂection. We can assume that
the verb has moved out of the VP to I. In (2a) we use an arrow to show verb
movement.
20
Cf. Chapter 4, Exercise 22 (9).
TSC03 07/18/2005, 03:37PM232
Lexical Projections and Functional Projections 233
21
For the discussion of the relevance of object shift in the derivation of the Germanic OV
languages see Exercise 18 of Chapter 2.
22
For alternative proposals see Holmberg (1986, 1999), Collins and Thráinsson (1993), Diesing
(1997).
Exercises 233
(2) a IP
I′NP
I VP
VPAdvP
V NP
ekkilasJón bækurnar
To derive the word order in (1b), we need an additional movement: we have to
shift the object NP, bækurnar (‘the books’), leftward to an IP-internal position. The
operation which moves the object to the left within the IP domain is called object
shift.21
One possibility would be to propose that the object NP left-adjoins to the
VP.22
The movement of the verb and that of the object are both indicated by
arrows in (2b).
(2) b [IP Jón [I′[Ilas] [VP [NP bækurnar] [VP ekki [VP [V ——] [NP ——]]]]]].
Consider the examples in (3). Has the lexical verb moved to I in (3a)? Has object
shift taken place in (3a)? Consider (3b). Is object shift possible?
(3) a Jón hefur ekki lesid bækurnar.
John has not read the books
‘John hasn’t read the books.’
b *Jón hefur bækurnar ekki lesid.
John has the books not read.
In (3) the participial form of the lexical verb lesiD (‘read’) has not left the VP. In
Icelandic there is a correlation between object shift to the left of the VP and movement
of V to I. If there is no V-to-I movement, then object shift is not possible. This
correlation is often referred to as Holmberg’s generalization (Holmberg, 1986).
TSC03 07/18/2005, 03:37PM233
234 Chapter 3
Exercise 17 Variations in subject positions (L)
Consider the following examples. Identify the subject of the underlined verbs.
(1) a Enough moralists were objecting to the theatre as it was.
(2) a A bomb is waiting to go off in consumers’ pockets.
(3) a Some men were selling special editions of the evening paper.
(4) a A lot of things are going on around you.
Now consider the underlined strings in the (b)-sentences, which were the source for
the (a)-sentences above. The (b)-sentences all contain a variant of the (a)-sentences.
Describe the differences between the (a)-sentences and the (b)-sentences.
(1) b Putting a rogue on the stage as a hero was really daring, there were enough
moralists objecting to the theatre as it was. (Independent 15.7.2004, p. 11,
col. 2)
(2) b William Ostrom . . . admits there is a “bomb” waiting to go off in consumers’
pockets. (Guardian, 6.5.2003, p. 14, col. 4)
(3) b When I came out there were men selling special editions of the evening
paper. (Guardian, Review, 31.5.2003, p. 5, cols 3–4)
(4) b If you are staying with a family there are a lot of things going on around
you. (New York Times, 1.8.2004, Travel section, p. 4, col. 5)
KEY AND COMMENTS
We will discuss example (1). In (1a) the subject is the NP enough moralists, and the
lexical verb is objecting. The NP enough moralists occupies what we call the canonical
subject position, the speciﬁer of IP. When we form a direct question based on (1a),
SAI affects the relative order of the NP enough moralists and the ﬁnite auxiliary
were:23
(1) c Were enough moralists objecting to the theatre?
(1d) isolates the underlined section in (1b):
(1) d There were enough moralists objecting to the theatre.
23
We turn to the position of the inverted auxiliary in Chapter 5.
TSC03 07/18/2005, 03:37PM234
Lexical Projections and Functional Projections 235
In (1d), the NP enough moralists follows the ﬁnite auxiliary. This means that it does
not occupy the speciﬁer of IP, the canonical subject position. Rather, in the canonical
subject position we ﬁnd the word there. When applied to (1d), SAI changes the
order of the auxiliary were and the element there.
(1) e Were there enough moralists objecting to the theatre?
On the other hand, in terms of its meaning, the NP enough moralists refers to the
AGENT
24
of objecting, and it agrees with the auxiliary were, suggesting that it too is
a subject. So we are faced with a sentence with two constituents that qualify as
subjects. In turn, this means that we must provide two positions for these subjects
in the structure of the sentence: the canonical subject position, which is the speciﬁer
of IP, and a second subject position which in (1b) is immediately next to the verb.
We will examine such sentences in Chapter 4, section 3.2.2.25
Exercise 18 Modern Greek word order (T, L)
Consider the following examples in Modern Greek (Alexiadou, 1997). Does the
Modern Greek verb move to I?
(1) O Petros egrafe panda megala grammata.
the Peter-NOM write-IMP-3SG always long letters
‘Peter always wrote long letters.’ (Alexiadou, 1997: 91 (17b))
(2) O Petros etroge sinithos sika.
the Peter-NOM ate-IMP-3SG usually ﬁgs
‘Peter usually ate ﬁgs.’ (Alexiadou, 1997: 93 (23a))
(3) O Janis agorase kthes to aftokinito.
the John-NOM bought-3SG yesterday the car
‘John bought the car yesterday.’ (Alexiadou, 1997: 109 (62a))
(4) O sismos katastrepse entelos to horjo.
the earthquake destroyed-PERF-3SG completely the village
‘The earthquake destroyed the village completely.’ (Alexiadou, 1997: 131 (14a))
Now discuss the word order displayed by the examples below. As shown by the
English translations, these are to be interpreted as declarative sentences.
24
Cf. section 3.2 for thematic roles.
25
Exercise 18 also suggests that we need more than one subject position.
Exercises 235
TSC03 07/18/2005, 03:37PM235
236 Chapter 3
(5) Filise o Petros ti Maria.
kissed-3SG the Peter-NOM the Mary -ACC
‘Peter kissed Mary.’ (Alexiadou, 1997: 57 (15b))
(6) Ektise i Maria to spiti.
built-3SG the Mary the house
‘Mary built the house.’ (Alexiadou & Anagnostopoulou, 1998: 495 (7c))
(7) Diavaze sihna o Janis to vivlio.
read-3SG usually the John-NOM the book-ACC
‘John was usually reading the book.’ (Alexiadou, 1997: 62 (29))
(8) Efage kala o Janis.
ate-3SG well the John
‘John ate well.’ (Alexiadou, 1997: 131 (13a))
KEY AND COMMENTS
Examples (1)–(4) suggest that the lexical verb moves to I in Modern Greek. As a
result of the movement, the verb can be separated from its object by an intervening
adverb. In (5) and in (6) the verb precedes the subject. Moreover, in (7) and in (8)
the verb is separated from the subject by an adverb. If the verb moves from V to I
and if the relevant adjuncts in (7) and (8) are VP-adjoined, then the position of the
subject in these examples is unexpected. The discussion concerning the relation of
the subject to the verb and the VP in section 3.3 bears on this.26
26
The English data in Exercise 17 above also suggest the need for postulating more than one
subject position. In Exercise 16 of Chapter 4 we return to the Modern Greek data discussed in
the exercise. In Chapter 4 we will introduce the hypothesis that in addition to the canonical
subject position, the speciﬁer of IP, there is a second subject position, the speciﬁer of VP. See
also Exercises 6 and 7 in Chapter 4.
TSC03 07/18/2005, 03:37PM236
Reﬁning Structures 237
4 Reﬁning Structures: From One Subject
Position to Many
Discussion
Contents
0 Introduction 238
1 Recapitulation 239
2 Mapping Form onto Meaning: A Problem 241
2.1 Mapping form to meaning: Core constituent and periphery 241
2.2 Speciﬁers 243
2.3 Interpretation 246
3 The VP-Internal Subject Hypothesis 247
3.1 Displaced constituents: Movement and copies 247
3.2 Empirical support for the VP-internal subject hypothesis? 250
3.2.1 Floating quantiﬁers 251
3.2.2 Existential sentences 253
3.2.2.1 Thematic subject and grammatical subject 253
3.2.2.2 A note on determiner choice 257
3.2.2.3 A note on verb choices and existential patterns 258
4 Subject Positions and Auxiliaries 262
4.1 Auxiliaries and step-by-step movement of the subject 262
4.2 Empirical support for step-by-step movement. 265
5 Summary 266
TSC04 07/18/2005, 03:39PM237
238 Chapter 4
0 Introduction
This chapter returns to the representation we have elaborated for the structure of
the sentence. We will try to reﬁne the representation in order to capture the mapping
of form to meaning. In a way this chapter is like a discovery trail through
sentence structure. We start out from known territory and we go out into unknown
territory. We will ﬁrst look at some empirical and theoretical evidence to suggest
that sentences must have more articulated structures than we had thought. In particular,
we will take up the conclusion reached in Chapter 3, section 3.3. Up until
that point we had been assuming that the subject occupies the speciﬁer position
of IP or SpecIP for short. As subjects typically occupy that position, SpecIP is the
canonical subject position in English.
In the discussion in Chapter 3 we observed that while we had been identifying
one position, SpecIP, as the subject position, it turns out that there are theoretical
reasons for assuming that we need to also provide for a subject position in the VP.
We will pursue this hypothesis here and take it to its consequences. While doing
this, we will keep constant the basic principles of structure building that we have
elaborated. Throughout the chapter, we will follow the same methodology: on the
basis of the premises we start out with, and coupled with new empirical discoveries,
we will formulate a hypothesis, test it, ﬁnd problems with it, reformulate it, etc.
We will ﬁnd arguments to introduce more subject positions. It is important to
bear in mind, though, that our theory as such will remain unchanged. What changes
is the implementation of the theory. This chapter provides an illustration of how
linguistic theories develop through the interplay of two components: evidence from
empirical data is one factor in the elaboration of a theory, but we also use conceptual
arguments, that is, arguments driven by the kind of theory we have already
developed.
We are trying to illustrate as precisely as possible the concrete process of elaborating
a theoretical proposal. The hypotheses we will be elaborating are compatible
with current thinking about syntax but they do not exhaustively represent the most
recent developments of the theory. Our goal is to examine how a certain conclusion
is arrived at using some speciﬁc analytic devices, but if we were to pursue the same
strategy of thinking yet further we would see that what we propose here can again
be challenged and must be modiﬁed. So even at the end of the discussion, there is
still work to be done. This is an inevitable component of scientiﬁc research. Even if
we were to present the results of the most recent work in syntax, we would still
have to evaluate them and test them and we would discover that modiﬁcations are
needed. Throughout all scientiﬁc work, an attitude of constructive criticism prevails.
That is the nature of scientiﬁc research: it is never really “ﬁnished.” We repeat
the quotation given in the exercise section of Chapter 1:
TSC04 07/18/2005, 03:39PM238
Reﬁning Structures 239
In any branch of science there are only two possibilities. There is either nothing left to
discover, in which case why work on it, or there are big discoveries yet to be made, in
which case, what the scientists say now is likely to be false. (Nigel Calder, author of
Magic Universe: The Oxford Guide to Modern Science. Cited in the Guardian, 3.6.2004,
p. 6, col. 2)
This chapter will be different from the previous ones in that it is more of a “narrative.”
The idea is to trace step by step the reasoning that guides the construction
of a theory. While doing this, we will point out the advantages and disadvantages
of each modiﬁcation as we go along.
The chapter is organized as follows: Section 1 is a recapitulation of the theory we
have elaborated so far. Section 2 discusses the mapping of the form of the sentence
to its meaning, that is, the ﬁt between the structure elaborated and the interpretation
associated with that structure. In particular, we raise the question why the
subject, which is thematically related to the verb, is located strictly outside the VP.
Section 3 provides a solution to the mapping problem and proposes that the subject
is inserted in the speciﬁer of the VP and that it is subsequently moved to the speciﬁer
of IP. Thus, the subject has a close link to V as well as being seen to occupy a VPexternal
position. Section 4 ties in the discussion in the preceding section
with the discussion of sentences with multiple auxiliaries in Chapter 3, section 4.
Section 5 is a summary.
1 Recapitulation
Discuss how the examples in (1) can be argued to offer support for the claim that
the underlined strings of words are constituents:1
(1) a Two decades of ﬁnancial squeeze has eroded academic standards and seriously
damaged common-room morale. (Guardian, 26.10.2002, p. 13, col. 1)
b You could study this pattern for years and still not wholly understand it.
(Ian Rankin, The Falls, 2001: 240)
c “They must talk about it, and talk about it they must,” he said. Food
for thought, there! It’s a phrase that could add a measure of gravity to any
press conference. “We must do this, and do this we must.” (Simon Hoggart,
Guardian, 29.1.2003, p. 2, col. 5)
d Lawyers who’ve handled arbitration claims for years . . . are getting very busy;
lawyers who never have are joining the fray. (Washington Post, 29.4.2003,
p. C2, col. 1)
1
For diagnostics of structure see Chapter 2, section 1.
TSC04 07/18/2005, 03:39PM239
240 Chapter 4
In these examples, the underlined strings of words can be seen to be affected by a
number of operations: in (1a–b), for instance, the relevant constituent is co-ordinated
with another constituent, in (1c) VPs are fronted, and in (1d) the underlined
string corresponds to the constituent omitted after the auxiliary have in the second
sentence. The underlined constituents are all verb phrases: their head is a verb.
On the basis of the data discussed in Chapter 3 we arrived at the conclusion that
sentences are put together by merging pairs of constituents, according to the following
schema:
(2) a IP
I′NP
VPI
As illustrated by (1c), the structures formed according to the schema in (2a) can be
modiﬁed by movement operations: in (1c) VPs are fronted, they are moved to a
position to the left of the subject. We might provisionally say that such VPs are
adjoined to IP.2
(2) b [IP [VP Do this] [IP [NP we] [I′ must [VP ——]]]].
The position in which a constituent is merged is called its base position. The position
that it attains after movement is called its landing site. In (2b) we show that
originally the VP do this is the complement of the modal must by means of the
arrow which links the base position [VP ——] with the landing site of the moved
element, a position adjoined to IP. To represent the relation between the fronted VP
and its base position, we could also use an alternative representation in which we
actually show the VP in two positions: its base position, where it is merged as the
complement of the modal auxiliary must, and the moved position adjoined to IP, in
which it is pronounced or spelt out. To show that the VP is not pronounced in its
base position we use the notational device of strikethrough: do this.
(2) c [IP [VP Do this] [IP [NP we] [I′ must [VP do this]]]].
We might ask ourselves if it wouldn’t be simpler to just use a representation in
which all of the original VP in the base position disappears?
(2) d [IP [VP Do this] [IP [NP we] [I′ must]]].
2
See discussion in Chapter 2, Exercise 15, and in Chapter 3, section 1.2.3.4.
TSC04 07/18/2005, 03:39PM240
Reﬁning Structures 241
(2d) suggests a radical change in the structure of the sentence. So far we had been
assuming that the core of the sentence is the VP and that on top of this we construct
a functional projection, IP. The original structure is fully preserved in representation
(2c): notably, VP is still the starting point of the construction and it remains the
complement of the modal. In representation (2d) the relationship between must and
the VP has been irretrievably destroyed: according to (2d), the sentence is simply a
projection of the modal in I, with the VP do this as an adjunct. (2d) does not
preserve the format of the sentence before the movement; the strikethrough representation
in (2c) does allow us to trace back the initial structure at the basis of the
sentence. (2c) gives us the full information on the derivation of the sentence: how it
has been built up by a combination of the operations Merge and Move. In this
chapter and in Chapter 5 we will discuss a number of instances of movement and
we will show the advantages of preserving the pre-movement structure.
2 Mapping Form onto Meaning: A Problem
2.1 Mapping form to meaning: Core constituent and
periphery
Consider examples (1a) and (1b) above. The co-ordinated VPs in these examples
are:
(3) a [VP eroded academic standards]
b + [VP seriously damaged common-room morale]
(4) a [VP study this pattern for years]
b + [VP still not wholly understand it]
In (3b) the adjunct seriously is part of the co-ordinated constituent, hence it is taken
to be part of the VP. We assume that seriously is left-adjoined to the VP. Similarly,
in (4b) the adjunct still, the negative marker not, and the degree adverb phrase
wholly are left-adjoined to VP. We have been working on the assumption that
adjoined constituents are less central to a phrase than the core constituents of that
phrase, i.e. the head (here the verb) and its complement (here the direct object NP).
A maximal projection is fully formed even if it lacks any adjuncts. The internal
layering of the components of a projection is mapped onto the interpretation: the
action described by the VP is deﬁned by the verb and the complement, while the
adjuncts provide additional ﬁne-tuning.
TSC04 07/18/2005, 03:39PM241
242 Chapter 4
(3) c VP
VPAdvP
seriously
NPV
common room moraledamaged
(4) c VP
VP3
AdvP
still
VPAdvP
not
VPAdvP
wholly
NPV
understand it
In (4c) the core VP understand it is “augmented” with three peripheral adjuncts, still,
not, wholly. The interpretation of the VP can be read off from the tree: in (4c), the
core meaning of the VP is encoded by the core VP, understand it, which denotes a
psychological state. The adjunct wholly signiﬁes that the state of the understanding
covers the complete extent of the object it. The negation marker not is added to the
unit wholly understand it and it denies that complete understanding. The adjunct still
speciﬁes a temporal interval. Observe that not negates the content of the VP wholly
understand it and that still indicates the time interval of the content of the augmented
VP not wholly understand it. We say that an adjoined constituent has scope over
the constituent it adjoins to. Still has scope over the VP not wholly understand it;
not has scope over the VP wholly understand it. If you compare the scope of the
words still and not, the scope of still is wider than that of not. Or conversely, the
scope of not is narrower. We can represent scope relations as in (4d):4
(4) d still > not
3
This representation of negation is a simpliﬁcation. See Chapter 3, Exercise 5. For more careful
analysis see Pollock (1989), Haegeman (1995), Rowlett (1998), Zanuttini (1997a, b).
4
Exercise 3 of Chapter 3 also illustrates how structure expresses scope relations.
TSC04 07/18/2005, 03:39PM242
Reﬁning Structures 243
Similarly not takes scope over the VP wholly understand it while wholly only takes
scope over the VP understand it. The scope of not is wider than that of wholly.
(4) e not > wholly
We can combine (4d) and (4e) to show the relative scope of the stacked VP-adjuncts:
(4) f still > not > wholly
The syntactic structure we have postulated can be seen as the input to the interpretive
mechanisms that map linguistic forms into the corresponding interpretations.
The structure allows us to assign scope to the adjuncts.5
2.2 Speciﬁers
Recall that in the preceding chapter6
we came across a number of theoretical questions
with respect to the representations of VP structures such as those in (3c) and
(4c). In fact, when elaborating the structure for the NP in Chapter 2 we ended up
with a structure in which, for instance, a prenominal genitive NP occupied a speciﬁer
position. We assumed that there is one such speciﬁer position per NP, and that the
same position also hosts the determiner, thus accounting for the complementary
distribution of prenominal genitive and determiner. We abbreviate “speciﬁer of NP”
as SpecNP. As discussed in Chapter 3, the subject NP is the speciﬁer of the sentence,
IP. “Speciﬁer of IP” is abbreviated as SpecIP. Again there is just one SpecIP: there is
one subject per sentence. (5a) shows the position of the speciﬁer in a maximal
projection: the speciﬁer is the constituent immediately dominated by the maximal
projection, XP.7
The speciﬁer combines with the intermediate constituent X′.8
(5b)
is the representation of the structure of an NP, (5c) that of a sentence. The boldfaced
NPs in (5b) and in (5c) occupy the speciﬁer positions.
(5) a XP
X′speciﬁer
complementX
5
Exercise 22 concerns scope relations. However, the exercise is very speculative and it takes us
well beyond the discussion. It should only be tackled at the end of the chapter.
6
Chapter 3, sections 2.1 and 3.3.
7
See the discussion of the format of the structure in Chapter 2, section 2.
8
The component of syntactic theory that sums up the format for syntactic structure is often
referred to as X-bar theory.
TSC04 07/18/2005, 03:39PM243
244 Chapter 4
b NP
N′NP
PPN
of the problemanalysisthe teacher’s
c NP
I′NP
VPI
analyze the problemwillThe teacher
If both NPs and sentences (“IPs”) have a speciﬁer, the question arises whether the
VP also has such a speciﬁer position. In the discussion of the structure of the VP in
Chapter 2,9
we did not have any clear empirical reason for postulating a speciﬁer
for the VP. However, simply stating that the VP does not have a speciﬁer because
our data had not led us to postulate such a position is not sufﬁcient. If the VP did
indeed not have a speciﬁer position at all, we would have to explain why this is. We
would have to account for the asymmetry across categories: NPs have speciﬁers, IPs
have speciﬁers, but VPs would not have them. Indeed, given the need to postulate
speciﬁers in at least two constituents, the simplest theory will be the one in which
no such additional stipulation is needed and in which all constituents, including VP,
have a speciﬁer. We have also seen that the simplest theory is preferred.10
So let us
examine whether the generalization of speciﬁer positions to all maximal projections
is workable. Taking a sentence such as (6a), what could be the ﬁller of the speciﬁer
position of the VP?
(6) a The teacher will [VP probably [VP analyze the problem]].
It is not plausible that the adjoined adverbial probably is the speciﬁer of the VP.
Recall that we assume that a maximal projection has just one speciﬁer. But we have
seen that adjunction is not restricted to just one constituent: a VP may have more
than one adjunct (see (4c)). We also assume that adjoined elements are outside the
core of the projection. The speciﬁer, though, was taken to be part of the core of the
9
Chapter 2, sections 1 and 2.
10
Chapter 1, section 1.2.3.
TSC04 07/18/2005, 03:39PM244
Reﬁning Structures 245
NP and of the clause. Following this line of thinking, we would assume that the
speciﬁer of VP is not an adjunct. Based on the models in (5) for the projection of
NP and IP, we would locate the speciﬁer position in VP lower than left-adjoined
VP-adjuncts, as shown in (6b):
(6) b VP
VPAdvP
V′speciﬁer
the problemanalyzeprobably
NPV
?
Speciﬁer positions typically can host subjects. So we could make a bold move here
and propose that the speciﬁer of the VP is the subject. This move has some interesting
consequences. In the preceding chapter,11
we had come to the conclusion that we
wanted to bring the subject closer to the verb because the subject usually receives a
thematic role from the verb. If thematic relations are symptomatic of there being a
close structural relationship between the assigner of the role (here the verb) and the
receiver of the role (here the subject), then it would seem reasonable to expect the
subject to be located somewhere in the VP. Suppose we merge the subject NP with
V′. This means that the subject of a sentence occupies the speciﬁer position of the
VP, abbreviated as SpecVP. This is done in (7b). How would (7b) be spelt out or
pronounced? Does (7b) provide the basis for the correct linearization of the sentence?
(7) a The teacher will probably analyze the problem.
b VP
VPAdvP
V′speciﬁer
the problemanalyzeprobably
NPV
the teacher
11
Chapter 3, section 3.3.
TSC04 07/18/2005, 03:39PM245
246 Chapter 4
(7b) does not lead to the correct linear order or spell out: (7b) corresponds to the
bracketing representation in (7c), which is spelt out as (7d).
(7) c [IP [I′ will [VP probably [VP the teacher analyze the problem]]]].
d *Will probably the teacher analyze the problem.
(7b/c) cannot be the ultimate structure for the sentence we are aiming for. Compared
with the intended sentence, (7a), the subject NP the teacher occupies the
wrong position.
There is also a theoretical objection to (7b). Recall that we postulate that I,
occupied by the inﬂection or by the auxiliary, serves as a linker between subject and
predicate, and that the content of I validates/qualiﬁes this link.12
In (7c) the I position
would not be able to perform its linking function: there is no constituent in
the speciﬁer of IP that could be linked up with the VP. (7c) is “unbalanced” because
I does not have its anchor.13
2.3 Interpretation
The semantic organization of the structure in (7b) is quite appealing.14
The VP is
hierarchically organized and the complement, which has the closest semantic connection
to V, is situated closer to V than adjoined elements, whose relation to V is looser.
Similarly, as we have seen, adjoined elements themselves are stacked so that the
closer the adjoined element is to the verb the closer its connection to the VP.15
For
instance, in (4c) wholly encodes the degree of understanding; not denies the content
of wholly understand it and still encodes a time span during which ‘not wholly understanding
it’ holds. Wholly is more closely connected to understand it than still.
The subject must be more closely connected to V than any of the adjoined constituents.
After all, in (7a) the core information of the sentence is essentially ‘who
analyzes what’, and the ‘probability’ of this event is in a sense secondary and
additional to that core information. Before qualifying the likelihood of an event, we
want to convey the nature of that event; speciﬁcally we want to identify the entities
involved in the event. So we would expect that the subject is part of the core VP, that
it is closer to V than an adjoined constituent such as probably in (6a). This relative
closeness of the subject to the verb is encoded by representation (7b). The question
is how to reconcile the linearization of the sentence in (7a), with the intuition
expressed by (7b) that the subject is closely related to VP, and with the proposal
that I links the subject and the VP.
12
Exercise 23 introduces sentences in which the predicate seems to precede the subject. However,
this exercise should not be tackled until you have reached the end of the chapter.
13
See also Chapter 3, sections 1.2.1 and 1.2.2.
14
See the discussion in Chapter 3, section 3.3.
15
For illustrations see also Chapter 3, Exercise 3.
TSC04 07/18/2005, 03:39PM246
Reﬁning Structures 247
3 The VP-Internal Subject Hypothesis
3.1 Displaced constituents: Movement and copies
So far, when dealing with integrating the subject into the sentence we have only
used the operation Merge. But recall16
that we also need the operation Move, if only
to allow for the fronting of constituents such as the VPs in our example (1c). Given
that this operation must be independently available in our theory, let us apply it to
the subject and at the same time reﬁne the operation somewhat. We start from (7b),
in which the subject has been merged VP-internally. We motivated this merger on a
theoretical basis, namely by invoking the close semantic connection between subject
and VP. We can now propose that in order to arrive at its position in (7a), the
subject moves leftward and merges again with I′, forming IP. Thus the subject
becomes the speciﬁer of IP as a result of movement:
(8) a IP
I′NP
VPI
the teacherprobably
probablyb The teacher
VPAdvP
will
will
V′NP
NPV
the problem
the problem.
analyze
analyze
Recall that we suggested in Chapter 3 that the subject in SpecIP functions as anchor
for the information in the sentence. The subject satisﬁes the linking function of I: I
validates the link between subject and predicate. The operation Move as proposed
above enables us to achieve the correct linear order and enables the subject to
anchor the sentence. It is important, however, that the movement should not undo
the semantic relation we postulated between the subject the teacher and the verb
analyze. In (8b) the subject NP continues to express the entity that carries out the
16
From Chapter 2, sections 1.4, 2.3.4, and also from Chapter 3, section 2.3.
TSC04 07/18/2005, 03:39PM247
248 Chapter 4
activity. The subject realizes the thematic role agent, to use the technical term.17
We conclude that the subject should somehow be represented twice in the sentence:
it originates in the VP and it occupies the canonical subject position, SpecIP. To
express the double afﬁliation of the subject in the sentence various technical devices
are available.
One approach, the one we will adopt here, is to say that we merge the subject in
the VP and that we move it to the speciﬁer of IP. The result of the movement
operation is that we leave a copy of the moved constituent in its original position.
The original position at which a constituent is merged is also called the base position.
The position that a constituent occupies after movement is called a derived
position. As a result of the Move + Copy operation, the sentence contains two copies
of the subject, one in the SpecVP and one in SpecIP. For reasons of economy,18
when
a sentence contains multiple copies of the same constituent, we pronounce (“spell
out”) only one such copy, in our example the one in the canonical position. We
represent non-pronounced copies of a constituent by strikethrough: [NP the teacher].
(8) c [IP [NP The teacher] [I′ will [VP probably [VP [NP the teacher] analyze [NP the
problem]]]]].19
It is important to underline once more that the motivations for this representation
remain at this point theoretical: (i) the representation allows us to generalize speciﬁers
across categories, thus keeping our theory of structure as simple as possible; (ii) it
allows us to match form and meaning hierarchies in the structure of the VP; (iii) it
allows the subject NP to be related to two positions in the structure. In the next
section we examine whether this proposal can be supported by empirical data.
In the discussion we present the subject as being merged in one position and then
moved upwards, with a copy left in the base position. We pronounce the subject at
its derived position; the copy in the base position remains unpronounced, though it
is still available for interpretation. This representation allows us to relate the subject
NP to two environments in the sentence. A slightly different way of presenting the
same analysis would be to propose that we merge the subject NP twice: once in the
17
For thematic roles see Chapter 3, section 3.
18
In the sense of linguistic economy as discussed in Chapter 1, section 2.2.3.
19
An alternative and older notation is that in which the original position of the moved constituent
is marked with “t” for trace (ia). Coindexation (NPi, ti) is often used to link a trace to
a moved constituent (ib).
(i) a [IP [NP The teacher] [I′ will [VP probably [VP [NP t] analyze [NP the problem]]]]].
b [IP [NPi The teacher] [I′ will [VP probably [VP [NP ti] analyze [NP the problem]]]]].
Instead of strikethrough, one can also use angled brackets < . . . > to surround non-pronounced
copies:
(ii) [IP [NP The teacher] [I′ will [VP probably [VP [NP <the teacher>] analyze [NP the problem]]]]].
TSC04 07/18/2005, 03:39PM248
Reﬁning Structures 249
speciﬁer of VP and once in the speciﬁer of IP. In other words, the sentence could be
argued to simply contain two identical copies of the subject, one merged in the VP
speciﬁer and one “re-merged” in the IP speciﬁer. In the course of the derivation of
the sentence, we ﬁrst merge an NP VP-internally, which allows V to establish a
thematic relation with the NP. (For instance, in (8) the NP the teacher is the agent
of the verb analyze.) Then we insert a copy of that NP higher, in the speciﬁer of
IP, which allows I to accomplish its linking function between subject and VP. For
reasons of economy, when a sentence contains multiple copies of the same constituent,
we pronounce (“spell out”) only one such copy, in our example that in the
subject position. We can again represent non-pronounced copies of a constituent by
strikethrough.
(8) d [IP [NP The teacher] [I′ will [VP probably [VP [NP the teacher] analyze [NP the
problem]]]]].20
Representation (8d) is identical to (8c), except that we do not derive it by movement.
What we do is merge the same NP twice; that is we insert one copy in SpecVP and
one copy in SpecIP. This procedure is sometimes referred to as multiple Merge. The
two proposals are very close, and it would probably be difﬁcult to ﬁnd empirical
differences between them. Still, it is conceivable that one might propose that
the multiple Merge proposal is “simpler” in that it invokes the application of an
operation “Merge” without requiring movement. Both theories need to postulate
copies. In what follows, we will continue to refer to “Move” operations because
this gives us a way to track the sequencing of the various positions of a constituent.
It is plausible though that whatever we say can be replaced by a theory that has
multiple Merge.
A further point that comes out of our analysis is more general. So far we have
been using the term “subject” as if it had a clear and unique reference. However,
having proposed two subject positions in the sentence, the concept “subject” is
now no longer a unitary phenomenon: whereas, before, we had postulated just
one position for the subject in the sentence, now we postulate two positions. The
VP-internal subject position encodes the semantic relation of the subject to the
situation expressed by the VP; in the IP speciﬁer position, the subject NP serves
as the anchoring point for the sentence. The constituent that we label “subject”
has a complex set of properties related to the semantics of the verb and to the
informational structure of the clause. Our representation enables us to identify
the different properties of the subject and associate them with different positions
in the tree.21
20
The “trace” notation illustrated in the preceding footnote appeals to an approach in terms
of movement and copying. Strikethrough and angled brackets are neutral between a Move
approach or an approach in terms of multiple Merge.
21
See McCloskey (1997) for a comprehensive and thorough discussion of the concept of “subject”
in linguistic theory. For discussion of subjects in Japanese and in English see Kuroda
(1986) and Kitagawa (1994).
TSC04 07/18/2005, 03:40PM249
250 Chapter 4
3.2 Empirical support for the VP-internal subject hypothesis?
We have now elaborated a theory in which a sentence contains two positions for the
subject: SpecVP, in which the subject is merged, and SpecIP, into which the subject
moves, leaving a copy in SpecVP. It is important to observe that our theory has not
been made more complex: we need the operations Merge and Move independ-
ently.22
We have simply made use of the same operations to reﬁne the structure. The
elaboration of the structure itself was motivated theoretically.
The question arises whether there are any indications in the language facts
to corroborate the hypothesis that the subject is merged VP-internally. That the
subject is in SpecIP is empirically founded: we literally “see” the subject in that
position. SpecIP is the spell-out position of the subject. But do we have any evidence
that the subject is merged in a lower position, i.e. in SpecVP? In this section we
look at some data that bear on this issue.
3.2.1 FLOATING QUANTIFIERS
Compare the form and the interpretation of the sentences in (9). Identify the subject
of each of the sentences. What is the category of the subject? Discuss how (9a) can
be derived using the Merge and Move technique elaborated above.
(9) a All astronauts don’t speak the same language.
b Astronauts don’t all speak the same language. (Washington Post, 29.4.2003,
p. A10, col. 3)
The two sentences in (9) have a similar interpretation, but their form is different in
that in (9a), all is part of the subject and in (9b), it appears somewhere to the right
of the subject. Still, in (9b), the quantiﬁer all bears on the subject NP, astronauts.
There is what we could call a discontinuous constituent: the subject of (9b) is a
quantiﬁed NP all astronauts, but the quantiﬁer is not adjacent to the NP which
it quantiﬁes over. When a quantiﬁer is not adjacent to the constituent which it
quantiﬁes over it is called a ﬂoating quantiﬁer.23
The constituents in the canonical subject positions in the sentences in (9) are NPs:
their most important element is the N astronauts. The quantiﬁer all is NP-adjoined
in (10a).
(10) a [NP all [NP astronauts]]
b [NP astronauts]
Remember that there is no contradiction in calling astronauts an NP, even though it
contains just one word. The unit astronauts is a constituent whose main (and only)
22
Recall that as an alternative to movement we could appeal to multiple Merge.
23
See also Sportiche (1988) and Koopman and Sportiche (1991). For an alternative view see also
Doetjes (1992).
TSC04 07/18/2005, 03:40PM250
Reﬁning Structures 251
component is a noun. Both the NP all astronauts in (9a) and astronauts in (9b) can
be substituted for by a pronoun:
(9) c They don’t speak the same language.
d They don’t all speak the same language.
Along the lines sketched in the preceding section we derive (9a) as illustrated in (11):
(11) (i) V = speak
(ii) V′
NP
the same language
V
speak
(iii) VP
V′Spec
[NP all [NP astronauts]] speak the same language
(iv) VP
VP
not [NP all [NP astronauts]] speak the same language
(v) I′
VPI
not [NP all [NP astronauts]] speak the same languagedo
(vi) IP
I′NP
do not [NP all [NP astronauts]]
speak the same language
[NP all [NP astronauts]]
TSC04 07/18/2005, 03:40PM251
252 Chapter 4
We assume that don’t results from contracting do and the negative marker not into
one word.
Consider step (vi): we merge the NP [NP all [NP astronauts]] ﬁrst as SpecVP and
then we move it to SpecIP, leaving a copy in SpecVP. This copy will be unpronounced.
How could we derive the alternative pattern in (9b)? Starting from (11v): observe
that in order to derive (11vi), we move the subject NP all astronauts. But astronauts
itself is also an NP, so we could have decided not to move the outer NP but rather
to move the inner NP, the core NP astronauts. As a result, we would leave the outer
layer, containing all, behind in SpecVP. This derivation will yield the pattern in
(9b). The quantiﬁer all remains in the speciﬁer of VP. It is sometimes said that the
quantiﬁer is stranded. The stranded quantiﬁer signals the original merger site of the
subject:
(11) (vi′) IP
I′NP
do not [NP all [NP astronauts]] speak the same language[NP astronauts]
The examples in (12) are based on the test sentence we were working on in the
earlier chapters. Paraphrase them using a pattern with a ﬂoating quantiﬁer.
(12) a All the customers in the corner will order their drinks before the meal.
b All the customers in the corner have ordered their drinks before the meal.
c All the customers in the corner are ordering their drinks before the meal.
The result of ﬂoating the quantiﬁers in (12) is given in (12′):
(12′) a The customers in the corner will all order their drinks before the meal.
b The customers in the corner have all ordered their drinks before the meal.
c The customers in the corner are all ordering their drinks before the meal.
Identify the subject NP and the ﬂoating quantiﬁer associated with it in the following
examples. Describe the relative position of auxiliary and quantiﬁer.
(13) a They’ve all worked so hard. (Ian Rankin, The Falls, 2001: 418)
b They will all give us lower prices, better ranges and more jobs. (Guardian,
1.5.2003, p. 12, col. 1)
c We cannot all drive into a city. (Guardian, 22.7.2002, p. 5, col. 3)
Each of the above examples contains an auxiliary; in each, the linear order is subject
– auxiliary – ﬂoating quantiﬁer – V. This sequencing is as expected given that our
assumption is that the ﬂoating quantiﬁer (abbreviated as FQ) occupies SpecVP.
TSC04 07/18/2005, 03:40PM252
Reﬁning Structures 253
(13′) a They ’ve all worked so hard.
subject FQ
b They will all give us lower prices, better ranges and more jobs.
subject FQ
c We cannot all drive into a city.
subject FQ
We conclude that sentences with ﬂoating quantiﬁers related to their subjects offer
empirical support for the proposal that the subject originates in the speciﬁer of VP.
The ﬂoating quantiﬁer is like a residue in the base position of the subject.24
3.2.2 EXISTENTIAL SENTENCES
3.2.2.1 Thematic subject and grammatical subject
Identify the lexical verb in the following example. What is the subject NP?25
Motivate
your answer.
(14) a Three students are now working on this project.
The subject in (14a) is three students, and the lexical verb is working. When we
form a direct question based on (14a), SAI affects the relative order of the NP three
students and the ﬁnite auxiliary are:26
(14) b Are three students now working on this project?
The NP three students determines the agreement of the inﬂection of the auxiliary:
three students is plural.
Let us insert are in I and assume that the progressive form working is inserted
as one word under V.27
Using the Merge and Move technique, and assuming that
the subject NP originates in the VP, show how we derive sentence (14a). What is
the position of the unpronounced copy of the subject? Following the derivation
outlined above we should proceed as in (15):
(15) a derivation of (14a)
(i) V = working
24
Exercises 3, 5, 8, and 9 are straight applications of the discussion. In Exercises 10, 11, and 12
we raise some problems. Exercise 21 looks at the agreement relation between a quantiﬁer and
the related NP in Hebrew.
25
Exercise 17 of Chapter 3 introduced data similar to the data which we are dealing with here.
26
We turn to the position of the inverted auxiliary in Chapter 5.
27
Cf. Chapter 3, Exercise 14 for reservations concerning this proposal.
TSC04 07/18/2005, 03:40PM253
254 Chapter 4
(ii) Merge V and its complement:
V′
NPV
on this projectworking
(iii) Merge V′ and the subject:
VP
V′NP
working on this project[NP three students]
(iv) Adjoin now to the core VP:
VP
VPAdvP
three students working on this projectnow
(v) Merge VP and I:
I′
VPI
now three students working on this projectare
(vi) Provide anchor for I by merging subject NP and I′:
(15) b Spell out higher copy of subject NP:
[IP[NP Three students] [I′ are [VP now [VP three students working on this
project]]]].
IP
I′NP
now three students working on this project[NP three students]
I VP
are
TSC04 07/18/2005, 03:40PM254
Reﬁning Structures 255
In derivation (15a) we merge the subject NP three students in SpecVP and subsequently
we move it up to SpecIP, leaving a copy in the original position.
Arrived at stage (v) of derivation (15a), could we have proceeded differently?
Suppose we had not moved the subject up to SpecIP: what would happen? We
cannot just leave the speciﬁer of IP empty, because the head I must link two components,
the VP and what will be the speciﬁer of IP, which, among other things,
we designated as the “informational” anchor of the sentence.28
(16) a [IP —— [I′ Are [VP now [VP three students working on this project]]]].
b *Are now three students working on this project.
Can we spell out (16a), but without moving the NP three students into SpecIP? To
maintain the linking function of I, we could try ﬁlling SpecIP with a different constituent.
We insert the element there in SpecIP:
(17) a IP
Spec I′
There
I VP
AdvP VP
NP V′
now three students on this project.are
V PP
working
In (17a) there acts as a ﬁller for the canonical subject position, SpecIP. The NP three
students remains in its base position, SpecVP. (17a) spells out as (17b).
(17) b There are now three students working on this project.
Observe that the auxiliary are is plural. If the VP-internal subject had been singular,
we would have found singular agreement on the auxiliary. In these examples agreement
is not realized in a speciﬁer – head relation.
(17) c There is now one student working on this project.
28
In Chapter 3, section 1.1.
TSC04 07/18/2005, 03:41PM255
256 Chapter 4
Recall that in direct questions the auxiliary typically inverts with the subject. If
there occupies the canonical subject position, then we predict that SAI will invert
the auxiliary with there. This prediction is borne out:
(18) Are there now three students working on this project?
In (17b) two properties normally associated with the unitary subject of a sentence are
split over two constituents. The auxiliary inverts with there in a question (18); the
auxiliary agrees with the NP three students.29
In addition, the NP three students in
(17b) refers to the entity engaged in the activity denoted by the verb working, three
students denotes the agent. There does not denote an entity, it is mere a ﬁller for
SpecIP. Observe that the word there as used here is not the place indication used in
opposition to here: to show this we may note that locative there can be added to
this sentence without creating redundancy; here can also be added without creating
a contradiction:
(19) a There, there are now three students working on this project.
b Here, there are now three students working on this project.
Apparently, there is not inserted in the subject position in (17b) to denote a place but
rather because the head I is a linking element which serves to anchor the sentence.
In the absence of a referential anchoring NP, we insert there. There is sometimes
called a dummy element. Its role is to stand in for the subject. Another term is
expletive there.30
We conclude that if we do not move the NP three students into SpecIP, we can
merge there in SpecIP. As a result, subjecthood is split over two constituents, and
over two positions. The NP three students denotes the entity that carries out the
action expressed by the verb working. In this way, we may say that it is the subject.
The NP three students also determines the agreement of the inﬂectional element, the
head of the sentence. On the other hand, with respect to grammatical operations
such as SAI, there functions as the subject. Sometimes it is said that three students is
the notional subject or the logical subject and that there is the grammatical subject,
the provisional subject or the expletive subject. Using terminology elaborated in
Chapter 331
we can add that three students is the thematic subject.
We may wonder whether there-sentences are associated with a speciﬁc interpretive
effect. One effect of inserting there in the canonical subject position of the sentence
is that the thematic subject does not move to the canonical subject position. We
have postulated that the subject NP is inserted in SpecVP. This is its base position.
29
For some discussion of agreement patterns in there sentences see also Sobin (1997) and Schütze
(1999). For a different pattern see also the examples in (6) of Exercise 2.
30
Exercise 2.
31
Section 3.
TSC04 07/18/2005, 03:41PM256
Reﬁning Structures 257
In sentences introduced by expletive there, the subject remains in its base position.32
Hierarchically, the thematic subject remains “lower” in the structure. Linearly, the
thematic subject remains in a position to the right of the canonical subject position.
In a there sentence, the thematic subject is expressed later in the sentence than if it
were to occupy the canonical subject position; when we hear or read such sentences
the thematic subject comes later.
In general when communicating information we tend to organize that information
in a particular way. We start on the basis of familiar information and we lead up
to new information.33
By locating a constituent toward the end of the sentence we
signal it is relatively new information. In (17b) we present as new the information
that three students work on the project. New information may also concern the
very existence of an entity whose existence we did not know about, an entity that
has recently come into existence, or at least of whose existence we have only recently
been made aware. In (17b) we draw attention to the existence of the students
and their participation in the action. Sentences with expletive there in the subject
position and with a thematic subject in SpecVP are sometimes called existential
sentences.
3.2.2.2 A note on determiner choice
Describe the difference in the realization of the subject NPs in the following sentences.
When could we use such sentences? How could we describe the differences in the
realizations of the subjects?
(20) a A French student has arrived.
b The French student has arrived.
In (20a) the subject NP a French student is introduced by an indeﬁnite determiner,
the indeﬁnite article a. In (20b) it is introduced by a deﬁnite article (the). The choice
of the determiner is associated with a difference in meaning. In (20a) we signal that
the entity denoted by the NP (‘a French student’) is new to the cognitive environment;
we have not discussed this referent yet; we introduce the entity to be discussed
by means of the sentence. In (20b) the entity denoted by the subject is presented as
being already accessible somewhere in our cognitive environment: perhaps we have
already mentioned the French student in the current discussion or perhaps we have
a certain expectation given our knowledge of the situation. In English (as in fact in
many languages), the deﬁnite article and the indeﬁnite article are linguistic devices
to indicate the discourse status of the entity referred to. When we wish to introduce
a novel entity into the discourse we use the indeﬁnite article for a singular NP. For
plural NPs, we either do not ﬁnd an article at all or we may use the quantiﬁer some
(unstressed and pronounced as [sm]).
32
For a complication see Exercise 14.
33
For instance, in this chapter we ﬁrst recapitulate what we know already and then move on to
new material. See Prince (1981).
TSC04 07/18/2005, 03:41PM257
258 Chapter 4
(20) c French students have arrived.
Some French students have arrived.
If the existential sentence pattern serves to introduce the subject as novel then we
predict that in existential patterns, subjects will typically be indeﬁnite. If you go
over the examples discussed so far once again, you will see that this prediction is
borne out.34
3.2.2.3 A note on verb choices and existential patterns
In the derivation of the English existential sentences above we saw that we can leave
a subject in its base position and insert there in the canonical subject position.
However, the account might wrongly lead us to expect that we can simply leave
every subject in the base position and insert the expletive there. We noted in the
preceding section that the application of this strategy is restricted to sentences with
indeﬁnite subjects. If we were to assume that the existential strategy generalizes to
all English sentences, however, we would wildly overgenerate: we would produce
lots of unacceptable sentences. Consider (21). Using the step-by-step derivation
discussed above to derive (17b), construct the sentences that would arise if we (i)
left the subject in the base position, and (ii) merged there in the canonical subject
position.35
(21) a Three students have worked on this project.
b Three students worked on this project.
In (22) we outline the derivation of (21a):
(22) a Derivation of (21a).
(i) V = worked
(ii) Merge V and its complement:
V′
PPV
on this projectworked
34
That subjects of existential sentences are usually indeﬁnite is referred to as the deﬁniteness
effect. For discussion of the existential pattern see also Milsark (1974, 1979), Saﬁr (1985),
Lumsden (1988), Belletti (1988), Law (1999).
35
You may insert inﬂected have under I and the past participle worked under V.
TSC04 07/18/2005, 03:41PM258
Reﬁning Structures 259
(iii) Merge V′ and the subject NP:
VP
V′NP
worked on this project[NP three students]
(iv) Merge VP and I (see note 35):
I′
VPI
three students worked on this projecthave
(v) Provide anchor for I by merging subject NP and I′:
IP
I′NP
three students worked on this project[NP three students]
VPI
have
b Spell out higher copy of subject NP:
[IP Three students [I′ have [VP three students worked on this project]]].
We have merged the subject NP three students in SpecVP and moved it up to SpecIP,
leaving a lower (non-pronounced) copy.
Arrived at step (iii) of (22a), could we have proceeded differently? Suppose we
had decided not to move the subject from SpecVP to SpecIP. As before, we cannot
just leave the speciﬁer of IP empty, because the head I must link two components,
the VP and the speciﬁer of IP, which serves as the anchor of the sentence. To spell
out this sentence without moving the NP three students, we might once again try
ﬁlling SpecIP with existential there:
TSC04 07/18/2005, 03:42PM259
260 Chapter 4
(23) a IP
I′Spec
three students worked on this projectThere
VPI
have
In (23a) there ﬁlls the position SpecIP, the NP three students remains in SpecVP.
(23a) would spell out as an unacceptable sentence (23b):
(23) b *There have three students worked on this project.
Though we have faithfully followed the steps of derivation (17a), the result is not
acceptable. The same problem will arise if we try to create existential sentences on
the basis of the second example in (21).
(23) c *There three students worked on this project.
This is a problem. It could mean that we need to reconsider our analysis of existential
sentences entirely, and discard derivation (17). However, before throwing
out the analysis (and losing the insights it had given us), let us look a little further
aﬁeld. In Belfast English, sentences resembling (23b) are attested, though not those
corresponding to (23c).36
(24) Belfast English (Henry, 2001):
a % There have hundreds of people phoned us.
b % There has something come in about this.
but
c % *There hundreds of people phoned us.
d % *There something came in about this.37
Belfast English allows the derivation of existential sentences with the auxiliary have,
though it does not allow the pattern in the absence of an auxiliary. (24e) and (24f)
give a partial representation of (24a) and (24b):
(24) e % [IP There [I′ have [VP hundreds of people phoned us]]].
f % [IP There [I′ has [VP something come in about this]]].
36
The percentage symbol, %, indicates that only a subsection of speakers accept the example.
When a percentage symbol is followed by an asterisk * this means that even for the relevant
subsection of speakers the sentence is ungrammatical.
37
See also Henry (2001) and Cottell and Henry (2004).
TSC04 07/18/2005, 03:42PM260
Reﬁning Structures 261
In other languages, existential sentences can be fairly freely derived even in the absence
of an auxiliary. Icelandic, for instance, allows sentences such as the following:38
(25) a ad qad hefur einhver bordad epli.
that there has someone eaten (an) apple (Vikner, 1995: 189)
‘that someone has eaten an apple.’
b ad qad bordadi einhver epli.
that there ate someone (an) apple (Vikner, 1995: 219)
‘that someone ate an apple.’
In (25a) the canonical subject position is occupied by expletive QaD, and the agent
of the action, the indeﬁnite NP einhver (‘someone’), follows the ﬁnite auxiliary
hefur (‘have’). Using our structures elaborated so far the sentence would have the
partial structure in (25c):
(25) c ad [IP qad [I′ hefur [VP einhver bordad epli]]].39
(25b) lacks an auxiliary but the existential pattern is possible. The ﬁnite lexical verb
borDaDi precedes the subject. Recall that depending on the strength of the inﬂection
the lexical verb will either remain in the position V and I will lower onto V, or,
alternatively, the lexical verb will raise to ﬁnite I.40
If the subject einhver (‘someone’)
occupies SpecVP then the position of the verb to its left suggests that V moves to I
in Icelandic.
(25) d ad [IP qad [I′ bordadi [VP einhver bordadi epli]]].
This would mean that the inﬂectional paradigm of Icelandic is strong, that is, that it
contains many distinct forms. Table 1 contains the paradigms for the present tense
and for the past tense of the verb hear.41
For some reason, which we won’t explore
here,42
Modern English does not allow the general application of the existential
pattern. If you go over all the acceptable sentences in the preceding text you will
ﬁnd that they have in common the presence of be.43
We assume then that there must
be a special property in English that will mainly restrict the existential pattern to
sentences with be.44
38
Examples such as (25a) and (25b) in which a transitive sentence is used in the existential
pattern are referred to as transitive expletive constructions. See Exercise 14 of this chapter for
complications.
39
For the position of the subordinating conjunction aD (‘that’) see Chapter 5, section 2.3.
40
See Chapter 3, sections 1.2.3.3 and 1.2.4. See also Chapter 3, Exercise 6.
41
See Chapter 3, Exercise 6.
42
See Bobaljik and Jonas (1996), Vikner (1995).
43
This is a simpliﬁcation. See Milsark (1974, 1979), Saﬁr (1985), Belletti (1988), and for introductory
discussion also Haegeman and Guéron (1999: chapter 2, section 3).
44
Exercises 6, 7, 13, and 16 offer additional empirical support for the hypothesis that the subject
originates in the VP. Exercises 14, 15, and 16 raise complications.
TSC04 07/18/2005, 03:42PM261
262 Chapter 4
4 Subject Positions and Auxiliaries
4.1 Auxiliaries and step-by-step movement of the subject
When discussing the structural properties of sentences containing auxiliaries in
Chapter 3,45
we had not yet introduced the VP-internal subject hypothesis. In
order to ensure that our theory remains internally coherent, we need to make sure
that our different proposals for the structure of the sentence are consistent with
each other. Recall from Chapter 1 that the knowledge acquired by scientists is
cumulative and is part of a system. We cannot simply present unrelated insights,
however interesting they may be. Let us return to the hypothesis that the subject
originates in the speciﬁer of VP and that it moves up to the canonical position, the
speciﬁer of IP, leaving a copy in the original position. Let us examine how this
hypothesis ties in with the representations that we elaborated for sentences with
multiple auxiliaries.
The auxiliaries have and be are verbs which head their own projection, VP.
Following our theory about structure, these VPs can also have a speciﬁer position.
Consider a sentence containing an auxiliary that takes as its complement a VP
headed by a lexical verb. The lexical verb assigns a thematic role to the subject; the
auxiliary does not assign any thematic role.46
By hypothesis, though, the VP headed
by the auxiliary can have a speciﬁer position.
Let us examine the movement of the subject from its VP-internal base position to
the canonical subject position, SpecIP. What happens if the sentence contains more
than one auxiliary? Sentences relevant for the discussion are (26a) and (26b).47
Table 1 Inﬂectional patterns in Icelandic
PRESENT PAST
1sg heyri heyrDi
2sg heyrir heyrDir
3sg heyrir heyrDi
1pl heyrum heyrDum
2pl heyriD heyrDuD
3pl heyra heyrDu
45
Section 4.
46
See discussion in Chapter 3, sections 3.3 and 3.4.
47
These sentences are discussed as (68) and (69) in Chapter 3, section 4.4.
TSC04 07/18/2005, 03:42PM262
Reﬁning Structures 263
(26) a These students would have considered other universities.
b We had been talking about this problem.
The question arises whether, starting from the speciﬁer position of the lexical VP,
the subject moves directly into SpecIP, or whether it transits via the intermediate
speciﬁer of the VP headed by the auxiliary. Consider the representations in (27),
which we elaborated in Chapter 3.48
(27) a IP
I′NP
These students
I VP
V VP
would have
V NP
other universitiesconsidered
b IP
I′NP
We
I VP
V VP
-ed have
V VP
about this problemtalking
V PP
been
Revise representations (27a) and (27b), signaling the copy of the subject in its base
position using the strikethrough notation. Also add in the speciﬁer positions for the
projections headed by the auxiliaries. Once you have done this, check the result
with the representations below.
48
(27a) and (27b) correspond to representations (68c) and (69b) in Chapter 3.
TSC04 07/18/2005, 03:42PM263
264 Chapter 4
(28) a IP
I′NP
These students
I VP
Spec V′
would
V VP
the students
NP V′
have other universities
V NP
considered
b IP
I′NP
We
I VP
Spec V′
-ed
V VP
Spec V′
have
V VP
been talking about this
NP V′
we
TSC04 07/18/2005, 03:42PM264
Reﬁning Structures 265
How does the subject NP move from the lower SpecVP to SpecIP? Does it move in
one step? Or does it move step by step via the intermediate speciﬁers? In the latter case,
we could assume that the subject leaves an unpronounced copy in the intermediate
speciﬁers. This would mean that in a sense, the subject NP is the subject of the
lexical VP but also of the VPs headed by the auxiliaries.
4.2 Empirical support for step-by-step movement
We cannot decide on the precise execution of the movement of the subject by simply
looking at the tree diagram or at a random sentence. The relevant intermediate
speciﬁer positions, which are located between the auxiliaries and which would
correspond to the speciﬁers of the VPs headed by auxiliaries, do not contain any
overt elements, but this is also expected if the subject moves through them because
the intermediate copies left by the subject would not be spelt out.
(29) a [IP These students would [VP these students have [VP these students considered
other universities]]].
= These students would have considered other universities.
b [IP We [I -ed] [VP we have [VP we been [VP we talking about this]]]].
= We had been talking about this.
What kind of evidence could we invoke in support of the step-by-step movement
hypothesis? The evidence we are looking for has to show that in addition to SpecIP
and SpecVP there are intermediate subject positions that contain unpronounced
copies of the moved subject. Earlier sections in this chapter introduced empirical
material to deal precisely with the issue of moved constituents and their residues.
We postulated two subject positions in the sentence: SpecVP and SpecIP. The
empirical support for the VP-internal subject position, SpecVP, was drawn from,
among other things, the distribution of ﬂoating quantiﬁers (section 3.2.1). In this
light, consider sentence (30a):49
(30) a All the students must have been sleeping.
In this example, the quantiﬁer all has moved with the subject. Try stranding
the quantiﬁer in a lower position. McCawley (1988: 90) gives the following
examples:50
(30) b The students must all have been sleeping.
c The students must have all been sleeping.
d The students must have been all sleeping.
49
These data are from McCawley (1988: 92).
50
Exercise 1.
TSC04 07/18/2005, 03:43PM265
266 Chapter 4
In these examples, the quantiﬁer all is stranded in intermediate positions. The positions
occupied by the quantiﬁer correspond exactly to the intermediate speciﬁer
position that we postulated above.51
(30) b′ [IP The students must [VP all the students have [VP all the students been [VP
all the students sleeping]]]].
c′ [IP The students must [VP the students have [VP all the students been [VP all
the students sleeping]]]].
d′ [IP The students must [VP the students have [VP the students been [VP all the
students sleeping]]]].
The following attested examples illustrate the stranding of a quantiﬁer in various
positions:52
(31) a They would have all loved to come. (Mick Jagger, interview, BBC4,
12.12.2003, 7 o’clock news)
b They’d all have gladly murdered Brooks. (Colin Dexter, The Daughters of
Cain, 1995: 266).
c Those leisured or ﬂexi-working people who have the chance to go to the
cinemas in the afternoon must all have wondered at some time if multiplexes
are a front for something else. (Guardian, 24.5.2003, p. 17, col. 7)
5 Summary
This chapter reassesses the representation of the structure of the sentence in terms
of the question of how well the structure we had elaborated so far can be mapped
onto its interpretation. In English, SpecIP is the canonical subject position; subjects
typically occupy that position. However, there are a number of theoretical reasons
for assuming that we must also provide for a VP-internal subject position. These
theoretical arguments are twofold. Firstly, there is an argument to be made for
generalizing the speciﬁer position to all maximal projections, thus eliminating the
exceptional status of the VP (which lacks a speciﬁer) as compared to NP and IP
(which have a speciﬁer). Secondly, there is a semantic argument that comes out of
our attempt to have structure match meaning. In particular, if the subject were
really located entirely outside the VP, it would come as a surprise that it can receive
a thematic role from the verb, while VP-adjoined constituents, which are structurally
closer to V, do not receive a thematic role.
In this chapter we elaborate the VP-internal subject hypothesis. We propose that
the subject is ﬁrst merged with V′. It is the speciﬁer of VP. Then, when I′ has been
51
Exercise 4.
52
Exercises 17, 18, and 19 examine the structure of passive sentences.
TSC04 07/18/2005, 03:43PM266
Reﬁning Structures 267
built, the subject leaves the speciﬁer of VP and moves to merge with I′, thus becoming
the speciﬁer of IP. The subject has a double afﬁliation in the sentence: it is both
the subject of the sentence and that of the VP.53
The chapter provides empirical support for the hypothesis that the subject originates
in the VP. That support comes from the distribution of ﬂoating quantiﬁers
associated with the subject and from the distribution of the subject NP in existential
sentences. Floating quantiﬁers associated with the subject can be seen to occupy a
position adjacent to the lexical verb. We propose that such ﬂoating quantiﬁers are
stranded in the base position of the subject, SpecVP. In existential sentences introduced
by expletive there, the logical subject of the sentence is adjacent to the verb.
Again we assume that it occupies its base position, SpecVP.54
If auxiliaries head independent VPs, their projections will also have a speciﬁer
position. Auxiliaries do not assign a thematic role, so there is no obvious ﬁller for
their speciﬁer position. We assume that on its way from the speciﬁer position of the
lexical VP to the canonical subject position, SpecIP, the subject, which receives a
thematic role from the lexical verb, moves through the intermediate speciﬁer positions
of the projections headed by the auxiliaries. Evidence for this step-by-step movement
is provided by the fact that quantiﬁers associated with the subject may be found in
the intermediate speciﬁer positions.
53
Exercise 23 introduces an additional word order pattern.
54
For a general discussion of the traditional terms “subject” and “object” in relation to sentence
structure see also the introduction and the papers in Davies and Dubinsky (2001b).
TSC04 07/18/2005, 03:43PM267
268 Chapter 4
Exercises
Contents
Exercise 1 Floating quantiﬁers (T) 269
Exercise 2 Existential patterns (T, E) 269
Exercise 3 Floating quantiﬁers and to inﬁnitives (T) 270
Exercise 4 Existential patterns in Belfast English (T) 270
Exercise 5 Floating quantiﬁers (T) 271
Exercise 6 Verb-Subject-Object (VSO) languages (T, E) 271
Exercise 7 VP ellipsis and Verb-Subject-Object (VSO) languages
(E, presupposes Exercise 6) 275
Exercise 8 Verb positions, ﬂoating quantiﬁers, and comparative
syntax (T) 277
Exercise 9 Floating quantiﬁers and adjuncts (T, E) 278
Exercise 10 Floating quantiﬁers and manner adjuncts (T, E) 279
Exercise 11 Floating quantiﬁers and adjuncts in French (T, E) 280
Exercise 12 Floating quantiﬁers and adjuncts in Hebrew (T, E) 281
Exercise 13 VSO patterns in Welsh (E) 283
Exercise 14 Subject position(s) in Icelandic transitive expletive
constructions (T, E) 284
Exercise 15 VSO orders and subject positions (E, presupposes
Exercises 6, 7, and 13) 285
Exercise 16 VSO orders and subjects in Modern Greek (E, presupposes
Exercises 7, 13, 14, and 15) 287
Exercise 17 Passivization and ﬂoating quantiﬁers (E) 288
Exercise 18 Passivization and ﬂoating quantiﬁers (E, presupposes
Exercise 17) 292
Exercise 19 Floating quantiﬁers and the base position of the subject
in passive sentences (T, presupposes Exercise 17) 292
Exercise 20 Passivization and existential patterns (E, presupposes
Exercises 17, 18, and 19) 293
Exercise 21 Floating quantiﬁers in Hebrew (E) 295
Exercise 22 Reﬁning structures: From one functional head to many
(T, E) 296
Exercise 23 Predicate inversion in English (E) 301
Exercise 24 Language typology (T, E, presupposes Exercises 6, 7, 13,
and 16) 303
TSC04 07/18/2005, 03:43PM268
Reﬁning Structures 269
Exercise 1 Floating quantiﬁers (T)
Consider the distribution of the ﬂoating quantiﬁers in the following examples (from
Cottell and Henry, 2004). Does the discussion in the chapter allow us to derive the
position of the quantiﬁers?
(1) The students should have been all doing the exam at that time.
(2) The students should have all been doing the exam at that time.
(3) The students should all have been doing the exam at that time.
Exercise 2 Existential patterns (T, E)
Identify the existential sentences in the following examples. Discuss the derivation
of the existential sentences.
(1) There was nothing happening and the market just drifted. No business was
going through. (Guardian, 22.6.2002, p. 4, col. 3)
(2) William Ostrom admits there is a “bomb” waiting to go off in consumers’
pockets. (Guardian, 6.5.2003, p. 14, col. 4)
(3) There are probably fewer than a dozen major agencies in North America
handling bookings for language schools world-wide. (New York Times,
1.8.2004, Travel section, p. 4, col. 4)
(4) When I came out there were men selling special editions of the evening paper.
(Guardian, Review, 31.5.2003, p. 5, cols 3–4)
(5) If you are staying with a family there are a lot of things going on around you.
(New York Times, 1.8.2004, Travel section, p. 4, col. 5)
Comment on the agreement patterns of be in the following examples:1
(6) a There’s no permanent jobs going, are there? (Josie Lloyd and Emlyn Rees,
Come Together, 1999: 164)
1
For some discussion of agreement patterns in there sentences see also Sobin (1997) and
Schütze (1999). See also Exercise 6 for comparative data on subject verb agreement.
Exercises 269
TSC04 07/18/2005, 03:43PM269
270 Chapter 4
b A lot more people are going out, there’s been a load of new clubs opening
and the music scene has really come together. (Observer Magazine,
21.11.2004, p. 35, col. 3)
Exercise 3 Floating quantiﬁers and to inﬁnitives (T)
In Chapter 32
we brieﬂy examined the structure of inﬁnitival clauses. In which
position did we insert the inﬁnitival marker to? Assuming that inﬁnitival clauses also
allow quantiﬁer ﬂoating, what should be the relative order of to and the ﬂoating
quantiﬁer? Consider the underlined sequences in (1) and (2). Is the distribution of
the ﬂoating quantiﬁer predicted by the analysis of to elaborated in Chapter 3?
(1) It is not exceptional for experts to disagree among themselves. In fact, it would
be exceptional if they were to all agree. (Guardian, 17.3.2004, p. 3, col. 4)
(2) We believe it is crucial for communities to each have their own beat ofﬁcer.
(Independent, Review, 30.7.2004, p. 2, col. 2)
Exercise 4 Existential patterns in Belfast English (T)
Recall that in Belfast English the existential pattern generalizes to all types of sentences
with auxiliaries. In particular, transitive sentences with the auxiliary have also allow
the pattern:3
(1) a Some students should have passed the tests.
b %There should have some students passed the tests. (Cottell and Henry,
2004 (4))
Draw a tree diagram for (1b). Use the derivation in (15) in the text as your model
and remember that auxiliaries head independent V-projections.
Consider example (2). Discuss how it provides empirical support for the step-bystep
movement of the subject:
(2) %There should some students have passed the tests. (Cottell and Henry,
2004 (4))
2
Section 1.3.
3
Section 3.2.2.3, example (24).
TSC04 07/18/2005, 03:43PM270
Reﬁning Structures 271
Exercise 5 Floating quantiﬁers (T)
Discuss the distribution of the underlined quantiﬁers in the following examples.
How would we derive their positions? Do these examples give rise to any problems?
(1) We all want to take part but we don’t necessarily all know how and where.
(BBC radio 4, phone-in listener, 29.10.2002, 13.45)
(2) Four of the ﬁve – Andrew Acred, Richard Blues, James Munk and James
Spooner – all celebrated as they picked up their results from the school yesterday.
(Independent, 23.8.2002, p. 9, col. 1)
(3) But it has not all been easy. (Guardian, G2, 22.11.2002, p. 10, col. 3)
(4) We cannot all drive into a city. (Guardian, 22.7.2002, p. 5, col. 3)
(5) There is certainly a perception that it doesn’t all ﬁlter through. (Guardian, G2,
2.10.2001, p. 13, col. 1)
Exercise 6 Verb-Subject-Object (VSO) languages (T, E)
Consider the following example, taken from Shlonsky (1997: 70). It illustrates the
neutral word order of Standard Arabic:
(1) Katab-at Mona risaalat-an.
wrote-3FS Mona a letter
‘Mona wrote a letter.’
As you can see, Standard Arabic displays VSO word order: the lexical verb katabat
(‘wrote’) precedes the subject Mona. The canonical subject position is here postverbal.
Assuming that an Arabic sentence is assembled in a way similar to an English
sentence, how could we derive (1)?
In Standard Arabic, VSO patterns alternate with SVO patterns, as illustrated in (2)
and in (3). As the examples show, the alternation in word order is not free. Examine
the glosses of the examples carefully. Discuss the correlation between verb morphology
and its relative distribution with respect to the subject. (Examples from Ouhalla
& Shlonsky, 2002: 13.)
(2) a Katab-a l-}awlaad-u l-risaalat-a.
write(PERF)-3MS the boys-NOM the letter-ACC
‘The boys wrote the letter.’
Exercises 271
TSC04 07/18/2005, 03:43PM271
272 Chapter 4
b *L-}awlaad-u katab-a l-risaalat-a.
the boys-NOM wrote(PERF)-3MS the letter-ACC
(3) a L-}awlaad-u katab-uu l-risaalat-a.
the boys-NOM wrote(PERF)-3MPL the letter-ACC
‘The boys wrote the letter.’
b *Katab-uu l-}awlaad-u l-risaalat-a.
wrote(PERF)-3MPL the boys-NOM the letter-ACC
KEY AND COMMENTS
For reasons of economy, the simplest theory is one according to which, crosslinguistically,
sentences are assembled in the same way, whether they be Arabic or
English. If we assume that the Arabic sentence is derived in the same way that the
English sentence is derived, then we start with the merger of the components of the
VP. V (katab ‘write’) ﬁrst combines with its complement (risaalat-an, ‘a letter’), and
then with its subject (Mona). Then the VP is merged with I. In order to derive the
VSO order we can propose that the verb moves to I, while the subject remains in
the speciﬁer of the VP. As before, to preserve the semantic contribution of V in the
VP and to preserve the structure of the sentence we use strikethrough to indicate
the unpronounced copy of the verb.
(1) b IP
I′
I VP
NP V′
risaalat-ankatab-at
V NP
katabMona
In (2) and (3) the subject is plural: l-]awlaad-u (‘the boys’). In (2) the lexical verb
does not agree in number with the subject: katab-a is singular and the subject
l-]awlaad-u (‘the boys’) is plural. In this example, the lexical verb moves out of VP
to I, but the subject remains in the speciﬁer of the VP. We obtain the VSO order.4
4
One point that remains unclear in this representation is why the speciﬁer position of I is not
ﬁlled. It could be the case that perhaps I is not universally a “linker” and that in some
languages it does not require a speciﬁer (see Goodall, 2001; McCloskey, 2001).
TSC04 07/18/2005, 03:43PM272
Reﬁning Structures 273
To preserve the semantic contribution of V in the VP and to preserve the structure
of the sentence we use strikethrough to indicate the unpronounced copy of the
verb.
(2) c [IP [I katab-a][VP l-}awlaad-u katab- l-risaalat-a]].
d IP
I′
I VP
NP V′
katab-Katab-a
V NP
l-}awlaad-u l-risaalat-a
In (3) the lexical verb (katab-uu) is plural: it agrees in number with the subject
(l-]awlaad-u, ‘the boys’). In this case, the subject itself also has to leave the VP.
We assume that it moves to the speciﬁer of IP.
(3) c [IP l-}awlaad-u [I katab-uu] [VP l-?awlaad-u katab l-risaalat-a]].
d IP
I′
I VP
NP V′
l-risaalat-akatab-uuV
NP
l-}awlaad-u
NP
l-}awlaad-u katab
Recall that we had been proposing that the speciﬁer – head relation in IP is the
typical conﬁguration for agreement. We can conclude from the data in (2) and (3)
that for the verb in I to agree with the subject in Standard Arabic, the subject must
occupy SpecIP. Or, putting it differently, moving the subject into the speciﬁer of IP
Exercises 273
TSC04 07/18/2005, 03:43PM273
274 Chapter 4
triggers agreement on the verb in I. The patterns in (2) and (3) conﬁrm our
hypothesis that agreement can be realized in a speciﬁer – head relation.
Consider (4) and (5) from Moroccan Arabic (Ouhalla and Shlonsky, 2002: 13). Compare
these data with the Standard Arabic data. What would you conclude concerning
the correlation between subject agreement and subject movement in this language?
(4) a Ketb-u l-wlad l-bra.
write(PERF)-3PL the children(PL) the letter
‘The children have written the letter.’
b *Ketb- l-wlad l-bra.
write(PERF)-3MSG the children(PL) the letter
(5) a L-wlad ketb-u l-bra.
the children(PL) write(PERF)-3PL the letter
‘The children have written the letter.’
b *L-wlad ketb- l-bra.
the children(PL) write(PERF)-3MSG the letter
KEY AND COMMENTS
In Standard Arabic there is a correlation between movement of the subject to
SpecIP and the realization of agreement morphology on the verb. For a verb to
agree in number with its subject, the subject must move to speciﬁer position of I.
This pattern cannot be generalized to Moroccan Arabic. In Moroccan Arabic, the
verb always agrees for number with the subject: with a plural subject l-wlad
(‘the children’), only plural kEtb-u (‘write’) is possible. Subject movement to the
speciﬁer of IP is, however, not obligatory.
To allow for number agreement in the VSO pattern in (4a) we have to conclude
that in some cases agreement morphology in I can also be realized in a matching
relation between a head, I, and a lower NP. In our example the relevant head is I
and the relevant NP occupies the speciﬁer of the VP. We conclude that agreement
does not always require a speciﬁer – head relation.
(4) d [IP [I Ketb-u] [VP l-wlad ketb l-bra]].
write(PERF)-3PL the children(PL) write the letter
We came across a similar agreement relation between a higher inﬂectional element
and a lower NP in our text examples (17b) and (17c), repeated here in (6). Could
we say that the higher inﬂected element and the subject NP must be adjacent?
(6) a There are now three students working on this project.
b There is now one student working on this project.
TSC04 07/18/2005, 03:43PM274
Reﬁning Structures 275
For proposals concerning the relation between agreement and movement see
Chomsky (1995).
Exercise 7 VP ellipsis and Verb-Subject-Object (VSO)
languages (E, presupposes Exercise 6)
Recall that the English VP can sometimes be omitted. Indicate VP ellipsis in the
following examples by means of the symbol ∅. Reconstruct the deleted material.
(1) Only those who were in the room know the absolute truth of this story. No
one else probably ever will. (Washington Post, 25.3.2004, p. D3, col. 5)
(2) If we could charge more money, we would. (Wall Street Journal, 29.3.2004,
p. A6, col. 6)
(3) Everyone says you can’t be scientiﬁc and fun, but we think you can. (New
York Times, 8.3.2004, p. C5, col. 2)
VP ellipsis is typically used in answers to yes/no questions (4a). If the context
supplies sufﬁcient information we may even omit the VP both in the question and
in the reply (4b).
(4) a Speaker A: Would you charge more money?
Speaker B: We would [VP ∅].
b Speaker A: Would you [VP ∅]?
Speaker B: We would [VP ∅].
In (4b) both the lexical verb and its object have been omitted in the question and
in the answer. Let us also represent VP ellipsis by means of strikethrough:
(4) c IP
I′NP
we charge more moneyWe
VPI
would
Consider the Irish examples in (5) (McCloskey, 2001: 161 his (4a), (5)). Using the
English glosses provided, describe the position of the lexical verb in relation to the
subject and the object.
Exercises 275
TSC04 07/18/2005, 03:44PM275
276 Chapter 4
(5) a Thóg sí teach dófa ar an Mhullach Dubh.
raised she house for them on the Mullaghduff
‘She built a house for them in Mullaghduff.’
b D’ól sí deaoch uische.
PAST drink she drink water
‘She drank a drink of water.’
Using the discussion of Arabic in Exercise 6 as a model, how could we derive the
VSO pattern in Irish?
Discuss ellipsis in the Irish question answer pair (6b, c). How does the pattern in
(6) differ from VP ellipsis in English (4)? Can we account for this difference?
(6) a Thóg sí teach.
raised she house
‘She built a house.’
b A-r thóg?
INTERROG-PAST raised
‘Did she?’
c Creidim gu-r thóg.
believe-1SG that-PAST raised.
‘I believe she did.’
KEY AND COMMENTS
In English VP ellipsis deletes the content of VP. That is it affects V′, the unpronounced
copy of the subject and VP adjuncts. VP ellipsis obviously leaves the subject in
SpecIP intact. In the Irish examples of ellipsis in (6), what seems to correspond to
English VP ellipsis paradoxically does not affect the verb while it does affect the
subject. In (6b) and in (6c) the verb thóg (‘raised’) is maintained; the subject
pronoun sí (‘she’) has been omitted. Let us examine how we can account for this
difference between English and Irish.
Irish is a VSO language. The canonical subject position is postverbal. If we try
to derive the VSO order in the way suggested in Exercise 6 for Arabic, this means
that in Irish the lexical verb evacuates the VP; it moves to I. On the other hand,
the subject remains in the speciﬁer of VP. The representation in (7) is based on
representation (1b) for the Arabic example (1a):5
5
As before, the question arises why the speciﬁer position of I apparently need not be ﬁlled. It
could be the case that I is not universally a “linker” and that in some languages it does not
require a speciﬁer (see Goodall, 2001; McCloskey, 2001).
TSC04 07/18/2005, 03:44PM276
Reﬁning Structures 277
(7) IP
I′
I VP
NP V′
teach
house
Thóg
built
V NP
thóg
build
sí
she
Applied to (7), “VP ellipsis” will delete the content of VP. As the lexical verb thóg
(‘built’) has “escaped” to I, VP deletion will not affect it. On the other hand, the
subject sí (‘she’) remains in VP, there is no movement to SpecIP. Thus the subject
will be affected by the ellipsis.6
Exercise 8 Verb positions, ﬂoating quantiﬁers, and
comparative syntax (T)
In this chapter we assume that ﬂoating quantiﬁers are stranded in the speciﬁer of
the VP. We saw in Chapter 3, section 1.2.4.2, that languages vary with respect
to the distribution of the lexical verb. In English the lexical verb remains in VP; in
French it moves to I. Using these theoretical assumptions, which would you predict
to be the acceptable word order in English: (1a) or (1b)? Is your prediction correct?
(1) a The students prepare all the text.
b The students all prepare the text.
Which would you predict to be the acceptable word order in French: (2a) or (2b)?7
(2) a Les étudiants préparent tous le texte.
the students-MPL prepare all-MPL the text
6
For complications see also Exercise 15.
7
For some discussion of French see the comments section in Exercise 11.
Exercises 277
TSC04 07/18/2005, 03:44PM277
278 Chapter 4
b Les étudiants tous préparent le texte.
the students MPL all MPL prepare the text.
KEY
While in English the ﬂoating quantiﬁer precedes the inﬂected lexical verb, in French
it follows it:
(3) a The students all prepare the text.
*The students prepare all the text.
b Les étudiants préparent tous le texte.
the students prepare all the text
*Les étudiants tous préparent le texte.
Exercise 9 Floating quantiﬁers and adjuncts (T, E)
Discuss the positions of the underlined quantiﬁers in the following examples. Discuss
any problems that you encounter for our approach to the distribution of ﬂoating
quantiﬁers.
(1) We should all not criminalise something, in ignorance of the facts. (Independent,
13.3.2001, p. 8, col. 3)
(2) We’re all still waiting. (Guardian, 29.5.2002, p. 6, col. 8)
(3) We’re all just scraping by. (Guardian, 15.9.2003, p. 14, col. 1)
(4) Brand managers are all now trying to establish an emotional connection with
their target markets. (Guardian, 4.1.2001, p. 7, col. 1)
(5) Supermarket wines may all soon taste the same. (Times, 25.10.2001, p. 19,
col. 8)
KEY AND COMMENTS
The data above are problematic if we assume that the quantiﬁer must be stranded
in the speciﬁer of the VP and if the negation marker not and the adverbial adjuncts
still (2), just (3), now (4), and soon (5) are left-adjoined to VP. The order which is
predicted by our theory would be adjunct/not – ﬂoating quantiﬁer – verb.8
8
For discussion of the position of ﬂoating quantiﬁers see also Bobaljik (1995, 2003), Bowers
(2001, 2002), and Boskovic (2004). See also Exercises 10, 11, and 12 for comparative data.
TSC04 07/18/2005, 03:44PM278
Reﬁning Structures 279
Exercise 10 Floating quantiﬁers and manner adjuncts
(T, E)
Provide the bracketed representation for the structure of sentence in (1), indicating
unpronounced copies of moved constituents by strikethrough:
(1) The girls will all wait for the teacher.
Now suppose we want to augment the VP in (1) with a left-adjoined manner
adjunct very patiently. Based on your structure for (1) and on the discussion in the
chapter, where would you predict that the manner adjunct will be located? Would
you expect (2a) or (2b) to be the resulting word order? Is the prediction correct?
(2) a The girls will all very patiently wait for the teacher.
b The girls will very patiently all wait for the teacher.
KEY AND COMMENTS
In (1) the subject is separated from the ﬂoating quantiﬁer by the modal will.
Following the discussion in the chapter we would represent the structure of (1) as
in (3).
(3) a [IP The girls [I′ [I will] [VP all the girls [V wait] for the teacher]]].
In (3a) we signal the base position of the subject by strikethrough (the girls). In (3b)
we show the position of left-adjoined VP adjuncts.
(3) b [IP The girls [I′ [I will] [VP _______ [VP all the girls [V wait] for the teacher]]]].
If we insert the adjunct very patiently in this position, we would expect that this
adjunct would precede the ﬂoating quantiﬁer.
(3) c Inserting the adjunct:
[IP The girls [I′ [I will] [VP very patiently [VP all the girls [V wait] for the teacher]]]].
So we would predict that (2b) is the acceptable word order, contrary to fact: (2a) is
acceptable, (2b) is not:
(4) a The girls will all very patiently wait for the teacher.
b *The girls will very patiently all wait for the teacher.
These data are obviously problematic for our analysis.
Exercises 279
TSC04 07/18/2005, 03:44PM279
280 Chapter 4
Exercise 11 Floating quantiﬁers and adjuncts in
French (T, E)
Using the glosses provided, represent the structure of the French sentence in (1) by
means of labeled brackets, indicating unpronounced copies of moved constituents
by strikethrough:
(1) Les étudiants préparent tous le texte.
the students prepare-3PL all the text
‘The students all prepare the text.’
Now suppose we want to augment the VP in (1a) with a left-adjoined manner
adjunct très soigneusement (‘very carefully’). Based on your structure for (1) and on
the discussion in the chapter, where would you predict that the manner adjunct will
be located? Would you expect (2a) or (2b) to be the acceptable word order? Is the
prediction correct?
(2) a Les étudiants préparent tous très soigneusement le texte.
the students prepare all very carefully the text.
b Les étudiants préparent très soigneusement tous le texte.
the students prepare very carefully all the text.
KEY AND COMMENTS
In (1) the subject is separated from the ﬂoating quantiﬁer by the ﬁnite lexical verb
préparent (‘prepare’). The ﬂoating quantiﬁer is stranded in the speciﬁer position of
the VP. We derive this order by assuming that the verb has raised leftward to I.9
Following the discussion in the chapter we would represent the structure of (1) as
in (3a).
(3) a [IP Les étudiants [I′ [I prépare-nt] [VP tous les étudiants [V prépare] le texte]]].
In (3a) we signal the base position of the subject by strikethrough (les étudiants).
Using (3a) as a starting point, we adjoin the adjunct très soigneusement to VP. In
(3b) we show the position of left-adjoined VP adjuncts. We would predict that this
adjunct will precede the ﬂoating quantiﬁer.
(3) b [IP Les étudiants [I′ [I prépare-nt] [VP _______ [VP tous les étudiants [V prépare]
le texte]]]].
9
See Chapter 3, section 1.2.4.2.
TSC04 07/18/2005, 03:44PM280
Reﬁning Structures 281
If we insert the adjunct very patiently in this position, it will precede the ﬂoating
quantiﬁer.
(3) c Inserting the adjunct:
[IP Les étudiants [I′ [I prépare-nt] [VP très soigneusement [VP tous les étudiants
[V prépare] le texte]]]].
So we would expect (2b) to be the word order of the sentence, contrary to fact:
(2a) is acceptable, (2b) is not:
(4) a Les étudiants préparent tous très soigneusement le texte.
the students prepare all very carefully the text
‘The students all prepare the text very carefully.’
b *Les étudiants préparent très soigneusement tous le texte.
These data are obviously problematic for our analysis. Observe that we encountered
exactly the same problem for English in Exercise 10.
Exercise 12 Floating quantiﬁers and adjuncts in
Hebrew (T, E)
Consider the Hebrew examples in (1).10
Using the English glosses provided as your
guideline locate the subject, the ﬂoating quantiﬁer associated with the subject, and
the verb. Do you think that Hebrew has V-movement to I? Motivate your answer.
(1) a Ha-yeladim hiku kul-am }et ha-mora.
the-children hit all-3MPL the-teacher
‘The children all hit the teacher.’
b Ha-banot hadfu kul-an }et ha-kadur.
the-girls hit-back all-3FPL the-ball
‘The girls all hit back the ball.’
Using labeled brackets, represent the structure of (1a) in order to show the position
of the subject ha-yeladim, of the verb hiku, and of the ﬂoating quantiﬁer kul-am.
In our examples we note that the Hebrew quantiﬁer shows overt agreement with
the NP to its left.11
10
All the Hebrew data in this exercise are from Shlonsky (1991: 172, his (21)).
11
For related examples see Chapter 2, Exercise 19.
Exercises 281
TSC04 07/18/2005, 03:44PM281
282 Chapter 4
Now suppose we want to augment the VP in (1a) with a left-adjoined manner
adjunct, be-]axzariyut (‘with cruelty’). Based on your structure for (1a) and on the
discussion in the chapter, where would you predict that this adjunct will be found?
Would you expect (2a) or (2b) to be the acceptable order?
(2) a Ha-yeladim hiku be-}axzariyut kul-am }et ha-mora.
the-children hit with cruelty all-3MPL the-teacher
b Ha-yeladim hiku kul-am be-}axzariyut }et ha-mora.
the-children hit all-3MPL with cruelty the-teacher
KEY AND COMMENTS
In (1a) the subject is separated from the ﬂoating quantiﬁer by the lexical verb hiku
(‘hit’). If the ﬂoating quantiﬁer is stranded in the speciﬁer position of VP, we
conclude that the verb has raised leftward to I. Following the discussion in the
chapter we would provisionally represent the structure of (1a) as in (3a).
(3) a First hypothesis:
[IP Ha-yeladim [I′ [I hiku] [VP ha-yeladim kul-am [V hiku] }et ha-mora]]].
In (3a) we signal the base position of the subject by strikethrough (ha-yeladim). As
mentioned above, NPs precede the associated (inﬂected) quantiﬁers in Hebrew. We
signal the base position of the verb by strikethrough (hiku). If we use the structure
in (3a) as a starting point and we adjoin an adjunct to VP, we predict that this
adjunct will occupy the slot indicated in (3b):
(3) b [IP Ha-yeladim [I′ [I hiku] [VP _______ [VP ha-yeladim kul-am [V hiku] }et
ha-mora]]]].
If we insert the adjunct with cruelty in this position, we predict that this adjunct will
precede the ﬂoating quantiﬁer:
(3) c Inserting the adjunct:
[IP Ha-yeladim [I′ [I hiku] [VP be-}axzariyut [VP ha-yeladim kul-am [V hiku] }et
ha-mora]]]].
So we would expect (2a), repeated as (4a), to be the acceptable word order of the
sentence, contrary to fact: (4a) is ungrammatical. The adjunct should follow the
ﬂoating quantiﬁer as in (2b), repeated as (4b):
TSC04 07/18/2005, 03:44PM282
Reﬁning Structures 283
(4) a *Ha-yeladim hiku be-}axzariyut kul-am }et ha-mora.
the-children hit with cruelty all-3MPL the-teacher
b Ha-yeladim hiku kul-am be-}axzariyut }et ha-mora.
the-children hit all-3MPL with cruelty the-teacher
‘The children all cruelly hit the teacher.’
These data are obviously problematic for our hypothesis. We were confronted with
exactly the same problem for English in Exercise 10 and for French in Exercise 11.
Exercise 13 VSO patterns in Welsh (E)
In Exercise 7 we discussed word order in Irish, a Celtic VSO language. Based on the
analysis for Irish, how could we account for the position of the verb and the subject
in Welsh, another Celtic language?
(1) Darllen-ais i y llyfr.
read 1SG I the book
‘I read the book.’ (Harlow, 1981: 219)
(2) Soni-ais i am y dyn.
talked 1SG I about the man.
‘I talked about the man.’ (Harlow, 1981: 219)
(3) a Gwelodd y plentyn ceffyl.
saw the child horse
‘The child saw a horse.’
b Gwnaeth y plentyn weld ceffyl.
did the child see horse
‘The child did see a horse.’ (Holmberg & Platzack, 1995: 57)
(4) a Ennill -odd John.
win-past John
‘John won.’
b Gwnaeth John ennill.
did John win
‘John won.’ (Harlow, 1981: 223)
For complications see also Exercise 15. For a detailed discussion of Welsh see
Roberts (2005).
Exercises 283
TSC04 07/18/2005, 03:44PM283
284 Chapter 4
Exercise 14 Subject position(s) in Icelandic transitive
expletive constructions (T, E)
Exercises 9–12 have shown that the proposals according to which a ﬂoating quantiﬁer
related to the subject occupies SpecVP is not without its problems. In particular, the
distribution of VP-adjoined adjuncts suggests that this hypothesis may have to be
revised. The problem illustrated in the exercises was that while we would predict
that the ﬂoating quantiﬁers follow VP-adjoined adjuncts, they actually precede
them.
In section 3.2.2 of the chapter we discussed evidence for the hypothesis that there
is a subject position in SpecVP based on so-called transitive expletive constructions.
First try to reconstruct the argumentation on the basis of the Icelandic examples in
(25) in the chapter, repeated here in (1).
(1) a ad qad hefur einhver bordad epli.12
that there has someone eaten (an) apple (Vikner, 1995: 189)
‘that someone has eaten an apple.’
b ad qad bordadi einhver epli.
that there ate someone (an) apple (Vikner, 1995: 219)
‘that someone ate an apple.’
Using the discussion in the chapter as our basis, how would we analyze the Icelandic
example (2)?
(2) qad klarudu margar mys ostinn.
there ﬁnished many mice the cheese
‘Many mice ﬁnished the cheese.’
(Alexiadou & Anagnastopoulou, 2001: 198, their (14b))
Now consider the following Icelandic examples. What problems do they raise for
the analysis we elaborated?
(3) a qad klarudu margar mys alveg ostinn.
there ﬁnished many mice completely the cheese
‘Many mice ﬁnished the cheese completely.’
b *qad klarudu alveg margar mys ostinn
there ﬁnished completely many mice the cheese
(Alexiadou & Anagnastopoulou, 2001: 198, their (15))
12
For the position of the subordinating conjunction aD (‘that’) see Chapter 5, section 2.3.
TSC04 07/18/2005, 03:45PM284
Reﬁning Structures 285
KEY AND COMMENTS
If the subject of the existential pattern is stranded in the speciﬁer position of VP
in the Icelandic examples above, we deduce that the verb klaruDu (‘ﬁnished’) in
(2) has moved leftward. We can assume it has raised to I. We would represent the
structure of (2) as in (4a). We signal the base position of the verb by strikethrough
(klarudu).
(4) a [IP qad [I′ [I klarudu] [VP margar mys [V klarudu] ostinn]]].
If we use (4a) as a starting point and we adjoin the adjunct alveg (‘completely’) to
VP, we predict that this adjunct will precede the subject in SpecVP.
(4) b *[IP qad [I′ [I klarudu] [VP alveg [VP margar mys [V klarudu] ostinn]]]].
So we would expect (3b) to be the correct word order of the sentence, contrary to
fact: (3b)/(4b) is ungrammatical. The VP-adjoined adjunct has to precede the
preverbal subject as in (3a). Since (3b) is ungrammatical, the postverbal subject
apparently does not remain in the speciﬁer of VP after all: it must move leftward to
a position which is itself lower than the position occupied by the verb. Since the
subject is a full projection its landing site will either be an adjoined position or a
speciﬁer position.
(4) c [IP qad [I′ [I klarudu] [??? margar mys [VP alveg [VP margar mys [V klarudu]
ostinn]]]]].
Obviously these data raise a problem for our analysis, and the problem raised is
similar to that which we encountered in the preceding exercises.13
Exercise 15 VSO orders and subject positions
(E, presupposes Exercises 6, 7, and 13)
In Exercises 6 and 7 we discussed VSO languages. To account for Irish ellipsis data
in Exercise 7, we proposed that, in the examples in (5), repeated here as (1), the
postverbal subject occupies the speciﬁer of VP.
(1) a Thóg sí teach dófa ar an Mhullach Dubh.
raised she house for them on the Mullaghduff
‘She built a house for them in Mullaghduff.’
13
For a ﬁrst tentative analysis see Exercise 22 (10a) below. For discussion see Alexiadou and
Anagnostopoulou (2001), Roberts (2005: 7–46), and the references cited there.
Exercises 285
TSC04 07/18/2005, 03:45PM285
286 Chapter 4
b D’ól sí deaoch uische.
PAST drink she drink water
‘She drank a drink of water.’
How would we represent the structure of (2)?
(2) Chuala Róise an t-amhrán sin.
heard Róise that song
‘Rosie heard that song.’
(Based on Alexiadou & Anagnastopoulou, 2001: 200, their (20))
Discuss the problems raised for the analysis by the Irish data in (3):
(3) a Chuala Róise go mini roimhe an t-amhrán sin.
heard Róise often before it that song
b *Chuala go mini Róise an t-amhrán sin.
heard often Róise that song
(McCloskey, 1996; in Alexiadou & Anagnastopoulou, 2001: 200, their (20))
Discuss the consequences of these data for the analysis of the ellipsis pattern in Irish
in Exercise 7.
KEY AND COMMENTS
If the postverbal subject is stranded in the speciﬁer position of VP in the Irish
examples in (1) and in (2), we infer that the verb chuala (‘heard’) in (2) has moved
leftward. We can assume it has raised to I. Following the discussion in the chapter
we would represent the structure of (2) as in (4a).
(4) a [IP [I′ [I Chuala] [VP Róise chuala an t-amhrán sin]]].
In (4a) we signal the base position of the verb by strikethrough (chuala). If we use
the structure in (4a) as a starting point and we adjoin the adjunct go mini roimhe
(‘often before it’) to VP, we predict that the adjuncts precede the postverbal subject
in the speciﬁer of VP.
(4) b *[IP [I′ [I Chuala] [VP go mini roimhe [VP Róise chuala an t-amhrán sin]]]].
So we expect (3b) to be the correct word order of the sentence, contrary to fact:
(3b)/(4b) is ungrammatical. The adjuncts precede the subject as in (3a). Since (4b)
is actually ungrammatical, we conclude that the subject cannot remain in SpecVP
after all. Apparently, it must move leftward to a position which is lower than the
position occupied by the verb. Since the subject is a full projection, its landing site
will either be an adjoined position or a speciﬁer position.
TSC04 07/18/2005, 03:45PM286
Reﬁning Structures 287
(4) c [IP [I′ [I Chuala] [??? Róise [VP go mini roimhe [VP Róise chuala an t-amhrán sin]]]]].
This conclusion also raises a problem for the formulation of the ellipsis pattern in
Exercise 7. If the postverbal subject in Irish occupied a higher VP-external speciﬁer,
then ellipsis would have to target a more comprehensive constituent. If on the
other hand, the subject is adjoined to VP, then we can maintain the analysis
proposed in Exercise 7.
Once again, the Irish data discussed here raise a problem for our analysis, and the
problem raised is also similar to that which we encountered in the preceding exercises.14
Exercise 16 VSO orders and subjects in Modern
Greek (E, presupposes Exercises 7, 13,
14, and 15)
In Exercise 18 of Chapter 3 we already introduced the examples in (1)–(2) below.
We concluded there that the fact that the inﬂected verb is separated from its direct
object by an adjunct suggests that V moves to I in Modern Greek.
(1) O Petros egrafe panda megala grammata.
the Peter-NOM write-IMP-3SG always long letters
‘Peter always wrote long letters.’ (Alexiadou, 1997: 91 (17b))
(2) O Peter etroge sinithos sika.
the Peter-NOM ate-IMP-3SG usually ﬁgs
‘Peter usually ate ﬁgs.’ (Alexiadou, 1997: 93 (23a))
In the light of the discussion of Exercises 6 and 7 above, can we account for the
patterns in (3) and (4), given as (7) and (8) in Exercise 18 in Chapter 3?
(3) Diavaze sihna o Janis to vivlio.
read-3SG usually the John-NOM the book-ACC
‘John was usually reading the book.’ (Alexiadou, 1997: 62 (29))
(4) Efage kala o Janis.
ate-3SG well the John
‘John ate well.’ (Alexiadou, 1997: 131 (13a))
Do these examples raise the problems for the VSO analysis that we were confronted
with in Exercises 14 and 15 above?
14
For discussion see Alexiadou and Anagnostopoulou (2001) and the references cited there. For
a ﬁrst tentative analysis see Exercise 22 (10b) below.
Exercises 287
TSC04 07/18/2005, 03:45PM287
288 Chapter 4
KEY AND COMMENT
The postverbal subject in Modern Greek can be argued to occupy the speciﬁer of
VP. The Greek VSO patterns do not give rise to the problems encountered before.
Unlike the transitive expletive patterns in Icelandic discussed in Exercise 14, or the
VSO patterns in Irish discussed in Exercise 15, the postverbal subject in Greek
follows VP-adjoined adjuncts:
(5) [IP [I Diavaze] [VP sihna [VP o Janis diavaze to vivlio]]].
read usually the John-NOM the book-ACC
‘John was usually reading the book.’ (Alexiadou, 1997: 62 (29))
(6) [IP [I Efage] [VP kala [VP o Janis efagre]]].
ate-3SG well the John
‘John ate well.’ (Alexiadou, 1997: 131 (13a))
Exercise 17 Passivization and ﬂoating quantiﬁers (E)
Once again this exercise is longer and more discursive than the preceding exercises.
Its purpose is to carry further the type of argumentation elaborated in the chapter
and to see where that leads us. The exercise takes up the discussion of passivization
in Exercises 9, 10, 11, and 12 of Chapter 3. These exercises are presupposed for the
current exercise.
Describe the differences between sentences (1) and (2) below in terms of the
realization of the arguments of the verb arrest. Speciﬁcally, identify the subject of
each sentence.
(1) The police ofﬁcer will arrest the burglar.
(2) The burglar will be arrested.
(1) is an active sentence; (2) is its passive counterpart. Informally put, the passive
version in (2) is characterized by the fact that the AGENT role assigned by the verb
arrest is not overtly realized, and that the argument which is assigned the THEME role
is realized by the NP the burglar in the canonical subject position, SpecIP.
What sort of evidence could we invoke to say that in (1) the NP the police ofﬁcer
is the subject, while in (2) it is the NP the burglar that is the subject? For one thing,
the NP the police ofﬁcer agrees with the verb in (1). This is not easy to see in a
sentence with a modal auxiliary because, typically, such auxiliaries do not show
overt agreement, but if we replace the modal auxiliary by the auxiliary have to form
the present perfect, then agreement is overt.
TSC04 07/18/2005, 03:45PM288
Reﬁning Structures 289
(3) a The police ofﬁcer has arrested the burglar.
b The police ofﬁcers have arrested the burglar.
(4) a The burglar has been arrested.
b The burglars have been arrested.
Intuitively, we want to say that the THEME argument of the passivized verb becomes
the subject.
For the derivation of the active sentence, we assume that the THEME argument is
realized by the NP the burglar. The verb arrest is merged with that NP to form V′.
The meaning of a passive verb is closely similar to that of its active counterpart, and
in particular, the verb assigns the same thematic roles. We might try to capture the
semantic parallelism between an active verb and its passive counterpart by proposing
that the thematic role THEME of a passive verb is also assigned in the complement
position of the verb. In other words, the NP which realizes the THEME argument of a
passive verb is merged with V to form V′.
If the THEME argument of a passive verb is ﬁrst merged in V′, we would have to
conclude that it subsequently moves to the speciﬁer of IP, the canonical subject
position. In a passive sentence, the AGENT of the verb need not be overtly expressed
(cf. (2)). Let us provisionally assume that in the passive sentence the NP that
expresses the AGENT argument is not merged as the speciﬁer of the VP. Consequently,
the speciﬁer of the VP is not ﬁlled.15
In the passive sentence, the NP which realizes the THEME argument ends up in the
canonical subject position, SpecIP. If the speciﬁer of the VP remains unﬁlled, the
question arises whether the NP moves directly from the complement position
dominated by V′ to the canonical subject position, SpecIP, as shown in (5a), or
whether it moves via the speciﬁer of the VP, a position which, by hypothesis, has
remained empty (5b).
(5) a [IP The burglar [I will] [VP be [VP arrested the burglar]]].
b [IP The burglar [I will] [VP be [VP the burglar arrested the burglar]]].
Recall that, based on the discussion in Chapter 3, section 4, we assume that the
passive auxiliary be projects its own VP. This raises an additional question. If the
auxiliary be projects its own VP, this VP will also have a speciﬁer position, through
which the THEME NP the burglar might also transit on its way to SpecIP. This is
shown in representation (5c):
15
See Chapter 3, Exercises 10–13.
Exercises 289
TSC04 07/18/2005, 03:45PM289
290 Chapter 4
(5) c [IP The burglar [I will] [VP the burglar be [VP the burglar arrested the burglar]]].
The spell-out of the representations in (5) will not differ, since the non-overt copies
of the moved NP (the burglar) are not pronounced. All three representations spell
out as (5d).
(5) d The burglar will be arrested.
What kind of arguments could we use to choose between (5a), (5b), and (5c)?
What would a theoretical argument in favor of one of the representations be
like? What type of empirical evidence could be invoked in support of one analysis
or the other?
Consider the data in (6) and (7) from McCawley (1988: 90). Do they provide any
arguments for choosing between (5a), (5b), and (5c). Why?
(6) a ?The children have been all vaccinated.
b The children have all been vaccinated.
(7) a We have been all robbed many times.
b We have all been robbed many times.
KEY AND COMMENTS
The sentences in (6) and in (7) provide empirical support for the hypothesis
represented in (5c), namely that the subject of a passive sentence can move via the
intermediate speciﬁer positions. For (6), we could postulate that the ﬂoating quantiﬁer
all is stranded in one of the two intermediate landing sites. In representation (8),
we assume that auxiliary have projects its own VP and moves to I.16
Write out the
sentences in their spell-out form:
(8) a ?[IP The children [I have] [VP the children have [VP the children been [VP all the
children vaccinated all the children]]]].
b [IP The children [I have] [VP the children have [VP all the children been [VP all
the children vaccinated all the children]]]].
c [IP The children [I have] [VP all the children have [VP all the children been [VP
all the children vaccinated all the children]]]].
As you can see: if you spell out (8b) and (8c), the resulting strings will be identical
and correspond to (7b): this is because there is no overt element in the V position
in which have is ﬁrst merged: the auxiliary have moves to I. What kind of data
16
For the position of auxiliaries see the discussion in Chapter 3, section 4.
TSC04 07/18/2005, 03:45PM290
Reﬁning Structures 291
might be able to distinguish between the pattern in (8b) and that in (8c)? What we
would need is for have to occupy the V-position in which it is ﬁrst merged, so that
we can see whether the ﬂoating quantiﬁer all is in the speciﬁer of the VP headed by
all or in that of the VP headed by be. How could we ensure that have remains in
V? We could prevent have from moving to I if we inserted an element in I. Let us
represent this element as X. In (9a) all will be found to the right of have, in (9b) it
will be to the left of have:
(9) a [IP The children [I X] [VP the children have [VP all the children been [VP all the
children vaccinated all the children]]]].
= The children X have all been vaccinated.
b [IP The children [I X] [VP all the children have [VP all the children been [VP all
the children vaccinated all the children]]]].
= The children X all have been vaccinated.
Which element could be inserted as X? If we were to insert a modal auxiliary under
I, then have would remain in the position V and the ﬂoating quantiﬁer appears
either to its right or to its left:
(10) a [IP The children [I will] [VP the children have [VP all the children been [VP all
the children vaccinated all the children]]]].
= The children will have all been vaccinated.
b [IP The children [I will] [VP all the children have [VP all the children been [VP
all the children vaccinated all the children]]]].
= The children will all have been vaccinated.
We tentatively conclude that there is some empirical support for postulating that
passive sentences involve step-by-step movement of a constituent from the
“complement position” in V′ to the canonical subject position, SpecIP.
The argumentation used here is analogous to that developed in section 4 of
Chapter 4. In that discussion, we argued that in a sentence with multiple auxiliaries
the subject NP moves from the speciﬁer of the lexical verb via the intermediate
speciﬁers of the VPs projected by the auxiliaries. Recall that in the discussion we
used the data in (11) as evidence (McCawley, 1988: 90):
(11) a The students must all have been sleeping.
b The students must have all been sleeping.
c The students must have been all sleeping.
We might also address the choice between representations (5a), (5b), and (5c) from
a theoretical point of view. Given that (11) offers evidence that the subject moves
step by step from its base position to its landing site, we may propose a general
hypothesis that movement always proceeds step by step.
Exercises 291
TSC04 07/18/2005, 03:45PM291
292 Chapter 4
Exercise 18 Passivization and ﬂoating quantiﬁers
(E, presupposes Exercise 17)
Discuss the relevance of the position of the underlined ﬂoating quantiﬁer in the
following examples for the derivation of passive sentences.
(1) It’s too bad they couldn’t have all been tried together. (Guardian, 10.9.2002,
p. 6, col. 2)
(2) This is no doubt all being facilitated by what British ministers have hailed as
the restrained behaviour of the Northern Alliance. (Guardian, 27.11.2001,
p. 7, col. 1)
(3) ‘Can they really all be so misled?’ (Guardian, 6.2.2002, p. 4, col. 6)
(4) These have now both been announced ahead of schedule and the company
says this is why Sir Peter has called it a day. (Guardian, 1.11.2001, p. 3, col. 3)
(5) The sums will only all be awarded if BT does better than its rivals over the next
three years in a league table based on a ﬁnancial metric known as total
shareholder return (TSR). (based on Guardian, 1.11.2001, p. 3, col. 8)
Exercise 19 Floating quantiﬁers and the base
position of the subject in passive
sentences (T, presupposes Exercise 17)
This exercise is again more discursive than some of the preceding exercises. Our
purpose is once again to carry further the type of argumentation elaborated in the
chapter and see where that would lead us. The exercise presupposes Exercise 17.
Return once more to the data in (6) discussed in Exercise 17, and the representations
in (8). The examples and their representations are repeated here in (1)
and (2):
(1) a ?The children have been all vaccinated. (McCawley, 1988: 90)
b The children have all been vaccinated. (McCawley, 1988: 90)
(2) a ?[IP The children [I have] [VP the children have [VP the children been [VP all the
children vaccinated all the children]]]].
b [IP The children [I have] [VP the children have [VP all the children been [VP all
the children vaccinated all the children]]]].
TSC04 07/18/2005, 03:45PM292
Reﬁning Structures 293
c [IP The children [I have] [VP all the children have [VP all the children been [VP
all the children vaccinated all the children]]]].
Though consistent with the hypothesis of stepwise movement, the sentences in (1)
raise a problem. Throughout the discussion of passive sentences we have been
assuming that the subject of a passive sentence originates as the complement of V:
it ﬁrst merges with V to form V′.
We also assume that ﬂoating quantiﬁers allow us to trace the route taken by the
moved constituent. In the discussion of the two subject positions in Chapter 4, we
proposed that the ﬂoating quantiﬁer signals the base position of the subject, SpecVP.
In the discussion of sentences with auxiliaries we concluded that a ﬂoating quantiﬁer
was to be found in each of the positions containing a non-pronounced copy of the
subject.
In line with our analysis of the relation of a ﬂoating quantiﬁer and the subject and
using representations such as those in (2), which would you expect to be the lowest
possible position of the quantiﬁer associated with the subject of the passive sentence?
If the ﬂoating quantiﬁer could be found in literally all the positions that host a nonpronounced
copy of the subject, we would also expect it to be available in the
thematic position of the subject. In passive sentences, the thematic position of the
subject is the complement position in V′, that is a position to the right of the verb.
Construct the appropriate sentence and check whether it is grammatical.
(2) d [IP The children [I have] [VP all the children have [VP all the children been [VP
all the children vaccinated all the children]]]].
= *The children have been vaccinated all.
Surprisingly, a ﬂoating quantiﬁer associated with the subject of a passive sentence
can apparently never occur in the base position of that subject, that is, the position
in which the NP is ﬁrst merged with V and in which its thematic role is assigned.
This obviously requires further investigation since we have been assuming that in
active sentences a ﬂoating quantiﬁer can signal the thematic position of the subject.17
Exercise 20 Passivization and existential patterns
(E, presupposes Exercises 17, 18, and 19)
Consider the underlined strings in (1a). Identify the lexical verbs. What is the
subject of the ﬁrst underlined string (i)? Motivate your answer. What is the subject
of the second string (ii)? Using the concepts elaborated in Chapter 4, how could we
relate the two patterns?
17
For discussion see, for instance, Sportiche (1988) and Koopman and Sportiche (1991).
Exercises 293
TSC04 07/18/2005, 03:46PM293
294 Chapter 4
(1) There have been changes in trends and (i) there’s a lot less pesticide being
used in the countryside these days than a decade ago. The problem is that,
although (ii) a lot less is being used, what is being used is a lot more effective
– so there’re less insects and seeds for birds to eat. (Guardian, 2.12.2003,
p. 6, col. 6)
Clearly, it is tempting to relate the underlined patterns in (1) along the lines of the
discussion of existential structures in section 3.2.2 in this chapter. (1a) is a simpliﬁed
version of (i) and can be compared to (2a), (1b) is a simpliﬁed version of (ii) and can
be compared to (2b).18
(1) a A lot less pesticide is being used.
(2) a Three students are working on the project.
(1) b There is a lot less pesticide being used.
(2) b There are three students working on the project.
We could argue that (1b) supports the proposal that the subject of a passive
sentence originates in a lower position and moves to the speciﬁer of IP. This
example would illustrate the pattern in which the NP a lot less pesticide, the THEME
argument of the verb use, has been moved to the speciﬁer of the VP headed by the
passive auxiliary be, while the expletive there is inserted in SpecIP. In (1a), the THEME
is moved to SpecIP. In the representations below strikethrough shows the copies of
the moved NP a lot less pesticide. We insert is in I to simplify the representation.
(3) a (= 1b) [IP There is [VP a lot less pesticide being [VP a lot less pesticide used a
lot less pesticide]]].
b (= 1a) [IP A lot less pesticide is [VP a lot less pesticides being [VP a lot less
pesticide used a lot less pesticide]]].
Discuss the derivation of the passive sentences in (4).
(4) a There has been some work presented in this workshop that really inspired
me and made me perhaps rethink what I want to do in my own work, she
said. (New York Times, 28.11.2002, p. B6, col. 1)
b There have been nearly two dozen people kidnapped since Nicholas E. Berg,
a radio-tower builder from Pennsylvania, was taken captive in Iraq in April
and later beheaded. (New York Times, 1.8.2004, p. 1, section 4, col. 4)
c There were already around 18,000 new titles a year being published in
1960. (Guardian, Review, 13.3.2004, p. 10, col. 2)
18
See section 3.2.2.1.
TSC04 07/18/2005, 03:46PM294
Reﬁning Structures 295
See also Basilico (1998) for more careful discussion of passive and existential sentences,
Caponigro and Schütze (2003) for extension to other sentence types and to
Italian passivization. See Law (1999) for critical discussion.
Exercise 21 Floating quantiﬁers in Hebrew (E)
Consider the examples in (1),19
which we discussed in Exercise 19 of Chapter 2. We
showed in that exercise that both the string kol ya-yeladim (‘all the children’), and
the string ha yeladim kul-am (‘the children all’) are constituents. Using the glosses
as a basis, describe the difference between the underlined NPs in (1a) and (1b).
(1) a Kol ha-yeladim zarku }avanim.
all the children threw stones
‘All the children threw stones.’
b Ha-yeladim kul-am zarku }avanim.
the children all-3MPL threw stones
‘The children all threw stones.’
Exercise 12 of the current chapter showed that quantiﬁers can be ﬂoated in Hebrew.
Consider the following examples: what restrictions on quantiﬁer ﬂoating do these
data reveal?
(2) a Ha-yeladim hiku kul-am }et ha-mora.
the-children hit all-3MPL the-teacher
‘The children all hit the teacher.’
a′ *Ha-yeladim hiku kol }et ha-mora.
the-children hit all the-teacher
b Ha-banot hadfu kul-an }et ha-kadur.
the-girls hit-back all-3FPL the-ball
‘The girls all hit back the ball.’
b′ *Ha-banot hadfu kol }et ha-kadur.
the-girls hit-back all the-ball
KEY AND COMMENTS
The quantiﬁer all has two forms in Hebrew: it either precedes the NP and then it is
invariant and takes the form kol or it follows the noun and then it is inﬂected for
19
The data in this exercise are based on Shlonsky (1991: 163–4). Thanks to Ur Shlonsky for help
with (1).
Exercises 295
TSC04 07/18/2005, 03:46PM295
296 Chapter 4
person, gender, and number. Based on the primed examples in (2) we conclude
that the invariant form of the quantiﬁer, kol, cannot be stranded. Only the inﬂected
postnominal quantiﬁer can be stranded. This suggests that for an NP to be able to
strand an associated quantiﬁer it must appear to its left, the position at which it
triggers the agreeing form.20
Exercise 22 Reﬁning structures: From one functional
head to many (T, E)
This is yet another long and discursive exercise. Its purpose is once again to explore
one line of argumentation and see where that leads us. The goal of the discussion
is not to provide a complete answer for the problem raised. Rather, the exercise
tries to show how new hypotheses come about through the interplay of data and
theory. The analysis elaborated in this exercise is not necessarily the deﬁnitive
one. It is one way of handling the data in terms of the theory we have been
elaborating.
In this exercise we explore the implications of the hypothesis that syntactic structure
determines interpretation. We ﬁrst examine the problems of interpretation raised by
one example and try to provide an analysis. After the discussion further examples
are given as an additional exercise. Consider example (1a).
(1) a Turkey could until very recently privately congratulate itself on narrowly
escaping a place in the front line. (Guardian, 21.11.2003, p. 16, col. 3)
We are interested in the underlined section of the example. Before we start the
discussion, we will simplify the example by removing distracting elements that are
not relevant to the point at issue. Such a modiﬁcation of the data is not “cheating.”
Simplifying the data is a perfectly legitimate operation in scientiﬁc work: we are like
the scientist who isolates the relevant data in the raw material he or she has collected,
removing material that is not relevant for the enquiry. The simpliﬁed example we
will be working on is (1b).
(1) b Turkey could until very recently privately congratulate itself.
Draw a tree diagram for (1b). Represent the base position of the subject by means
of the strikethrough notation. Consider the resulting structure in the light of the
discussion of adjunct scope in section 2.1 of the chapter. Do the positions of the
adjuncts correspond to their relative scope? What would be the scope domain of
the modal could?
20
See Shlonsky (1991) for an account.
TSC04 07/18/2005, 03:46PM296
Reﬁning Structures 297
KEY AND COMMENTS
The structure of (1b) can be represented as in (2a):
(2) a
IP
I′NP
I VP
VPPP
AdvP VP
Turkeyuntil very recentlycouldTurkey
NP V′
privately
V NP
itselfcongratulate
The positions of the adjuncts until very recently and privately reﬂect their relative
scope. Privately modiﬁes the VP domain; it speciﬁes the manner in which ‘Turkey
congratulated itself’. As a ﬁrst approximation we could say that the temporal PP
until very recently delimits the period in which ‘Turkey privately congratulated
itself’. The PP until very recently has wider scope than the adverb phrase privately.
(2) b until very recently > privately
We can read off the scope of these two adjuncts directly from the structure.
While the adjuncts occupy positions that mirror their scope, this is not true for
the modal auxiliary could in I. In this example the modal could expressed an ability.
If we read the scope of an element off from the structure then the modal could
would have to have scope over both adjuncts. However, we will naturally paraphrase
(1) as in (3), where until very recently in fact modiﬁes the period in which Turkey
had the ability to congratulate itself in private. So until very recently should not just
have scope over VP, it should also have scope over could.
(3) Until very recently it was the case that Turkey was able to privately congratulate
itself.
until very recently > could > privately
Exercises 297
TSC04 07/18/2005, 03:46PM297
298 Chapter 4
This means that the scope relations of until very recently and could cannot be
directly read off from the structure. The modal auxiliary could occupies the position
in I. What we would really need is for could to be represented in a lower
position than the adjunct until very recently. That lower position would determine
its scope.
The problem that arises for our representation is not new. It is similar to that
encountered in the discussion of the position of the subject NP in Chapters 3 and 4.
At some point we were faced with a situation in which the subject was seen to
occupy a position in the string, SpecIP, but for reasons of interpretation we also
wanted it to be associated with another, lower, position. As a result of the movement
analysis elaborated in Chapter 4, subject properties are split over two positions:
SpecIP and SpecVP. Both positions are input to semantic interpretation: in SpecVP
the subject is thematically related to V; in SpecIP it serves as the anchor of the
information in the sentence.
If we want the representation of (1b) to represent the relative scope of the
adjuncts and the modal could we have to associate the modal also with a lower
position. Let us pursue this point and see what happens. We elaborate an alternative
representation that includes a position X to represent the lower position of the
modal auxiliary. A ﬁrst very approximate representation is given in (4). We propose
that the modal be inserted in X as in (4a). To attain the position it occupies in the
spell-out of the sentence (that is, the way the sentence is pronounced), the modal
moves up to the position I. As always, we assume that a moved constituent leaves
an unpronounced copy in its base position (4b). We cannot yet detail the brackets
here as we will ﬁrst have to determine which constituents they are.
(4) Provisional representations:
a [IP Turkey [I] until very recently [X could] [VP privately [VP Turkey congratulate
itself]]].
b [IP Turkey [I could] until very recently [X could] [VP privately [VP Turkey
congratulate itself]]].
In its lower position (X), could is in the scope of the PP until very recently. We
propose that the modal moves from X to I, which is a head position. So far, we
have been assuming that heads (auxiliaries, verbs in French) move to I. Let us
generalize this idea and assume that the modal in X is a head and that X is a head
position. If X is a head then we expect it to project a phrase, “XP.”
In our earlier representation (2a) the PP until very recently was VP-adjoined. But
in the new representation, the PP until very recently can no longer be left-adjoined
to VP. The head X intervenes and it heads its own projection, XP. As a result, we
propose that the adjunct until very recently is adjoined to the projection XP. (4c)
completes the provisional representation above:
(4) c [IP Turkey [I could] [XP until very recently [XP [X could] [VP privately [VP Turkey
congratulate itself]]]]].
TSC04 07/18/2005, 03:46PM298
Reﬁning Structures 299
We have to determine the category of the new head X. Is X a head of the category
V? This would mean that we treat the modal could as another auxiliary with its
own VP-projection.21
However, one objection to this is that English modals do not
have any non-ﬁnite forms: could cannot be found as a participle, for instance:22
(5) a *He has so far could ﬁnish the book.
b *Turkey has until very recently could privately congratulate itself.
Another proposal is to say that X is functional head in the IP domain. Recall that the
content of I encodes the link between subject and VP in terms of tense, modality,
etc. On the basis of the preceding discussion we could tentatively “decompose”
I and postulate that there are two “inﬂectional heads” in the sentence: I1 and I2.
We might, for example, specify that I2 encodes modality and that I1 encodes Tense,
though we would obviously have to look at many more examples to substantiate
this analysis.
(4) d I1P
I1′NP
I1 I2P
I2PPP
I2′
until very
recently
couldTurkey
I2 VP
could
AdvP VP
NP V′
V NP
congratulate itselfTurkeyprivately
21
See Chapter 3, section 4.3.
22
See the discussion of (67) in section 4.3 of Chapter 3.
Exercises 299
TSC04 07/18/2005, 03:46PM299
300 Chapter 4
The PP until very recently is now adjoined to the projection of I2, meaning it
delimits the duration of the ‘ability’. Privately continues to have scope only over the
VP congratulate itself. Privately does not restrict the ability. The relative scope of
the two adjuncts as represented in (3) is retained in our new representation: until
very recently has scope over a larger constituent (I2P) than privately.
That modals may be associated with a lower position is conﬁrmed by those
dialects of English that allow double modals:23
(5) He’ll can get you one. (Brown, 1992: 75)
We might object that our representations are becoming more and more complicated.
This is not obviously the case, though. Though the structures we have are deﬁnitely
more articulated, it is not necessarily true that the theory we are elaborating is
itself becoming more complex. In order to arrive at the articulated structure
above, we have merely implemented the argumentation that we had developed
throughout. To use a metaphor: we have only used the tools that we have in our
toolbox.
In the light of the preceding discussion, comment on the relative scope of the
modal auxiliaries and the underlined adverbs in the following examples:
(6) a TB can usually successfully be treated with a course of antibiotics. (Guardian,
13.8.2004, p. 6, col. 8)
b There are a few technical things the president could probably learn. (New
York Times, 1.8.2004, Section 4, p. 4, col. 2)
Below we repeat some of the examples from Exercise 15 in Chapter 3. Could the
discussion above help us to account for the unexpected position of the underlined
adjunct?
(7) Michigan coach Lloyd Carr says sorry, but he simply can’t go against the
coaching association’s policy and vote for Southern California as college football’s
No 1 team. (Chicago Tribune, 3.1.2004, S3, p. 1, col. 1)
(8) I never could make out what those damned dots meant. (Independent,
14.04.2001, p. 20, col. 6)
(9) Today Gulliver still can barely stand to be among the Yahoos. (Guardian, G2,
31.8.2004, p. 12, col. 1)
23
For double modals in Scots see Brown (1992) and Miller (1993). For American varieties see for
instance Battistella (1995). For modals in earlier stages of English see Lightfoot (1979) and
Roberts (1985).
TSC04 07/18/2005, 03:46PM300
Reﬁning Structures 301
The hypothesis that a functional head such as I can be decomposed into a more
articulated structure is due to Pollock (1989) and has become known as the Split
Inﬂ Hypothesis.24
The Split Inﬂ hypothesis has further consequences. Recall that in previous exercises
we discovered the need for an additional position for the subject. This was the case
for instance, with respect to the transitive expletive construction in Exercise 14 and
with respect to the Irish VSO patterns in Exercise 15. Pursuing the Split Inﬂ hypothesis
we might postulate that the lower subject positions identiﬁed in the earlier exercises
correspond to the speciﬁer positions of lower functional heads.
(10) a [I1P qad [I′ [I klarudu] [I2P margar mys [VP alveg [VP margar mys [V klarudu]
ostinn]]]]].
b [I1P [I′ [I Chuala] [I2P Róise [VP go mini roimhe [VP Róise chuala an t-amhrán
sin]]]]].
Exercise 23 Predicate inversion in English (E)
Consider the following examples:25
they illustrate a construction type referred
to as predicate inversion or “inversion around be” (Emonds, 1976; Birner, 1992,
1994; Birner & Ward, 1992).
(1) a Complicating matters is cost. (Washington Post, 10.12.2002, p. A16, col. 1)
b Helping to run the house were a cook, a housemaid and a manservant.
(Carol Shields, Jane Austin, 2001: 123)
c (The thieves paid the £6 entrance fee and made their way to the staircase
hall, where they overpowered the guide and took the painting from the
wall.) Waiting for them outside were the Volkswagen and at least one other,
but probably two, accomplices. (Guardian, G2, 28.8.2003, p. 6, col. 2)
These examples can be contrasted with the more neutral word order patterns in (2).26
(2) a Cost is complicating matters.
b A cook, a housemaid and a manservant were helping to run the house.
c The Volkswagen and at least one other, but probably two, accomplices
were waiting for them outside.
24
Pollock’s own analysis was based on other empirical material. See also Cormack and Smith
(2002) for the scope of modals. See Svenonius (2002), Ernst (2002, 2004), Alexiadou (2004b),
Cinque (1999, 2004a), and Nilsen (2004), among others for the distribution of adverbials. See
Cinque (1999) for further decomposition of I.
25
Examples (1a) and (1b) were discussed in Exercise 12 of Chapter 1.
26
For the discourse function of the predicate inversion construction see Birner (1992, 1994),
Birner and Ward (1992).
Exercises 301
TSC04 07/18/2005, 03:46PM301
302 Chapter 4
How do we derive the sentences in (2)? Assuming that the sentences in (1) have a
similar underlying structure, how could we derive them?
KEY AND COMMENTS
Following the analysis elaborated we could assume that (2a) is derived in the by
now familiar way, with the subject NP cost being merged as the speciﬁer of the VP
and moving out to the speciﬁer position of IP. The auxiliary was is merged as the
head of a VP and moves to I, the subject NP cost undergoes step-by-step movement
to the speciﬁer of IP:27
(3) a [IP [NP Cost] [I was] [VP cost be [VP cost [V′ complicating matters]]]].
To derive the predicate inversion pattern in (1a), we might propose that the
constituent containing the verb complicating and the object matters moves, while
the subject NP cost is stranded in the speciﬁer of the VP:
(3) b Hypothesis
[IP [V′ Complicating matters] [I was] [VP [V′ complicating matters] be [VP cost
[V′ complicating matters]]]].
Observe that in this case we have to move an intermediate projection (V′). This is
potentially worrying, as we have not really used such a projection for any movement
operation yet.
TWO MORE COMPLICATIONS
Assuming the analysis in (3b), we would predict that there is relatively free variation
between the two patterns discussed here, i.e., that, depending on the informational
organization of the sentence we want to put forward, we can either front the
subject to SpecIP (as in the examples in (2)) or we can front V′ (as in the examples
in (1)). Is this prediction correct?
(4) a I wonder whether three students from Romania will be waiting in the corridor.
b *I wonder whether waiting in the corridor will be three students from Romania.
(5) a I expect the candidates for the exam to be waiting in the corridor.
b *I expect waiting in the corridor to be the candidates for the exam.
Discuss the problem raised for the analysis in (3b) by the application of predicate
inversion in the following examples. Could we claim that V′ has been fronted,
stranding the subject in the speciﬁer of VP?
27
See section 4 of this chapter.
TSC04 07/18/2005, 03:47PM302
Reﬁning Structures 303
(6) Still peeking through the attitudes of Labour politicians high and low are
glimpses of doubt about their right to be where they are. (Guardian, 9.4.2002,
p. 8, col. 3)
(7) Conﬁdently riding the horse is the beautiful, dark-haired Tamsin. (Observer,
24.20.2004, Review, p. 5, col. 2)
(8) A recent survey of senior-class presidents in high schools around the nation
has shown that 73% approve of draft registration for 18-year-old men and
51% favor prayer in public schools. Sharply dividing the class presidents was
the issue of abortion – 50% supported a woman’s right to terminate an
unwanted pregnancy; 32.5% opposed it. (Philadelphia Inquirer, 3.9.1983,
from Birner & Ward, 1992: 9, their (21a))
(9) Ashenden duly distributed the Welcome Trusthouse Forte forms, already
completed for the sections dealing with Company, Next Destination, Settlement
of Account, Arrival, Departure and Nationality. Only remaining for the
tourists to ﬁll in were the four sections headed Home Address, Telephone,
Passport Number and Signature. (Colin Dexter, The Jewel that was Ours,
1992: 23)
(10) Now living with the Morrises was Janey’s sister, Bessie Burden, who had
joined them in the last few months at Red House, after her father died.
(Fiona MacCarthy, William Morris, 1995: 198)
Exercise 24 Language typology (T, E, presupposes
Exercises 6, 7, 13, and 16)
In this chapter we have discussed a number of different languages. We have
identiﬁed two properties that determine their word order:
(i) The verb moves to I.
(ii) The subject moves to SpecIP.
Based on the discussion in the chapter and on the data in the preceding exercises,
complete Table 1, writing + if the language displays the property and – if it does not.
The combination of two variable properties leads to four possible combinations.
In the blanks in Table 2 ﬁll in the languages that display the corresponding combination.
Are all combinations attested?28
Exercises 303
28
For a discussion of the typology see Baker (2002).
TSC04 07/18/2005, 03:47PM303
304 Chapter 4
Table 1 Language typology
Language V to I Subject to SpecIP
Arabic
English
French
Irish
Modern Greek
Welsh
Table 2 Inventory of languages
V to I V remains in V
Subject to SpecIP
Subject does not move to SpecIP
TSC04 07/18/2005, 03:47PM304
The Periphery of the Sentence 305
5 The Periphery of the Sentence
Discussion
Contents
0 Introduction 307
1 Recapitulation 308
2 Constructing the Periphery of the Sentence 313
2.1 Direct yes/no questions 313
2.2 A ﬁller for SpecXP? 314
2.3 Conjunctions and SAI 316
2.4 Some terminology 320
2.5 Evidence for copies 320
2.6 Indirect constituent questions 322
2.7 Uniqueness of speciﬁers and multiple questions 325
2.8 Subject interrogatives 327
2.8.1 Indirect questions 327
2.8.2 Direct subject interrogatives and inversion 328
2.9 A note on that-omission in English 330
3 How Far Can You Move? And How Do You Get There? 331
3.1 “Long” movement 331
3.2 Intervention effects and step-by-step movement 334
3.2.1 Blocking movement 334
3.2.1.1 A hypothesis 334
3.2.1.2 Testing the hypothesis 335
3.2.2 Empirical evidence for step-by-step movement 337
3.2.3 Subjects and objects 338
3.2.3.1 The complementizer that 338
3.2.3.2 Blocking subject movement 340
4 The Periphery of Non-Finite Clauses 340
4.1 For as a conjunction for non-ﬁnite clauses 341
4.2 Non-ﬁnite interrogative clauses 341
TSC05 07/18/2005, 03:47PM305
306 Chapter 5
4.2.1 Whether 341
4.2.2 Fronted interrogative constituents 343
5 Relative Clauses: An Introduction 344
5.1 Movement of the relative pronoun 344
5.2 Constraints on movement: Some predictions 346
5.2.1 Intervention effects on the movement of the relative
pronoun 346
5.2.2 Subject movement and the complementizer that 348
5.3 A non-overt relative pronoun 349
5.3.1 The data: Relative clauses without a relative pronoun? 349
5.3.2 Evidence for movement of a non-overt pronoun 350
5.3.2.1 Step-by-step movement 350
5.3.2.2 Subject movement and the complementizer that 351
6 Summary 352
TSC05 07/18/2005, 03:47PM306
The Periphery of the Sentence 307
0 Introduction
This ﬁnal chapter completes the overview of the structure of the sentence. So far, we
have mainly concentrated on sentences whose leftmost constituent is the subject.
We did occasionally discuss examples in which the subject is preceded by some other
constituent, but the discussion did not go into details. In this chapter, we explore how
interrogative constituents and auxiliaries to the left of the canonical subject position
are structurally integrated into the sentence. We will also examine how a fully formed
sentence can be integrated into a larger structure as an embedded clause.
We continue to assume that structures are derived by means of the operations
Merge (‘assemble, put together’) and Move. As before, we use the binary branching
format for structure: a head merges with a complement, the resulting constituent
merges with a speciﬁer to complete the projection, and two fully formed projections
can be combined by adjunction. The current chapter focuses on the importance of
the operation Move. One of the issues that will be raised is how far a constituent
can be moved and whether there are any obstacles to the Move operation, that is,
whether there are factors that can stop a constituent from moving from one position
to another.
The chapter returns to one of the empirical issues raised in Chapter 1, the derivation
of questions. In English, question formation usually requires that there be some
material to the left of the canonical position of the subject. So we will investigate
how to integrate this area of the sentence into the structure we have been assuming.
In addition, we brieﬂy look at the derivation of relative clauses, which also implicate
the area of the sentence to the left of the subject. The mechanisms elaborated for the
derivation of questions will be extended to the derivation of relative clauses.
In the chapter we will put forward a number of constraints on the way a constituent
is moved and this will lead to the prediction that sentences violating these
constraints should be ungrammatical. That is to say, such sentences should not be
generated by our grammar. The analysis of grammatical sentences exemplifying
movement may conﬁrm that their derivation obeys the constraints we have postulated.
But the fact that all the sentences we come across obey the constraints is not
sufﬁcient to conﬁrm the prediction that all sentences violating such constraints are
ungrammatical. At various points in the chapter we will test the prediction made by
our analysis by experiment: we will construct examples that contain precisely the
patterns that violate the constraints. According to the prediction following from our
hypothesis such sentences should be unacceptable.
The chapter is organized as follows. Section 1 recapitulates the main issues covered
in the previous chapters. In section 2 we examine how ﬁnite interrogative sentences
are derived in English. We will see that the operation Move can be used to account
for the derivation of such sentences. In section 3 we examine some of the constraints
TSC05 07/18/2005, 03:47PM307
308 Chapter 5
1
As mentioned in Chapter 4, note 8, the component of syntactic theory which sums up the
format for syntactic structure is referred to as X-bar theory.
on the operation Move. In section 4 we brieﬂy look at non-ﬁnite interrogative
clauses. Section 5 deals with relative clauses. Section 6 is a summary of the chapter.
1 Recapitulation
Throughout this book, we have elaborated a set of hypotheses for the structure of
sentences. These hypotheses form a theory, a system of principles of syntax, which
we take to be the instructions for the derivation of sentences. Ideally, the theory we
elaborate must be powerful enough to allow us to derive all and only sentences that
are acceptable. In other words, the theory should not be too powerful: it should not
overgenerate, that is, it should not allow the derivation of unacceptable sentences.
At this point it may be important to underline that a scientiﬁc theory is not static,
it is not like a painting, for instance, which is an entity that, once it is ﬁnished, is
“ﬁxed” and does not change any more though our perception of it may change. A
scientiﬁc theory is dynamic. Confronted with new empirical data or with new theoretical
hypotheses, scientists may well reconsider some of their earlier hypotheses. A
scientiﬁc theory must continuously be evaluated and adapted. Evaluating a theory
of syntax may be done with respect to two types of questions:
(i) What is the internal organization of the theory? Are the principles that make
up the theory internally consistent? Are there any redundant principles that we
could eliminate, leading to a more economical theory?
(ii) What is the empirical coverage of the theory? Does the theory generate the
right type of sentence? Does the theory generate any unacceptable sentences?
Try to summarize the various components of the theory that we have elaborated
and illustrate how they account for the derivation of sentence (1a). Does the current
version of our theory allow us to derive sentences (1b), (1c), and (1d)?
(1) a The spy should destroy these instructions.
b Obviously, the spy should destroy these instructions.
c These instructions, the spy should destroy.
d Should the spy destroy these instructions?
In the discussion we have adopted a general hypothesis about linguistic structure
summarized schematically in (2).1
Our hypothesis is that all constituents are derived
by merging (= putting together) two constituents. We can read the derivation of
representation (2a) from the bottom to the top in the following way. We ﬁrst merge
the head of the constituent, X, with its complement, forming X′; then we merge X′
TSC05 07/18/2005, 03:47PM308
The Periphery of the Sentence 309
with another constituent, the speciﬁer, to form XP. The complement and the speciﬁer
are closely related to the head X and together they form the core constituent.
(2) a X-bar theory: The blueprint for structure
XP
X′speciﬁer
complementX
In addition, a fully formed constituent (which we could label YP, ZP, etc.) may be
adjoined to another constituent. Adjunction is another application of Merge. Its
speciﬁc property is that it merges constituents that are already complete in themselves.
We assume that adjunction may apply both to the left and to the right. Adjunction
is represented in (2b).
(2) b left-adjunction right-adjunction
XP
XPYP
XP
YPXP
In addition to assembling a structure by the operation Merge, that is by merging
two constituents, we can modify the structure obtained at some point by the operation
Move. This operation removes a constituent from one position, and merges it
in another position in the same structure. Move allows us to represent the fact that
one constituent “belongs” to two positions in the structure. Continuing to work our
way upward in the sentence, we take a constituent from a lower position and move
it into a higher position. For instance, to account for the fact that the subject of a
sentence has a close semantic relationship with the verb, we proposed in Chapter 4
that the subject is ﬁrst merged in a VP-internal position. To explain how the subject
ends up in its canonical position, the speciﬁer of IP, we propose that it moves to that
position. In order to represent that the moved constituent also “belongs” in the
lower position we use the device of copies: a copy of the moved constituent remains
in the base position, while the moved constituent is merged in the higher position. It
is the higher “moved” copy that eventually is spelt out (pronounced). We use the
strikethrough representation (subject) to signal the position of the non-pronounced
copy.2
Using the format in (2), we have elaborated a representation of the basic
structure of the sentence. In the ﬁrst stage of the analysis, we had arrived at the
structure in (3) below. A projection of the lexical head (V) is dominated by a projection
of a functional head that encodes inﬂectional features (see Chapter 3, section
1.2.3). VP is a lexical projection; IP is a functional projection.
2
We also saw in Chapter 4, section 4, that movement proceeds stepwise. We will come back to
this point in detail in section 3.2 below.
TSC05 07/18/2005, 03:47PM309
310 Chapter 5
(3) The structure of the sentence
IP
I′Spec
VPI
V′Spec
. . .Vsubject
subject
The system allows us to derive (1a). The semantic nucleus of (1a) is the verb destroy.
The verb denotes the situation we are dealing with and it determines which constituents
are obligatory in the sentence.3
So, in order to build a sentence, we start off
with the verb destroy (i). We merge it with its complement, here the NP these
instructions, which serves to narrow down the action, denoting the entity affected
by the act of destroying (ii). The next step is to insert the subject the spy, which
denotes the agent of destroy (iii). The completed VP will give us a complete picture
of the activity expressed in the sentence: the VP tells us who did what.
(4) (i) V = destroy
(ii) V′
NPV
these instructionsdestroy
(iii) VP
V′NP
destroy these instructionsthe spy
Next we build the dominating functional structure, the projection IP. I merges with
VP. We insert the modal auxiliary should in I. Should locates the situation expressed
in the sentence in the future and shows that the action expressed in the VP is
3
See Chapter 3, sections 3.1, 3.2, 3.3.
TSC05 07/18/2005, 03:47PM310
The Periphery of the Sentence 311
desirable to the speaker. I also determines a perspective, an anchor for the information
given in the sentence. I projects and we create its speciﬁer. The subject moves
up to the speciﬁer of I (iv).
(4) (iv) IP
V′
NP
NP
the spy these instructionsdestroyThe spy
NPV
I′
I VP
should
We build the structure starting from the semantic core, the verb, and by using
the operation Merge, which combines two units, and the operation Move, which
displaces units that are already part of the structure to a higher position.
Can we also derive the other examples in (1) using the above representations?
With respect to (1b), the derivation can proceed as in (4), but there is an additional
constituent in initial position, the adverbial phrase obviously.
(5) a [IP the spy should destroy these instructions]
b + [AdvP obviously]
This additional constituent is optional; it does not receive a thematic role from the
verb. The communicative function of obviously is to express the speaker’s attitude
to the content of the sentence. Recall that we can adjoin constituents to maximal projections.
We could assemble (1b) by (left)-adjoining the AdvP to the projection IP.
(6) IP
IPAdvP
the spy should destroy these instructionsObviously
Compare (1a) and (1c), repeated here as (7a), in terms of their form and their
meaning.
TSC05 07/18/2005, 03:47PM311
312 Chapter 5
(7) a These instructions, the spy should destroy.
(7a) has the same components as (1a) and the two sentences are very similar in
interpretation. However, the order of the constituents is different. In (7a), the direct
object NP, these instructions, does not occupy its canonical position to the immediate
right of the verb. Rather, it occupies a sentence-initial position. The fronted NP
these instructions signals what the information contained in the sentence is going to
be about, it is the sentence topic.4
But the NP these instructions also denotes the
entity affected directly by the action denoted by the verb destroy. So, following the
assembly procedure outlined above, we would expect the NP these instructions to
be merged as the complement of the verb. Following the line of reasoning adopted,
we could say that the NP these instructions ﬁrst merges with V and that it is subsequently
moved to the left-peripheral position.5
Tentatively, we could say that the
object moves out of VP and adjoins to IP:6
(7) b IP
IPNP
the spy should destroy these instructionsThese instructions
(1d) is familiar from the ﬁrst chapter of the book; it illustrates a direct question. The
subject NP the spy is preceded by the inverted modal auxiliary, should. Let us
assume that, as before, the auxiliary should is ﬁrst merged in I. The fronting of the
auxiliary to the left of the subject signals the illocutionary force of the sentence, that
is, it shows that the sentence is a question.7
In Chapter 38
we concluded that in SAI
the content of I inverts with the subject. In this chapter we will integrate the inverted
constituent into the structure. We will do this using the format we have elaborated.
As we will see, we will not need to invoke any novel mechanisms: we will be able
to derive the word order of interrogative sentences by using a combination of the
operations Merge and Move. As before, we elaborate a hypothesis and test it by
examining its predictions.
4
For an early discussion of the concept “topic” see Reinhart (1981).
5
See also Chapter 3, section 1.2.3.4, and for more illustrations Exercise 15 in Chapter 2.
6
One question that arises is whether a fronted object should really be presented as having the
same relation with IP as the adverbial obviously in (1b). We will not address this issue here
as it would lead us to a lengthy discussion beyond the scope of an introductory text. For
some discussion of object fronting see Rizzi (1997), Haegeman (2003), and the references cited
there.
7
Recall from Chapter 1, section 2.3.2, that SAI does not always give rise to question interpretation.
8
Section 1.2.4.
TSC05 07/18/2005, 03:47PM312
The Periphery of the Sentence 313
2 Constructing the Periphery of the Sentence
2.1 Direct yes/no questions
In Chapter 3 we observed that to form a direct question in English we move the
ﬁnite auxiliary from I to a position to the left of the subject (SAI). When a ﬁnite
sentence lacks an auxiliary we move the content of I and we insert the auxiliary do
as a supporting element to allow the inverted inﬂectional morpheme to survive in
the position to the left of the subject. In French direct questions, the verb itself is
able to move to a position to the left of the subject (SVI). This is compatible with
our hypothesis that it is the content of I that moves leftward: we had seen that in
French the lexical verb can move to I.
Let us assume that the inverted auxiliary in English and the inverted verb in
French end up in the same position to the left of the subject.9
SAI (or SVI in French)
means that the content of I moves to the left of the subject. I is a head, so SAI/SVI
is movement of a head. So far, we have been assuming that a head moves to a head
position. For instance, we proposed in Chapter 3, section 4.3, that English auxiliaries
are merged in V and move to I; French auxiliaries as well as French lexical verbs
merge in a head position and move to I. If SAI/SVI is head movement, the landing
site of SAI/SVI must be a head position to the left of the subject position. Let us
provisionally label this head “X.” The head X merges with IP. By the blueprint for
structure outlined above, X and IP form a constituent X′.
(8) a X′
IPX
you will buy the paper
tu achèteras le journal
Eng. Will
Fr. Achèteras
Would it be possible to let the structure terminate here? Given the way we conceive
the structure of constituents, is (8a) a legitimate representation? The answer is both
“yes” and “no.” (8a) does correspond to our blueprint (2a). However, once we
have postulated a head, which combines with a complement, we would expect there
to be a completed projection (“XP”). This would also make a speciﬁer position
available. The complement of the head X in (8a) is IP, that is, the sentence. (8b) is
what we would expect to generate following the system we have set up.
9
Exercise 23 in Chapter 4 shows that we should not derive all sentences in which an auxiliary
precedes the subject by means of SAI. Exercise 15 in the current chapter shows that not all
postverbal subjects in French necessarily occupy the same position.
TSC05 07/18/2005, 03:47PM313
314 Chapter 5
(8) b XP
X′Spec
IPX
At this point, (8b) is only motivated on a theoretical basis. We have not provided
any empirical support for projecting the XP level; nor have we discussed what kind
of elements could ﬁll the speciﬁer position. A further question is to determine the
nature of the head X. We turn to these issues immediately.
2.2 A ﬁller for SpecXP?
What kinds of constituents could occupy speciﬁer of X in (8b)? Or, putting it differently,
is there any type of constituent that naturally tends to precede the inverted
auxiliary in direct questions? We are looking for a constituent that “speciﬁes” the
question as expressed by the inversion of the auxiliary. The underlined examples in
(9) suggest one answer. It is tempting to associate the interrogative constituent, or
the wh-constituent, to the left of the auxiliary with the speciﬁer position. Would this
make sense in terms of the relation of this constituent to the question it introduces?
(9) a “Which way are you going? Shall we walk a bit?” he began, putting the
second question before the ﬁrst was answered. (Edith Wharton, The House
of Mirth, 1998: 201)
b How much can I afford? (New York Times, 28.4.2003, p. A22, advertisement
Fleet)
Compare the interpretations of the questions (10a) and (10b), which are both based
on (9a).
(10) a Are you going to the park?
b Which way are you going?
Question (10a) could be said to be general: it is about whether or not a certain
activity (‘going to the park’) will take place. (10a) is called a yes/no question: the
expected answer is yes or no. (10b) is different: when asking (10b), we take it for
granted (or we presuppose) that the activity of going somewhere will take place, but
we don’t yet know its goal. The initial wh-constituent which way speciﬁes the
focus or the scope of the question: it speciﬁes the domain of application of the
question, what the question is about. It is also an indication of the type of answer
expected: as a reply to (10b) we expect to be provided with some constituent
to match the interrogative constituent. Questions like (10b) are called constituent
TSC05 07/18/2005, 03:48PM314
The Periphery of the Sentence 315
questions. Because most interrogative constituents contain a word beginning with
wh- such questions are also called wh-questions.
If we assume that the initial wh-constituent in a constituent question occupies the
speciﬁer position of XP in (8b), it has a speciﬁer–head relation with the head X,
which hosts the inverted auxiliary. The fronted wh-constituent interacts with the
illocutionary force: its function is to specify the domain of the question. On the
other hand, remember that the fronted wh-constituent also has a function within
the IP domain. More precisely, the wh-constituent in our example has a semantic
relationship with the verb. The NP which way indicates the goal argument of the
verb going. So we also want to link the wh-constituent to a VP-internal thematic
position. We can do this by ﬁrst merging the NP which way VP-internally as a
complement of the verb. Which way is merged with V to form V′, and it then moves
to the speciﬁer position of XP, leaving a copy in the VP. As shown by (11), we can
use the same analysis to derive example (9b).10
The wh-constituent how much is
both the complement of the verb afford and the element that marks the scope of the
question. Again it ﬁrst merges with V and then moves to SpecXP, leaving a copy.
(11) XP
X′Spec
X IP
I′Spec
I
you
I
are
can
a Which way
b How much
are
can
Spec V′
V NP
you
I
VP
going
afford
which way
how much
= CLAUSE/SENTENCE
What label shall we use to designate the head X? The ﬁllers for X we have come
across so far are auxiliaries which have moved from the position I to X. Would it
make sense to re-label X as “Aux”? What would be the problems with this label?
10
For subject movement see Chapter 4.
We will insert inﬂected auxiliaries under I to simplify the representations.
TSC05 07/18/2005, 03:48PM315
316 Chapter 5
Though the label Aux might at ﬁrst sight seem appropriate when we restrict the
scope of our enquiry to English, it becomes less adequate once we take into consideration
the fact that French lexical verbs can precede the subject as a result of
inversion.11
Recall that this was also possible in Old English.12
Labeling the head
position occupied by the inverted head Aux would mean that we only capture a
subset of facts.13
To derive SAI/SVI the content of I moves to the left of the subject. Whether an
auxiliary or a verb moves to the left of the subject depends on whether an auxiliary
or a verb occupies I. However, the label I is also not suitable to designate the
landing site X of the inverted head because I is already used as the label for the
linking head that merges with VP and whose speciﬁer is the subject. We do want to
keep the two heads apart. For one thing, as can be seen in (11), the speciﬁer of X
need not be a subject.14
One problem for labeling X is due to the fact that so far, we have only been
looking at heads that are moved to X but which themselves originate in another
position. For the labeling of X it would be important to ﬁnd out whether there are
any elements that are merged directly in X. We turn to this point in the next section.
2.3 Conjunctions and SAI
Discuss the realization of the objects of the underlined verbs in the following
examples.
(12) a What do you like and hate about your job? I love that I am talking about
skiing all day. (Independent, Traveller, 25.9.1999, p. 8, col. 7)
b Well, he said, I like that there are so many pretty girls. (New York Times,
2.1.2003, p. B2, col. 1, letter to the editor from Miles Fisher, 19, Dallas)
c Your ﬁrst brush with the capital is incredibly disorienting. You don’t know
whether one Tube stop is going to be 100 yards or if it’s going to be miles.
(The Times, 25.11.2000, p. 12, col. 1)
For each of the examples the object of the verb is itself realized by a sentence. In
(12c) the object of know consists of the co-ordination of two sentences:
(12) a′ I love [that I am talking about skiing all day].
b′ I like [that there are so many pretty girls].
c′ You don’t know [[whether one Tube stop is going to be 100 yards]
or [if it’s going to be miles]].
11
Chapter 1, section 3.1 and Chapter 3, section 1.2.4.
12
Chapter 1, section 3.2 and Chapter 3, section 1.2.4.
13
We will discover a further objection to using the label Aux for X in section 2.3.
14
On whether the speciﬁer of XP actually can be a subject see the discussion in section 2.8.
TSC05 07/18/2005, 03:48PM316
The Periphery of the Sentence 317
When a sentence functions as a constituent inside another sentence it is often referred
to by means of the term clause; it is a subordinate clause or an embedded clause.
The bracketed subordinate clauses in the primed examples above function as complements
of the verbs and are therefore also called complement clauses. As these
clauses realize the direct objects of the verbs, another term that is used is object
clauses. The subordinate clauses are introduced by the words that, whether, and if.
(13) a that I am talking about skiing all day
b that there are so many pretty girls
c whether one Tube stop is going to be 100 yards or
if it’s going to be miles
Traditionally, words such as that, if, whether, which introduce embedded clauses,
are labeled (subordinating) conjunctions. In this book, we will use the more recent
term complementizer. Complementizers serve to introduce subordinate or embedded
clauses; they enable a clause to function inside another clause, for instance as a
complement of a verb. Complementizers are functional elements.15
The class of words
that function as complementizers is closed:16
any given language has a restricted
set of words that have this “complementizing” function. Complementizers serve to
encode illocutionary force. The complementizers if and whether signal that the subordinate
clause is interrogative, while that indicates that the clause is declarative.17
As shown by (12c), a subordinate interrogative clause can be introduced either by
if or by whether. Could we combine both conjunctions in one and the same clause?
Try inserting if in the ﬁrst embedded clause in (12c) and try adding whether to the
second. This gives us (12c″):
(12) c″ You don’t know *whether if one Tube stop is going to be 100 yards or
*if whether it’s going to be miles.
15
See Chapter 3, section 3.5 on the difference between lexical elements and functional elements.
16
See Chapter 1, section 2.4, and Chapter 3, section 4.3 on “closed classes.”
17
In Chapter 1, note 10, we signaled that a more careful semantic analysis should make a distinction
between the concepts “interrogative clause” and “question.” For a good and accessible
discussion see Huddleston (1994). For a more advanced semantic analysis see Ginzburg
(1999). See McCloskey (forthcoming) for some syntactic consequences of the typology of
interrogatives.
It is also not the case that all embedded clauses introduced by the complementizer that are
assertions. Compare, for instance, (ia) and (ib):
(i) a John explained [that we cannot afford this car].
b John regrets [that they cannot afford this car].
While in (ia) the content of the embedded clause is asserted (by John), in (ib) it is presupposed.
See Kiparsky and Kiparsky (1971) and Hooper and Thompson (1973) for early discussion of
the syntactic consequences of the typology. See Hegarty (2003) for a recent discussion.
TSC05 07/18/2005, 03:49PM317
318 Chapter 5
Apparently, the conjunctions if and whether are in complementary distribution.18
To account for the fact that we cannot insert both if and whether to the left of a
subject, we could propose that if and whether occupy the same position.
Replace the two co-ordinated indirect questions in (12c) by two co-ordinated
direct questions.
(14) a Is one Tube stop going to be 100 yards
b or is it going to be miles?
In each of the co-ordinated direct yes/no questions in (14), the subject is preceded
by the auxiliary; in indirect yes/no questions the subject is preceded by the interrogative
complementizer if or whether. Can we combine SAI and the conjunction if
or whether? To check whether this is possible, we could try to ﬁnd relevant examples
in which SAI coincides with the insertion of an interrogative complementizer. However,
even if we do not ﬁnd any examples, this would not ultimately prove that
the combination of the conjunction if/whether with subject auxiliary inversion is
impossible. We can also try to run an experiment in which we control the data, we
construct the type of sentence we want to study, and we examine the result. We
need a sentence in which subject-auxiliary inversion has taken place and which is at
the same time introduced by a conjunction. Our experiment could consist of three
stages. First we construct a direct question with SAI; then we turn that example into
an embedded interrogative; ﬁnally, we insert a conjunction. Unfortunately, the second
step of the experiment is not straightforward. We might start from the direct questions
in (14), which display SAI. However, when we embed the sentences in (14) as
complements of a verb the resulting sentences are unacceptable for most speakers
of English:19
(14) a′ *You don’t know is one Tube stop going to be 100 yards.
b′ *You don’t know is it going to be miles.
Inserting a complementizer does not improve the sentences (14a″, b″), but this
does not mean that the combination of the complementizer and the inversion is the
cause of the ungrammaticality. Even without the complementizer the examples were
ungrammatical.
(14) a″ *You don’t know if is one Tube stop going to be 100 yards.
b″ *You don’t know if is it going to be miles.
To test the compatibility of SAI with complementizers, we need to work with
legitimate instances of embedded inversion and then we can check whether such
18
See Chapter 2, section 3.1.2.2 for another illustration of complementary distribution.
19
Hiberno English, the variant of English spoken in Ireland, does allow for this pattern. See
Henry (1995). Exercise 5 illustrates this variant of English. The exercise can be tackled at the
end of the present section.
TSC05 07/18/2005, 03:49PM318
The Periphery of the Sentence 319
examples allow insertion of the complementizers if or whether. Consider the
following examples:20
(15) a Had I known you were coming, I would have baked a cake.
b Had the money not been returned, the evidence would have pointed strongly
to a conclusion that the NRCC “ﬁnanced” the Forum. (Washington Post,
29.4.2003, p. A18, col. 3)
In (15a, b) SAI is found in conditional clauses. The examples can be paraphrased as
in (16). The examples in (15) could be considered as the second step of the experiment:
the sentences display embedded SAI. (16) shows that the conjunction if can
also introduce a conditional clause.
(16) a If I had known you were coming, I would have baked a cake.
b If the money had not been returned, the evidence would have pointed
strongly to a conclusion that the NRCC “ﬁnanced” the Forum.
We can now get on to the third step of the experiment, that is, we can insert the
conjunction if in the conditional clauses in (15). Is the result acceptable?
(16) a′ *If had I known you were coming, I would have baked a cake.
b′ *If had the money not been returned, the evidence would have pointed
strongly to a conclusion that the NRCC “ﬁnanced” the Forum.
The ungrammaticality of (16a′) and (16b′) suggests that the inverted auxiliary and
the complementizer are in complementary distribution. Why would this be? One
plausible account could be that there is a single head position to the left of the
subject and that this position hosts either the inverted auxiliary or the conjunction.
Recall that in the preceding section we had already identiﬁed the head position X
in representation (8). This head was argued to host the inverted auxiliary. The
complementary distribution of the inverted auxiliary and the complementizer leads
us to the hypothesis that X also hosts the complementizer. Whereas the auxiliary
moves to X from I, the complementizer does not originate in another position. Let
us propose that the complementizer is directly merged in the position X. Let us
therefore re-label the position X as “C.” C stands for complementizer. The position
C is the position where complementizers are merged with the clause they introduce.
The choice of C (if, whether, that) signals illocutionary force. In the absence of a
conjunction, the position C may host the inverted auxiliary.21
If we re-label X in (8) as C, then we will also re-label X′ as C′, and XP as CP. The
speciﬁer of XP in (8b) becomes the speciﬁer of CP (or SpecCP).
20
We discussed them in Chapter 1, section 2.3.2, examples (32) and (33).
21
Representation (17) is simpliﬁed in that we have not indicated the unpronounced copy of the
subject in the speciﬁer of VP. This is because the focus of the discussion is on the CP area.
TSC05 07/18/2005, 03:49PM319
320 Chapter 5
(17) CP
C′Spec
a Which way
b How much
c
d
e
f
g
IPC
you are going which way
I can afford how much
there are so many pretty girls
it’s going to be miles
one Tube stop is going to be 100 yards
I had known you were coming
I had known you were coming
are
can
that
if
whether
If
Had
= CLAUSE/SENTENCE
2.4 Some terminology
Sentences are derived by means of two operations: Merge and Move.22
Merge
assembles a structure of hierarchically organized constituents. Move displaces a
constituent from the position in which it has been merged and merges it at some
higher position in the structure. Moved constituents leave a copy.23
Merge is used to
insert a novel element into a structure, and also to (re)insert a moved element into
the structure.24
2.5 Evidence for copies
In interrogative sentences, the sentence-initial wh-constituent is moved from a position
inside the sentence. Moved constituents leave copies to ensure that the relations
they have with various constituents in the sentence can remain encoded. For
instance, when a direct object NP is moved to the sentence-initial position, its copy
in the VP allows us to establish the thematic relation with the verb, while the
pronounced copy in the initial position speciﬁes the scope of the question. This
motivation for copies is theoretical or conceptual. It follows from the way we set up
our theory. We assume that the syntactic structure of a sentence maps onto its
interpretation, that there is a direct correlation with the form of the sentence and its
interpretation.
22
For a brief discussion of a system that only uses Merge (and “Multiple Merge”) see Chapter 4,
section 3.1.
23
Chapter 4, section 3.1.
24
Exercises 1, 3.
TSC05 07/18/2005, 03:49PM320
The Periphery of the Sentence 321
In the discussion of the movement of the subject25
we provided some empirical
evidence for our movement hypothesis. One type of evidence was provided by
sentences in which (part of) the subject had been stranded in a lower position, giving
rise to a discontinuous constituent. Would there be any similar empirical support
for the hypothesis that moved interrogative constituents leave copies? Consider the
two formulations of Mrs Pettigrew’s question in the following extract. Identify the
object of the verb mean in both formulations. In what way could such an example
be relevant for the theory of movement that we have been elaborating here?
(18) “What do you mean by that exactly?”, said Mrs Pettigrew, “What exactly do
you mean?” (Muriel Spark, Memento Mori, 1977: 81)
The two formulations of Mrs Pettigrew’s question are very similar:
(19) a What exactly do you mean?
b What do you mean exactly?
In (19a), the object of mean is what exactly, an interrogative constituent or a
wh-constituent. This wh-constituent occupies an initial position. We analyze (19a)
as an instantiation of leftward movement: we have displaced the wh-constituent
what exactly from the complement position of the verb mean to the speciﬁer position
of the projection CP. The fact that the string can be moved suggests that what
exactly is a constituent.
(19) c [CP What exactly [C′ [C do] [IP you [present] [VP you mean what exactly]]]]?26
In the alternative formulation of the question we ﬁnd a discontinuous constituent
what . . . exactly. The direct object NP of mean is split up: its interrogative component,
what, is moved to SpecCP and is separated from exactly. The two variants of
the sentence seem to have the same interpretation. We can relate the derivation of
the two variants if we assume that in (19b) the moved interrogative pronoun what
and the adverb exactly are ﬁrst merged as one constituent and then movement of
what strands the adverb exactly. In (19d) exactly is represented as a residue of the
moved constituent.27
(19) d [CP What [C′ [C do] [IP you [present] [VP you mean [what exactly]]]]]?
25
See Chapter 4, section 3.2.
26
For present tense inﬂection we use the symbol -s for the third person singular morpheme. We
use the symbol present for other person and number combinations, because they lack overt
manifestations. In (19d) and similar examples, do in C spells out the moved present tense ending.
For more careful discussion of the inﬂection of “bare verbs” in English see Pollock (1994).
27
For more careful discussion of these patterns see also McCloskey (2000: 63–4, note 8 and the
reference cited there).
TSC05 07/18/2005, 03:49PM321
322 Chapter 5
McCloskey (2000) provides the West Ulster English examples in (20) (McCloskey,
2000: 58, his (2)). Identify the objects of the underlined verbs. How do we derive
the word order in these questions? Represent the structure of the examples, using
labeled bracketing and indicating non-pronounced copies of moved constituents by
strikethrough:
(20) a What all did you get for Christmas?
b Who all did you meet when you were in Derry?
Based on the discussion so far, we would provide the representations in (21):
(21) a [CP [NP What all] [C′ [C did] [IP you [I -ed] [VP you get [NP what all] for
Christmas]]]]?
b [CP [NP Who all] [C′ [C did] [IP you [I -ed] [VP you meet [NP who all] when you
were in Derry]]]]?
McCloskey (2000: 58) reports that West Ulster English usage offers some interesting
variants of the sentences above:
In addition to [20], though, West Ulster English allows [22]:
[22] a What did you get all for Christmas?
b Who did you meet all when you were in Derry?
The quantiﬁer all in [22ab] is construed with the interrogative pronoun . . . that is, the
examples in [22] are synonymous (completely so, as far as I have been able to tell) with
those in [20].
Using the examples in (20) as a starting point, how would you represent the alternative
formulations for the questions as given in (22)? How do these examples bear on
our current discussion?
Once again, it seems plausible that what . . . all in (22a) and who . . . all in (22b) are
discontinuous constituents. While the interrogative pronouns what and who have
moved to the speciﬁer of CP, the associated quantiﬁer all has been stranded in the
base position. The ﬂoated quantiﬁer would then signal the base position of the object.
(23) a [CP What [C′ [C did] [IP you [I -ed] [VP you get [what all] for Christmas]]]]?
b [CP Who [C′ [C did] [IP you [I -ed] [VP you meet [who all] when you were in
Derry]]]]?
2.6 Indirect constituent questions
We saw in our discussion of question formation28
that indirect questions do not
tend to give rise to SAI. Identify the indirect questions in (24):
28
Chapter 1, especially section 2.3.
TSC05 07/18/2005, 03:50PM322
The Periphery of the Sentence 323
(24) a David S. Chu, under-secretary of defense for personnel and readiness, said
last week the Pentagon needs a freer hand in determining who is hired,
how much they are paid and what types of jobs they do in order to shape
a department that is more agile in carrying out its mission of national
defense. (Washington Post, 29.4.2003, p. Q21, col. 5)
b You could rehash the night before, talk about what adjustments you need
to make, whether it was great or whether you caved and did something
you probably shouldn’t have. (Washington Post, 10.12.2002, p. F4, col. 2)
c Mrs Smegma showed me to a room, then gave me a tour of the facilities
and outlined the many complicated rules for residing there – when breakfast
was served . . . which hours of the day I would have to vacate the
premises and during which brief period a bath was permitted (these seemed,
oddly, to coincide), how much notice I should give if I intended to receive
a phone call or remain out after 10 p.m. (Adapted from Bill Bryson, Notes
from a Small Island, 1996: 15–16)
d She knew how many novels she would write and what they would be
about. (Guardian, 2.8.2003, Review, p. 14, col. 4)
In the following examples, the embedded questions do display SAI, which is not the
most usual order.29
Rephrase the indirect questions in (25) to undo SAI:
(25) a All he wants to know is which boxes have I ticked on the forms he keeps
giving me to ﬁll in. (Guardian, G2, 15.3.2001, p. 9, col. 8)
b People ask why was I not at Coniston when Bluebird was raised. (Guardian,
15.3.2001, p. 5, col. 8)
In this section, we make the derivation of indirect questions more precise. We
continue to adopt the hypotheses elaborated already. We will ﬁrst experiment on
the basis of an invented example. This is because in such an example we control the
material and we ensure that we are not distracted by complications that are perhaps
not relevant for the point at issue.
Let us look at the underlined string in (26a). Is this string a constituent? What is
its grammatical function? Consider the internal structure of the underlined string.
Identify the lexical verb. What is the direct object? Identify the subject. Do subject
and object occupy their canonical positions?
(26) a The students wondered which analysis they should adopt.
Among other things, the underlined string contains a lexical verb, adopt, and a
subject NP, they. The subject occupies its canonical position: it precedes the modal
auxiliary should, which occupies I. The NP which analysis, which functions as the
29
For an (accessible) discussion of inversion in embedded questions in Hiberno English see
Henry (1995).
TSC05 07/18/2005, 03:50PM323
324 Chapter 5
object of adopt, does not occupy its canonical position in the VP headed by adopt,
but it has moved leftward to a peripheral position. Could the object have remained
in its canonical position to the right of the verb? If we simply return the interrogative
object NP to its base position the result is not acceptable:
(26) b *The students wondered they should adopt which analysis.
Why should this be? What could be the reason why the NP which analysis has
to move leftward? And where does it move? Replace the object in (26b) by a
non-interrogative object. Is the resulting string acceptable?
(26) c *The students wondered they should adopt this analysis.
Can we now make (26c) grammatical by moving the object NP this analysis leftward?
(26) d *The students wondered this analysis they should adopt.
How can we rephrase (26c) to make it acceptable? One strategy is to insert the conjunction
if or whether to the immediate left of the subject of the embedded clause:
(26) e The students wondered if they should adopt this analysis.
f The students wondered whether they should adopt this analysis.
An alternative strategy is to replace the verb wonder by believe:
(26) g The students believed they should adopt this analysis.
The data suggest that there is a correlation between the presence of the verb wonder
in the main clause and the internal make-up of the embedded clause. With wonder
as a main verb, we either introduce the embedded clause by means of the conjunction
if/whether or we move a wh-constituent to the left of the subject. These two
manipulations are related: the verb wonder selects an interrogative clause as its
complement. The conjunctions if and whether are one way of encoding illocutionary
force, they signal that the embedded clause they introduce is interrogative. An alternative
way of encoding interrogative illocutionary force is to shift a wh-constituent to
the left in the embedded clause. In the latter case, the shifted wh-constituent will
specify the scope of the question.
For direct interrogatives, we have already proposed that interrogative force is
encoded in the part of the structure labeled CP. In keeping with this hypothesis,
it would be simplest (and therefore it would be preferable) to use the structures
that we have in place to account for the position of the fronted wh-constituent
in indirect questions. The function of the fronted wh-constituent is to specify the
scope of the question. We can propose that, as was the case in direct questions, the
wh-constituent in indirect questions is moved to the speciﬁer of the embedded CP.
TSC05 07/18/2005, 03:50PM324
The Periphery of the Sentence 325
In indirect questions in which we front a wh-constituent, the position C remains
obligatorily empty in Modern English. However, we have to postulate a head C
because, in the absence of the head C, we would no longer be able to project a constituent
and create the relevant speciﬁer position. In (27) below we give a schematic
representation.30
(27) V′
CPV
Spec C′
I′NP
I
theywhich analysiswonder should
Spec V′
V NP
they
VP
adopt which analysis
C IP
2.7 Uniqueness of speciﬁers and multiple questions
The canonical position of the subject is SpecIP: among other things, the subject
speciﬁes the domain of application of the VP, the subject says what the sentence
is about. For instance in (28a), the property “run an Indian restaurant” is applied
to the referent of the subject NP John:
(28) a John runs an Indian restaurant.
We know that IP has only one speciﬁer (28b). If we want two NPs to function as
subjects of a sentence, we co-ordinate them, thus turning them into one constituent
(28c).
(28) b *[NP Mary] [NP John] will run an Indian restaurant.
c [NP [NP Mary] and [NP John]] will run an Indian restaurant.
30
Exercises 4, 5, 6, 7, 8.
TSC05 07/18/2005, 03:50PM325
326 Chapter 5
Similarly, the speciﬁer of CP encodes the scope of the question, the domain to which
the question is applied. Identify the fronted interrogative constituents in embedded
questions in the following examples.
(29) a It has become a matter of dispute what John said to George.
b It has become a matter of dispute to whom John said that he had to leave.
c It has become a matter of dispute who said to George that he had to leave.
Consider (30). What is wrong with these examples?
(30) a *It has become a matter of dispute what to whom John said.
b *It has become a matter of dispute to whom who said these things.
c *It has become a matter of dispute what who said to John.
The examples in (30) all contain two sentence-initial wh-constituents in the embedded
clause. Apparently this is not possible. How could we account for this? Observe that
we cannot claim that a question cannot bear on more than one constituent. (31) contains
attested examples of questions with more than one interrogative constituent:
(31) a I don’t know who spotted who ﬁrst, but within a couple of days The Bruiser
was Jeffrey’s minder and conﬁdante. (Guardian, G2, 7.4.2003, p. 2, col. 2)
b For the past few months, KidsPost has had articles about who does what at
a newspaper. (Washington Post, 29.4.2003, p. C13, col. 2)
c The chronological structure seems sensible in theory, but in fact is dull.
As she marches doggedly forward from Praxitiles’ Aphrodite of Knidos to
Madonna, we can predict which names will occur in which order. (Guardian,
29.3.2003, p. 14, col. 5)
d There was still much confusion, and many yawns, about exactly who had
said what when, and sent which emails to whom when. (Guardian, 2.3.2002,
The Editor, p. 5, col. 2)
e I will be letting people know more about exactly who knew what and when.
(Independent, 27.2.2000, p. 8, col. 4)
f The question of who gives how much is of intense interest to non-proﬁt
groups as they search for sources to make up for falling government, foundation
and corporate funding. (Washington Post, 29.4.2003, p. A7, col. 3)
Using (31) as a model, let us try to rephrase the questions in (30) to make them
acceptable while retaining the two interrogative constituents. One way of “repairing”
the examples in (30) is illustrated in (32). Compare the two sets of examples.
(32) a It has become a matter of dispute what John said to whom.
b It has become a matter of dispute who said these things to whom.
c It has become a matter of dispute who said what to John.
TSC05 07/18/2005, 03:50PM326
The Periphery of the Sentence 327
Both (30) and (32) contain sentences with more than one interrogative constituent.
The difference between them is that in the acceptable examples in (32) there is only
one sentence-initial interrogative constituent, while in the unacceptable examples in
(30) there are two. One way of accounting for the unacceptability of the examples
in (30) is to assume that each projection has a unique speciﬁer position. If more
than one constituent moves to the speciﬁer of CP there will be a “collision” and, to
use a term used in recent Minimalist approaches to syntax,31
the derivation will crash.
In the acceptable examples in (31) and (32), there is one such sentence-initial whconstituent,
who, and the other wh-constituents occupy their base position.
2.8 Subject interrogatives
2.8.1 INDIRECT QUESTIONS
In fact, on closer inspection (32b) and (32c) and the attested examples in (31) raise
a question. In these examples, the sentence-initial interrogative constituent is the
subject. We may wonder which position the sentence-initial subject wh-constituent
occupies. Two hypotheses for the representation of (31a) are given in (33). Discuss
the difference between the two representations in (33).
(33) a I don’t know [CP [IP who spotted who ﬁrst]].
b I don’t know [CP who [IP who spotted who ﬁrst]].
Does a sentence-initial interrogative subject occupy its canonical position, SpecIP,
as in (33a)? Or is it in SpecCP, as in (33b)? The same question arises in a sentence
in which the only interrogative constituent is a subject. For (34a) too, we could
propose either representation (34b) or (34c):
(34) a I don’t know who spotted the thief ﬁrst.
b I don’t know [CP [IP who spotted the thief ﬁrst]].
c I don’t know [CP who [IP who spotted the thief ﬁrst]].
There is no direct evidence in the examples to help us decide. For instance, we
cannot simply determine the structure by inspecting the word order derived by the
representations above: both representations lead to a correct spell-out. There may
be some theoretical arguments in favor of the hypothesis that the interrogative
subject has moved to SpecCP (i.e. in favor of representations (33b) and (34c)). First,
recall that we propose that illocutionary force is encoded in the CP domain. In a
sentence that contains just an interrogative object NP, that NP has to be moved
to the left periphery to signal illocutionary force in SpecCP (cf. (26a)). If we fail to
move the interrogative constituent, the sentence is unacceptable (cf. (26b)). So,
31
See Chomsky (1995).
TSC05 07/18/2005, 03:50PM327
328 Chapter 5
on the assumption that interrogative force has to be signaled in the CP domain,
representations (33b) or (34c) are preferable.
We proposed that each constituent has one speciﬁer. This hypothesis ties in with
the idea that the function of the speciﬁer is to specify a unique domain of application.
If the subject who has been fronted to SpecCP in (33b) or in (34c), we predict
that no other interrogative constituents can move there; hence we correctly rule out
(33c) in which the interrogative object is also fronted. Assuming representation
(33b) there would be no room to front the object interrogative.
(33) c *I don’t know who who spotted ﬁrst.
If we adopt (33a) or (34b), we have two problems: (i) we do not encode interrogative
force in the CP domain, and (ii) we fail to explain why the object interrogative who
cannot move to the periphery of the clause (33c). (33d) would be the representation
based on (33a) with fronting of a second wh-constituent in (33a):
(33) d *I don’t know [CP who [IP who spotted who ﬁrst]].
In (33b)/(34c) movement of who to SpecCP does not lead to any reorderings in
the sentence. This type of movement is sometimes referred to as string-vacuous
movement.
2.8.2 DIRECT SUBJECT INTERROGATIVES AND INVERSION
In the preceding section we discussed the question whether a subject wh-constituent
undergoes movement to SpecCP. The examples discussed there were embedded
interrogatives. The same question arises with respect to direct subject questions.
Consider direct question (35a) and compare the three (partial)32
representations,
(35b), (35c), and (35d). Each of the three representations makes a different claim.
Discuss and compare these claims.
(35) a Which candidate will ﬁnish ﬁrst?
b [CP Which candidate [IP which candidate will ﬁnish ﬁrst]]?
c [CP Which candidate will [IP which candidate will ﬁnish ﬁrst]]?
d [CP [IP Which candidate will ﬁnish ﬁrst]]?
In (35b) the claim is that the interrogative subject which candidate has moved to the
speciﬁer of CP. In that position it speciﬁes the scope of the question. The auxiliary
will remains under I. According to (35c), the interrogative subject which candidate
has moved to SpecCP and the auxiliary will has also moved to C. The movement of
the auxiliary from I to C would be a manifestation of a phenomenon that we have
come across already: movement of an interrogative constituent to SpecCP coincides
32
We do not represent the VP-internal copy of the subject as this is not relevant here.
TSC05 07/18/2005, 03:51PM328
The Periphery of the Sentence 329
with SAI, movement of the content of I to C. According to (35c), the auxiliary
inverts with the (non-pronounced) copy of the subject in SpecIP, which candidate.
According to (35d), ﬁnally, neither subject interrogative nor auxiliary move.
Which of these representations is preferable? Again, there is no direct evidence in
the examples to help us decide. All three representations lead to the same word order.
Let us ﬁrst address the question whether the auxiliary remains in I (35b) or moves
to C (35c). In our discussion of direct questions in Chapter 133
we introduced SAI.
SAI is represented in (35c). Suppose we do opt for (35c) as our representation of a
direct question about the subject. In this example the auxiliary is ﬁrst merged in I
and moves to C. We can ask ourselves what the analysis represented by (35c) leads
us to predict for direct questions with an interrogative subject and without auxiliaries.
In order to assess the issue, let us turn to direct questions lacking an auxiliary.
Consider the derivation of the questions in (36).
(36) a Which candidate will they invite?
b Which candidate did they invite? (unstressed did)
c Which candidate did they invite? (stressed did)
In (36a) the modal auxiliary will moves from I to C. In (36b), the content of I (here
the past tense morpheme) inverts with the subject and in order to rescue the stranded
past tense inﬂection we insert do as a “last resort.” Observe that the auxiliary do
can but need not be stressed in this example. The stressed variant of do would occur
when we want to contrast the content of the question with an assumption or expectation
in the context (36c).34
(36b) shows that when a direct question lacks an aspectual or a modal auxiliary,
SAI gives rise to do insertion. This is because movement of I to C will lead to a
stranded inﬂection. Because I-to-C movement in sentences without auxiliaries leads
to do insertion, we can deduce that in direct questions where do-support is required
I-to-C movement has taken place. Speciﬁcally, if sentences with interrogative
subjects and without auxiliaries require do insertion we conclude that there has
been I-to-C movement. Conversely, if sentences with interrogative subjects and
without auxiliaries do not require do insertion we conclude that there has not been
I-to-C movement.
Let us see what happens in the interrogative sentences introduced by a subject
interrogative constituent and lacking an auxiliary. Based on (36) we create (37a).
The acceptable version of (37a) lacks do; if we insert unstressed do, the sentence
becomes unacceptable. Stressing do renders the sentence acceptable.
(37) a Which candidate invited them?
b *Which candidate did invite them? (with unstressed did)
c Which candidate did invite them? (with stressed did)
33
Section 2.3.
34
Cf. Chapter 1, Exercise 7, and Chapter 3, section 1.2.3.2.
TSC05 07/18/2005, 03:51PM329
330 Chapter 5
Suppose there were indeed movement of the content of I to C in direct questions
such as (35a). In (37), an example lacking an auxiliary, the application of
I-to-C movement would move just the inﬂectional ending of the verb (here the
past tense morpheme -ed). This should lead to do insertion, where do can be
unstressed. This prediction is not borne out: unstressed do is not present in (37b).
Only stressed do is available (37c), but as we have seen, stressed do is inserted
for emphasis and is not a reﬂex of the movement of I to C. In other words, the
absence of do-insertion in subject interrogatives suggests that they are not affected
by I-to-C movement. This means in turn that there is no SAI. We conclude that
in (35) the auxiliary will does not move to C and that representation (35b) is
preferable to (35c) because the former does not display movement of I to C while
the latter does. Apparently then, if we assume that interrogative subjects move to
SpecCP we have to admit that this movement does not trigger movement of the
content of I to C.35
(35b) is preferred to (35c). What about (35d)? The difference between (35b) and
(35d) is like that between (34c) and (34b) discussed in the preceding section. In
(35d), which matches (34b), neither the subject nor the auxiliary undergo movement.
What would be the predictions of this representation for a sentence without
an auxiliary? Are there any drawbacks to this analysis?
If we assume that the interrogative subject does not move, then the absence of
inversion and of do-insertion is as expected. But this alternative analysis also presents
a number of drawbacks, as we saw in the preceding section. First, there are some
theoretical questions. If the interrogative subject does not move to SpecCP, we
would have to explain (i) why it need not move there, unlike non-subject interrogative
constituents, and (ii) how an interrogative subject wh-constituent in SpecIP
can specify the scope of the question. Normally, the scope of a constituent question
is encoded in the speciﬁer position of C. Furthermore, assuming that the SpecCP
still remains available, we might expect that another interrogative constituent could
actually move there. If this movement triggers SAI, the result is (35e). This example
spells out as the ungrammatical (35f):
(35) e *[CP Which books [C will] [IP who [I will] [VP who buy which books]]]?
f *Which books will who buy?
2.9 A note on that-omission in English
We saw that declarative subordinate clauses are introduced by the complementizer
that. Often, the complementizer that can be absent:
35
The question arises why subject interrogatives do not trigger movement of I to C. See Rizzi
(1996) and for a recent account see Pesetsky and Torrego (2000).
TSC05 07/18/2005, 03:51PM330
The Periphery of the Sentence 331
(38) a Boyle testiﬁed that she told Malvo four times that he could be silent or see
an attorney. (Based on Washington Post, 29.4.2003, p. B1, cols 3–4)
b Boyle testiﬁed she told Malvo four times he could be silent or see an attorney.
Though there are no subordinating conjunctions in (38b), we will assume that the
CP level of the subordinate clauses is projected. This is because the embedded
clauses are declarative, and we assume that it is the CP layer that encodes illocutionary
force.
(38) c Boyle testiﬁed [CP [C] [she told Malvo four times [CP [C] [he could be silent or
see an attorney]]]].
When the C position of a ﬁnite embedded clause is not ﬁlled, it gets a “default”
declarative reading.36
3 How Far Can You Move? And How Do You Get
There?37
3.1 “Long” movement
Consider examples (39). Are they direct questions or indirect questions? How are
the arguments of the verb eat realized in (39a)? Does the direct object of eat occupy
its base position?
(39) a What does your cat eat?
b What do you think your cat eats? (Based on Observer Magazine, 7.9.2003,
p. 41, col. 1)
The object of eat in (39a) has been fronted because it is interrogative. The present
tense inﬂection is also inverted with the subject (SAI) and do is inserted to allow the
fronted inﬂection to survive. The resulting structure is as in (40a):
(40) a [CP What [C does] [IP your cat [I -s] [VP your cat [V′ eat [NP what]]]]]?
36
See also section 3.2.3.1 and Exercise 14 for some constraints on the insertion of that. The
exercise should be tackled after you’ve worked your way through the chapter.
37
In this book we can tackle only a few of the many issues raised by movement of interrogative
(and relative, see section 5 below) constituents. For a survey of a number of issues see Richards
(2001) and the references cited there.
TSC05 07/18/2005, 03:51PM331
332 Chapter 5
In (39b) the interrogative pronoun what continues to be interpreted as an argument
of the verb eat, but it precedes you, the subject of the verb think. Think selects a
clause as its complement: that your cat eats. Something is missing in this string: the
object of eat has moved out of the embedded clause. In (39a) the object of eat moves
to the SpecCP of the clause constructed around the verb eat. In (39b) the object of
eat moves to the SpecCP of the clause constructed around the verb think. In (39b)
the sentence constructed around think is a direct question: the speaker expects the
interlocutor to provide an answer ﬁlling in the speciﬁcation of what. What, the
interrogative element that narrows down the scope of the question, originates inside
the embedded clause as the object of eat. Observe that the embedded clause itself is
not a question. Using the description above, insert CP, IP, and VP brackets in (39b),
and represent the deleted copies of moved constituents by strikethrough.
(40) b Provisional representation for (39b).38
[CP What [C do] [IP you [I PRESENT] [VP you think
[CP [C] [IP your cat [I -s] [VP your cat [V′ eat-s [NP what ]]]]]]]]?
(40b) represents the result of the derivation: what occupies the periphery of the
higher clause, specifying the scope of the question. The wh-constituent what has
undergone long distance movement, or “long movement.”
Below are some more attested examples of long movement.39
Identify the moved
interrogative constituent. What is its base position? As you will see, interrogative
constituents that have undergone long movement can have a range of grammatical
functions:
(41) a It bafﬂes me as to who Tony Blair imagines will work in the universities
of the future. (Guardian, 27.11.2002, p. 9, col. 8, letter to the editor from
Hannah Cooke, Broadbottom, Cheshire)
b Who do you think is the more moderate politician? (Guardian, 25.9.2003,
p. 9, col. 6)
c What do you think was the great appeal of the Tramp? (Guardian, Review,
1.11.2003, p. 12, col. 3)
d What did they think they were making with those girls in there? (Guardian,
13.9.2003, p. 14, col. 2)
e At the end of the day everybody eats meat . . . What do you think your cat
or dog eats? Where do you think that meat comes from? Where do you
think Pedigree Chum comes from? (Observer Magazine, 7.9.2003, p. 41,
col. 1)
38
See section 2.9 for the hypothesis that even in the absence of the complementizer that the
embedded CP is projected.
39
Some of these examples were introduced in Chapter 3, Exercise 8, examples (11)–(15).
TSC05 07/18/2005, 03:52PM332
The Periphery of the Sentence 333
f Where does the Chancellor think the party is going? (Based on Guardian,
27.9.2003, p. 4, headline)
g Who does he think he is? (Guardian, 8.10.2003, p. 14, col. 4)
h Some are born great, some achieve greatness and some have greatness
thrust upon them . . . which do you think you were? (Guardian, G2,
27.10.2003, p. 6, col. 2)
i Of Pam Teare . . . Lord Hutton asked: “From what source or sources did
you think the name would leak?” (Guardian, 21.8.2003, p. 8, col. 3)
j Why do you think the Daily Mail and others have invested so much effort
in discrediting me personally? (Guardian, 30.9.2003, G2, p. 11, col. 5)
k Given current reserves and consumption, when is it predicted that the
world’s oil will run out? (Guardian, 14.1.2004, p. 15, col. 4, letter to the
editor from Jeff Lewis, Exmouth, Devon)
The moved constituent is a subject in (41a–c); it is an object in (41d). In (41e) what
is the object of the lower verb eat; where is the complement of the preposition
from in the lower clause. In (41f) where is a directional complement of the verb go;
in (41g) and (41h) an interrogative predicate is moved from the lower clause; in
(41i–k) adjuncts undergo long movement. In the examples below we represent the
position from which the underlined interrogative constituent has been moved by
dashes:
(42) a who Tony Blair imagines —— will work in the universities of the future
b Who do you think —— is the more moderate politician?
c What do you think —— was the great appeal of the Tramp?
d What did they think they were making —— with those girls in there?
e What do you think your cat or dog eats ——?
Where do you think that meat comes from ——?
Where do you think Pedigree Chum comes from ——?
f Where does he think the party is going ——?
g Who does he think he is ——?
h Which do you think you were ——?
i From what source or sources did you think the name would leak ——?
j Why do you think the Daily Mail and others have invested so much effort
in discrediting me personally ——?
k when is it predicted that the world’s oil will run out ——?
In the next section we will examine how long movement such as that illustrated
above proceeds.40
40
Exercise 2.
TSC05 07/18/2005, 03:52PM333
334 Chapter 5
3.2 Intervention effects and step-by-step movement
3.2.1 BLOCKING MOVEMENT
3.2.1.1 A hypothesis
An interrogative constituent undergoes leftward movement to the CP domain, where
it determines the scope (i.e. the domain of application) of the question. As shown by
the long movement examples (41), the interrogative constituent can cross clause
boundaries on its way to its landing site, a SpecCP position. Recall also that we
assume that the structure of the sentences is tripartite:
(43) CP > IP > VP
We assume that each projection of a head makes available a speciﬁer position.
To account for the derivation of interrogative sentences with long movement, we
must address the question of how the moved constituent proceeds. Consider our
provisional representation (40b). The arrows in the representation suggest that the
moved constituent what moves directly from its base position in the embedded
clause, in which it is the object of the verb eat, to the speciﬁer position of a CP of
the main clause, whose verb is think.
However, this is not the only derivation possible. Remember that even in the
absence of that the embedded CP is available.41
So, rather than going directly to the
speciﬁer of the CP constructed on the verb think, we could also imagine that movement
goes stepwise. In a ﬁrst step, what moves to the speciﬁer of the embedded CP
built around the verb eat. Then, in the second step, what could move to the higher
clause. The second application of movement would leave a copy which, being a
lower copy than that in the main clause SpecCP, will not be pronounced. (44) is a
representation of this alternative step-by-step derivation:
(44) [CP What [C do] [IP you [I PRESENT] [VP you think
[CP what [C] [IP your cat [I -s] [VP your cat [V′ eats [NP what]]]]]]]]?
What kind of data could be used as evidence that the interrogative constituent has
transited through the intermediate SpecCP? Could we devise any experiments to
help us decide? Recall that step-by-step movement leaves an unpronounced copy in
the intermediate landing site. How could we determine if there is a copy of the
moved constituent what in the lower SpecCP in (44)?
41
See section 2.9.
TSC05 07/18/2005, 03:52PM334
The Periphery of the Sentence 335
We postulate that a projection of a head has only one speciﬁer. In Chapter 4,42
we
examined subject movement and we proposed that the subject moves step by step,
via the available intermediate speciﬁers. If these two assumptions are generalized,
then (i) there is one speciﬁer position per CP,43
and (ii) movement of interrogative
constituents applies stepwise. From these two assumptions, we deduce that long
movement proceeds via the speciﬁer positions of intermediate CPs. The analogy
with step-by-step subject movement would be a theoretical argument in favor of
step-by-step movement of the interrogative constituent. If step-by-step movement
applies to interrogative constituents then, once one constituent occupies a SpecCP
position, we predict that this should block long movement of additional interrogative
constituents out of the relevant CP.
(45) [CP [CP wh- [CP
****
3.2.1.2 Testing the hypothesis
Our hypothesis is that movement of interrogative constituents proceeds stepwise. In
other words, no movement can skip an intervening SpecCP. The analysis predicts that
the grammar will not generate sentences in which movement of an interrogative
constituent has crossed an interrogative constituent in a SpecCP of an intermediate
sentence. Examples formed in this way should be ungrammatical and should not
occur. The attested examples of long movement in (41) are all compatible with the
hypothesis: there is no intervening interrogative constituent between the fronted
interrogative constituent and its base position. However, these and similar acceptable
examples of long movement do not prove conclusively that movement could
not have crossed an intervening interrogative constituent. The observation that we do
not ﬁnd any examples in which long movement crosses an interrogative constituent
is not conclusive evidence either. Perhaps their non-occurrence is a mere side-effect
of the kind of material we have examined. Perhaps the construction is stylistically
highly marked and therefore extremely rare though not impossible.44
We must
be sure that moving an interrogative constituent across a ﬁlled SpecCP (45) really
leads to an unacceptable output. What we can do is construct examples that use
the derivation in (45) and see if they are ungrammatical, as our hypothesis predicts.
Consider the grammatical (41d) repeated here as (46a):
(46) a What did they think they were making —— with those girls in there?
The interrogative constituent what has moved from the embedded clause to the
main clause. In principle, it can transit via the embedded SpecCP, which is empty.
In (46b) we provide only the labeled brackets relevant to our discussion.
42
Sections 3 and 4.
43
See also section 2.7 for motivation.
44
Recall the discussion of the generalization “all swans are white” in Chapter 1, section 2.3.
TSC05 07/18/2005, 03:52PM335
336 Chapter 5
(46) b [CP What did [IP they think
[CP what [IP they were making what with those girls in there]]]]?
(46b) is compatible with the hypothesis that movement cannot skip an intervening
SpecCP but it does not conﬁrm its general validity. What we need to show is that if
the constraint on stepwise movement is not respected the sentence is ungrammatical.
In order to test the prediction implied by (45) we modify (46b). We ﬁll the embedded
SpecCP with an interrogative constituent. For instance, we can turn the place adjunct
in there into the interrogative variant, where, and front it to the embedded SpecCP.
This will turn the embedded clause into an interrogative complement clause. In
(46b) the main verb, think, selects a declarative complement. To allow for an interrogative
complement clause, we have to change the verb of the main clause. We
replace think by wonder.
(46) c *[CP What did [IP they wonder [CP where [IP they were making what with those
girls where]]]]?
Though we can ﬁgure out what (46c) is intended to mean, it is not a grammatical
sentence.
Suppose that there is a “shortest step” constraint on movement: preferentially
movement must apply step by step, landing the moved constituent in intermediate
potential landing sites between its base position and its ultimate landing site. A
moved constituent will ﬁrst go to the nearest landing site and then move on to the
next one. If a moved constituent is forced to cross a potential landing site, then it
cannot proceed through the shortest steps possible and hence it is in violation of the
shortest step constraint (cf. (45)). In (46c), the speciﬁer of the CP in the embedded
clause is ﬁlled by the interrogative constituent where. It will therefore be impossible
for the interrogative object NP what to move from the embedded clause to the
speciﬁer of the CP of the main clause, because this movement would have to take a
big leap and cross the intermediate position. In (46d) we label the lower clause CP2
and the higher clause CP1. This is to make it easier to read the representation.
(46) d
*[CP1 What did [IP they wonder
[CP2 where [IP they were making what with those girls where]]]]?
*****
On the assumption that there is just one speciﬁer per CP, we account for the
degradation in (46c/d) by means of the hypothesis that wh-movement proceeds
stepwise. If an intermediate speciﬁer position is already ﬁlled, this will block movement
of another constituent. In other words, the degraded status of (46c/d) offers
indirect support for the hypothesis that all movement, including the long movement
of interrogative constituents, proceeds step by step.
The hypothesis that movement is done step by step can be seen as one implementation
of a more general property of the language system. Recall that in Chapter 4
TSC05 07/18/2005, 03:52PM336
The Periphery of the Sentence 337
we proposed that the subject has its base position in the speciﬁer of VP. This
hypothesis enabled us to deﬁne thematic relations in terms of local relations: a head
(here the verb) assigns its thematic roles in the domain that it controls, the VP. We
could say that the constraint that all movement must go step by step is another
instantiation of a general locality requirement on syntactic relations. Thematic role
assignment is local: that is, a head assigns a thematic role “locally,” within its
projection. Movement is “local”: that is, a constituent moves into a “local” SpecCP
before moving into a remote SpecCP.
Assuming step-by-step movement for the fronting of wh-constituents allows us to
account for the degradation of examples in which we try to extract an interrogative
constituent from an interrogative clause (46c/d). This evidence is indirect and hinges
on a number of theoretical assumptions. We have not provided any direct overt
evidence for the step-by-step movement hypothesis. Using indirect evidence is a
legitimate strategy in scientiﬁc research and it is used in other sciences. For instance
astronomers use it to pinpoint the existence of planets and planetary systems:
Until 1995, there was no evidence at all of planets orbiting other stars. Since the
ﬁrst dramatic discovery eight years ago, researchers have identiﬁed more than 100
planetary systems within 150 light years of Earth.
No one has seen any of these planets: researchers infer the presence of an orbiting
planet from a kind of wobble in the light from the parent star. (Guardian, 4.7.2003,
p. 7, col. 1, Tim Radford, Science Editor)
With respect to movement of interrogative constituents, the intervention effect that
arises when an interrogative constituent occupies an intermediate SpecCP and
hinders or blocks movement of another interrogative constituent (46c/d) could be
compared to the “wobble” created by a planet in the light of its parent star.
3.2.2 EMPIRICAL EVIDENCE FOR STEP-BY-STEP MOVEMENT
In addition to the indirect evidence for stepwise movement, there is also some direct
evidence. In section 2.5 of this chapter we discussed examples like those in (47), in
which movement of an interrogative constituent had stranded an element. (47a)
shows the stranding of exactly; (47b) and (47c) from West Ulster English illustrate
the stranding of the quantiﬁer all.
(47) a What do you mean exactly? (cf. (18))
b What did you get all for Christmas? (cf. (22a))
c Who did you meet all when you were in Derry? (cf. (22b))
The stranded elements, exactly and all, were taken to remain in the base positions
of the moved direct objects. In his article on wh-movement in West Ulster English,
McCloskey also discusses the examples in (48) (McCloskey, 2000: 63, note 8). Try
to provide the bracketed representations of these three examples. Can you see why
these examples bear on the current discussion?
TSC05 07/18/2005, 03:52PM337
338 Chapter 5
(48) a What exactly did he say that he wanted?
b What did he say that he wanted exactly?
c What did he say exactly that he wanted?
The data in (48) can be interpreted as offering empirical support for the hypothesis
of step-by-step movement of the interrogative constituent. In (48a) what exactly
moves to SpecCP. In (48b) we might propose that exactly is stranded in the base
position of the object, and in (48c) we might propose that exactly is stranded in the
intermediate speciﬁer position:45
(49) a [CP [NP What exactly] [C did] [IP he [I -ed] [VP he say
[CP what exactly that [IP he [I -ed] [VP he want-ed what exactly]]]]]]?
b [CP [NP What] [C did] [IP he [I -ed] [VP he say
[CP what that [IP he [I -ed] [VP he want-ed [NP what exactly]]]]]]]?
c [CP [NP What] [C did] [IP he [I -ed] [VP he say
[CP [what exactly] that [IP he [I -ed] [VP he want-ed [NP what exactly]]]]]]]?
Discuss the relevance of the West Ulster English sentences in (50) for the hypothesis
of step-by-step movement (data from McCloskey, 2000: 61, his (8)).
(50) a What all did he say (that) he wanted?
b What did he say (that) he wanted all?
c What did he say all (that) he wanted?
We can again interpret the data in (50) in line with our hypothesis. If we assume
that the quantiﬁer all is stranded after the movement of what in (50b) and in (50c),
then it is reasonable to say that in (50b) all occupies the object position. In (50c) all
could then be said to be stranded in the speciﬁer of the intermediate CP:
(51) a [CP [NP What all] [C did] [IP he [I -ed] [VP he say [CP [NP what all] (that) [IP he
[I -ed] [VP he want-ed [NP what all]]]]]]]?
b [CP [NP What] [C did] [IP he [-ed] [VP he say [CP what (that) [IP he [I -ed] [VP he
want-ed [NP what all]]]]]]]?
c [CP [NP What] [C did] [IP he [I -ed] [VP he say [CP [NP what all] (that) [IP he
[I -ed] [VP he want-ed [NP what all]]]]]]]?
3.2.3 SUBJECTS AND OBJECTS
3.2.3.1 The complementizer that
We saw that subordinate declaratives are introduced by the complementizer that.
Often, the complementizer that can be absent. We repeat the examples discussed in
section 2.9 here.46
45
There are complications, though, which are discussed in McCloskey (2000: 64, note 8).
46
For an early discussion of the data dealt with here see Perlmutter (1971).
TSC05 07/18/2005, 03:53PM338
The Periphery of the Sentence 339
(52) a Boyle testiﬁed that she told Malvo four times that he could be silent or see
an attorney. (Based on Washington Post, 29.4.2003, p. B1, cols 3–4)
b Boyle testiﬁed she told Malvo four times he could be silent or see an attorney.
In our earlier example of long extraction, the complementizer that can sometimes
be inserted in the embedded C:
(53) a What do you think your cat eats? (= (39b))
b What do you think that your cat eats?
c [CP What [C do] [IP you [I PRESENT] [VP you think
[CP what [C that] [IP your cat [I -s] [VP your cat [V′ eats [NP what]]]]]]]]?
In the attested examples of long movement in (41), the embedded CP does not contain
an overt complementizer. In some of the examples, though, the complementizer
that can be inserted; in others, on the other hand, inserting that will lead to sharp
ungrammaticality. Try inserting that in the examples in (41). Try to identify the
factor that sets apart the examples in which inserting that leads to ungrammaticality.
(54) a *It bafﬂes me as to who Tony Blair imagines that —— will work in the
universities of the future.
b *Who do you think that —— is the more moderate politician?
c *What do you think that —— was the great appeal of the Tramp?
d What did they think that they were making —— with those girls in there?
e What do you think that your cat or dog eats ——?
Where do you think that that meat comes from ——?
Where do you think that Pedigree Chum comes from ——?
f Where does he think that the party is going ——?
g Who does he think that he is ——?
h Which do you think that you were ——?
i From what source or sources did you think that the name would leak
——?
j Why do you think that the Daily Mail and others have invested so much
effort in discrediting me personally ——?
Inserting that leads to ungrammaticality in those sentences in which a subject has
been extracted from the embedded clause. We have discovered a subject/object
asymmetry: while subjects cannot be extracted across the complementizer that
(54a, b, c), there is no problem for object extraction (54d, e). It would be important
to explain this constraint on subject extraction. One possibility would be to relate
the constraint on subject extraction to the fact that the subject is extracted from a
position adjacent to the complementizer while the object is not extracted from such
a position. Alternatively, we might wish to relate the observed asymmetry to the
TSC05 07/18/2005, 03:53PM339
340 Chapter 5
fact that while the object is extracted from its VP-internal thematic position, the subject
has ﬁrst been moved from its base position, SpecVP, to SpecIP, a VP-external
position. We will not pursue this issue here.47
3.2.3.2 Blocking subject movement
Recall that it is difﬁcult to move an interrogative pronoun out of a CP whose speciﬁer
is already ﬁlled by a wh-constituent. Consider (55a), a constructed example.
What is the function of the underlined interrogative constituent?
(55) a *Which book did you wonder on which day they will publish?
Compare (55a) and (55b). Both examples in (55) are degraded, but (55b) is generally
felt to be unacceptable, while some speakers may marginally tolerate (55a).
Why should this be?
(55) b **Which book did you wonder on which day will appear?
Let us represent the derivations of both examples, representing the unpronounced
copies of the moved wh-constituents by strikethrough. In order not to overload the
representation in (56), we do not represent the unpronounced copies of the subjects
of wonder and of publish because these do not concern us here.
(56) a *[CP Which book did [IP you wonder [CP on which day [IP they will publish
which book]]]]?
b **[CP Which book did [IP you wonder [CP on which day [IP which book will
[VP which book appear]]]]]?
We discover a further subject/object asymmetry: moving an object out of a clause
whose SpecCP contains an interrogative constituent (56a) leads to some degradation,
but moving a subject out of such a clause leads to a worse degradation. Obviously,
it would be important to provide an account for this asymmetry, and preferably
one that ties in with the subject/object asymmetry that we discovered in relation to
that insertion. However, such an account would go well beyond the scope of this
introduction.
4 The Periphery of Non-Finite Clauses
So far we have only discussed the periphery of ﬁnite interrogative clauses. Let us
look at the structure of non-ﬁnite interrogative clauses. We will examine to what
47
Exercise 11. For a discussion of adjacency and anti-adjacency see Culicover (1991, 1993),
Rizzi (1997), Pesetsky and Torrego (2000). For the role of the extraction site see Rizzi (1990)
and Richards (2001) and the references cited there.
TSC05 07/18/2005, 03:53PM340
The Periphery of the Sentence 341
extent we can ﬁt non-ﬁnite interrogative clauses into the structures that we have
already elaborated for ﬁnite clauses.
4.1 For as a conjunction for non-ﬁnite clauses
Provide arguments for considering the underlined strings in the following examples
as constituents.
(57) a You realise he hates for you to call him “Ollie”, don’t you? (Marcia Muller,
Edwin of the Iron Shoes, 1993: 42)
b He did intend for Endrina to join him as a runner. (Guardian, G2,
15.2.2001, p. 4, col. 2)
The underlined strings function as the objects of the verbs hate and intend. We can
paraphrase them by means of a ﬁnite sentence:
(58) a He hates it that you call him “Ollie”.
b He did intend that Endrina should join him as a runner.
In the underlined strings in (57), the inﬂection of the clauses is non-ﬁnite, it is
realized by to.48
The subject is realized by an NP (you, Endrina). While the ﬁnite
counterparts of these clauses would be introduced by the conjunction that, the nonﬁnite
variants are introduced by for: for ﬁlls the C position in a non-ﬁnite clause.
(57) c [IP He [I -s] [VP he hate-s [CP [C for] [IP you [I to] [VP you call him “Ollie”]]]]].
4.2 Non-ﬁnite interrogative clauses
4.2.1 WHETHER
Consider (59a), a constructed example whose form is familiar from the preceding
discussion. The underlined embedded clause is interrogative, it is introduced by
whether. Using labeled bracketing, represent the structure of the example. (59b) is
an attested example. Compare the underlined segment of (59b) with that in (59a).
Identify similarities and differences between these two segments. Try to ﬁt the
underlined segment of (59b) into the structure you have devised for (59a).
(59) a The politicians are deciding whether they should sever their ties with the
arrested businessman.
b The convictions are likely to have far reaching consequences, with a number
of senior British politicians, including Lord Steel, having to decide whether
48
See Chapter 3, section 1.3.
TSC05 07/18/2005, 03:53PM341
342 Chapter 5
to sever their ties with his extensive network of companies. (Based on
Guardian, 15.11.2003, p. 8, col. 7)
(59a) and (59b) both contain an embedded interrogative clause. That in (59a) is
ﬁnite, its lexical verb is sever, the subject is they, the inﬂection is realized by the
modal auxiliary should, and the illocutionary force is signaled by whether. The verb
sever assigns an agent thematic role to the subject they. We can represent the
derivation of (59a) by (60a):
(60) a [IP The politicians [I are] [VP the politicians be [VP the politicians deciding
[CP whether [IP they [I should] [VP they sever [NP their ties with the arrested
businessman]]]]]]].49
In (59b) the embedded interrogative clause is also introduced by whether, but the
clause is non-ﬁnite: the lexical verb is sever and the inﬂectional marking is realized
by to. The non-ﬁnite variant of the embedded clause lacks an overt subject, but we
interpret the embedded clause as if there were an implicit subject. Speciﬁcally, the
verb sever will assign its agent thematic role to a non-overt NP that is interpreted
as coreferential with the main clause subject a number of senior British politicians,
including Lord Steel. We represent this non-overt NP as [NP Ø]. Let us assume that
this subject, [NP Ø], also originates in the speciﬁer of VP and moves to the speciﬁer
of the non-ﬁnite IP to allow I to fulﬁl its linking function.
(60) b [CP whether [IP [NP Ø] [I to] [VP [NP Ø] sever [NP their ties with his extensive
network of companies]]]].
The attested examples below illustrate non-ﬁnite interrogative clauses introduced
by whether.
(61) a Each European country will be free to choose whether to include records
of treatment received, photographs, and biometric data. (Guardian,
15.11.2003, p. 16, col. 4)
b The RMT executive is to meet on Monday to decide whether to call
underground-wide strikes. (Guardian, 21.11.2003, p. 6, col. 2)
c Lieberman said he was already in the State Senate when he mused about
whether to open the ice cream shop at a “perfect” site on the Yale campus
49
We assume that they moves step by step from the SpecVP headed by sever to SpecIP.
Similarly, the subject NP the politicians moves step by step from the SpecVP associated with
deciding to the speciﬁer of IP. There is an unpronounced copy of the moved subject in the
speciﬁer position associated with the projection of the auxiliary be. See Chapter 4, section 4 for
discussion.
TSC05 07/18/2005, 03:53PM342
The Periphery of the Sentence 343
or continue to practice law. (Atlanta Journal-Constitution, 23.11.2003,
p. A4, col. 5)
In non-ﬁnite interrogatives whether cannot be replaced by if.50
4.2.2 FRONTED INTERROGATIVE CONSTITUENTS
Consider (62a). In this example the embedded clause is interrogative and is introduced
by the fronted interrogative pronoun who. Who originates in the complement position
of the preposition on. Using labeled bracketing, represent the structure of this
example. Indicate non-pronounced copies of moved constituents by strikethrough.
(62) a They don’t know who they should rely on.
(62b) is an attested example. How does it differ from (62a)? Try ﬁtting the components
of (62b) into your representation for (62a).
(62) b They are panicking because they don’t know who to rely on. (Washington
Post, 29.4.2003, p. A18, col. 5)
For (62a) the representation will be as in (63a):
(63) a [IP They [I don’t] [VP they know [CP who [IP they [I should] [VP they rely [PP on
[who]]]]]]].
If we try to ﬁt (62b) into this structure we end up with the following partial
representation:
(63) b (First approximation)
[IP They [I don’t] [VP they know [CP who [IP —— [I to] [VP rely [PP on [who]]]]]]].
We can again ﬁt the constituents of the non-ﬁnite clause in (62b) fairly easily into
the representation designed for its ﬁnite counterpart (62a). The only problem is that
the subject of the inﬁnitival clause in (62b) is non-overt. Again, because the subject
of the embedded clause in (62b) remains implicit, we represent it by the symbol
[NP Ø]. Again, we assume that this non-overt subject starts out as the speciﬁer of the
verb and that it moves to the speciﬁer of IP to provide a linking element for the I′
constituent. The non-overt subject of the inﬁnitive is interpreted as coreferential
with the subject of the main clause, i.e. they.
(63) c [IP They [I don’t] [VP they know [CP who [IP [NP Ø] [I to] [VP [NP Ø] rely [PP on
[who]]]]]]].
50
See also Exercise 3 for some problems with the syntax of whether.
TSC05 07/18/2005, 03:53PM343
344 Chapter 5
5 Relative Clauses: An Introduction
5.1 Movement of the relative pronoun
Section 3 dealt, among other things, with constituent questions or wh-questions
such as those in (64).51
(64) a Who had he met at the party?
b I wonder [CP who he had met at the party].
We postulated that these sentences are derived by moving an interrogative constituent
to SpecCP in order to signal the scope of the question. In addition, the
fronted wh-constituent has a grammatical function inside the clause. Both in the
direct question (64a) and in the indirect question (64b), who is the direct object
of met. Following the argumentation elaborated above the representation of the
structure of the embedded interrogative clause in (64b) is as in (64c).
(64) c [CP who [IP he [I had] [VP he have [VP he met who at the party]]]]52
Now consider (65a):
(65) a I interviewed [NP the man who he had met at the party].
The direct object of the verb interviewed is the NP the man who he had met at the
party. We can replace this string by the pronoun him; the constituent can be the
focus of a cleft sentence.
(65) b I interviewed him.
c It is the man who he had met at the party that I interviewed.
The head of the NP is a noun, man. This noun is preceded by a determiner (the) and
it is followed by a string, who he had met at the party, whose function is to narrow
down the reference of the noun phrase. The string who he had met at the party
helps to identify which entity with the property ‘man’ we are talking about. The
embedded clause tells us ‘which man I interviewed’.
51
The aim of section 5 is to offer a ﬁrst overview of the derivation of relative clauses. The
section offers the “traditional” analysis. For a more comprehensive introductory discussion see
Haegeman and Guéron (1999). For a more recent and very inﬂuential analysis see also Kayne
(1994: chapter 8).
52
Recall that we assume that there is an unpronounced copy of the moved subject in the speciﬁer
position associated with the projection of the auxiliary have. See Chapter 4, section 4, for
discussion.
TSC05 07/18/2005, 03:54PM344
The Periphery of the Sentence 345
The string who he had met at the party is itself a ﬁnite clause: it contains a subject
(he), a ﬁnite lexical verb (met), and the I position is ﬁlled by the auxiliary had. The
pronoun who precedes the subject and we will assume that once again it occupies
the speciﬁer of CP.
(65) d [CP who [IP he [I had] [VP he have [VP he met who at the party]]]].
There are a number of formal parallelisms between the embedded clause in (64b)
and that in (65a). Both in (64b) and (65a) the embedded clause starts with a whconstituent
(who). In both examples, this wh-phrase precedes the subject NP of the
embedded clause. In both clauses the wh-constituent is the direct object of the verb
met. We assume that in both cases the wh-constituent occupies the speciﬁer of CP.
In our earlier example (64b), the embedded clause was interrogative. It was the
complement of the V wonder, which selects an interrogative clause as its complement.
The status of the embedded clause in (65a) is different: the clause is not the
complement of a verb; rather it serves to modify a head N (here man). The embedded
clause in (65a) is a relative clause. In (65a), the pronoun who narrows down the
reference of the N man. When used to introduce a relative clause, who is called a
relative pronoun. In its use as a relative pronoun, who indicates the type of the
embedded clause (here relative), and it also has a grammatical function inside that
clause. The noun phrase to which a relative pronoun refers is called its antecedent.
For human antecedents we use the relative pronoun who, for non-human antecedents
we use the relative pronoun which:
(66) a I like the book [CP which you gave me]. which = object
b They reviewed the books [CP which arrived ﬁrst]. which = subject
Observe that many wh-pronouns double up as relative pronouns and as interrogative
pronouns, and the clauses they introduce will accordingly be either relative
clauses or interrogative clauses. In (67) we give a few examples.53
(67) a I wonder [CP who told him about Bill’s departure]. Interrogative
This is [NP the person [CP who told him about Relative
Bill’s departure]].
b I wonder [CP which I should buy him: Interrogative
the green sweater or the blue one].
He did not like [NP the sweater [CP which I bought him]]. Relative
c I asked them [CP whose car they had used]. Interrogative
[NP The woman [CP whose car they had used]] will be paid. Relative
The moved constituent whose car in (67c) contains both the possessive pronoun
whose and the noun car. Whose is the speciﬁer of the NP whose car; it denotes the
53
Exercise 9. For relative clauses introduced by that see Exercise 14. This exercise should be
tackled after you have ﬁnished section 5.
TSC05 07/18/2005, 03:54PM345
346 Chapter 5
possessor of the entity that the NP refers to. If we want to question a possessor of
the entity referred to by an NP, we apparently cannot just move an interrogative
possessor on its own (68a), but we have to move the containing NP along. We say
that we move along or we pied-pipe the NP.
(68) a *I asked them [CP whose they had used whose car].
Analogously, when forming a relative clause for which the antecedent corresponds
to a possessive pronoun, we must pied-pipe the NP:
(68) b *[NP The woman [CP whose they had used whose car]] will be paid.
In the same way that an interrogative pronoun can be moved out of its own clause
and function as the marker of interrogative force in a higher clause, a relative
pronoun can move out of its own clause. (69a) is a constructed example, (69b) is an
attested example.
(69) a I know the candidates [CP who [IP they said [CP that [IP they will nominate
this year]]]].
b A Pakistani jeweler said today that his picture is among those of ﬁve men
[CP who the F.B.I. says [CP [IP may have entered the United States on doctored
passports]]]. (New York Times, 2.1.2003, p. A9, col. 2)
5.2 Constraints on movement: Some predictions
Our hypothesis is that relative clauses are derived essentially along the lines of
interrogative clauses. In particular, we assume that relative pronouns undergo movement
similar to the movement of interrogative pronouns. This hypothesis leads us
to a number of predictions. Given that movement of an interrogative pronoun is
made difﬁcult by certain factors, the same factors should also lead to a degradation
for the movement of a relative pronoun. In section 3.2.1 we introduced the shortest
step constraint on movement. Long movement of an interrogative pronoun was
shown to lead to ungrammaticality if the speciﬁer of an intervening CP was already
occupied by a wh-constituent. We predict that, in the same way, the movement of a
relative pronoun will degrade if it crosses a ﬁlled speciﬁer of an intervening CP.
In section 3.2.3, we saw that the presence of the complementizer that in the
position C blocks the movement of an adjacent interrogative subject. We predict
that the presence of that should also block extraction of an adjacent relative subject.
We will examine these two predictions here.
5.2.1 INTERVENTION EFFECTS ON THE MOVEMENT OF THE
RELATIVE PRONOUN
When we apply long movement in interrogative clauses, the moved wh-constituent
moves step by step via the intermediate speciﬁers of CP. Coupled with the assumption
TSC05 07/18/2005, 03:54PM346
The Periphery of the Sentence 347
that each CP has only one speciﬁer, this allowed us to predict the degraded status of
(55a) repeated here as (70):
(70) *[CP Which book did [IP you wonder [CP on which day [IP they will publish]]]]?
If the derivation of relative clauses proceeds like that of interrogative clauses, a relative
pronoun also moves stepwise. We predict that relative fronting out of an interrogative
clause will lead to a degradation. Observe that once again attested examples
which are compatible with a derivation using stepwise movement do not constitute
conclusive evidence for our hypothesis. What we have to show is that examples in
which movement cannot proceed stepwise are ungrammatical. For instance, relativization
out of an interrogative clause should not be grammatical. Again, we cannot
test our hypothesis by looking for examples to conﬁrm this. By deﬁnition, being
ungrammatical, the relevant examples should not occur. It is not because we don’t
ﬁnd any examples of relativization out of an interrogative clause that we can
conclude that such examples do not exist.54
To test our hypothesis we can run an experiment and create the very conditions
that should lead to the degradation. We construct a relative clause in which the relative
pronoun undergoes long movement. In other words, the pronoun starts from the
embedded clause and moves up to a higher clause. If we then insert an interrogative
constituent in the speciﬁer of an intermediate CP, we should ﬁnd a degradation.
Starting from (69a) above, repeated here as (71a) let us test our prediction.
(71) a I know the candidates [CP who [IP they said [CP that [IP they will nominate
who this year]]]].
To construct the decisive kind of example, we have to modify the declarative
embedded clause in (71a), and turn it into an interrogative clause. However, the verb
said in (71a) does not select an interrogative complement clause. In order to allow
for an interrogative complement clause, we have to replace the verb said by a verb
selecting an interrogative clause. Let us replace said by wondered. Within the complement
clause we replace the temporal speciﬁcation this year by a wh-constituent,
which we move to the speciﬁer of the embedded CP. (71b) is a partial representation:
(71) b *I know the candidates [CP who [IP they wondered [CP when [IP they will
nominate who]]]].
As predicted, (71b) is not acceptable: the interrogative constituent when in the
speciﬁer of lower CP blocks the passage of who to the higher speciﬁer position.
These data support the hypothesis that both relative pronouns and interrogative
pronouns move step by step.55
54
See section 3.2.1.2.
55
Exercise 10A. Exercises 12 and 13 introduce another pattern which blocks movement.
TSC05 07/18/2005, 03:54PM347
348 Chapter 5
5.2.2 SUBJECT MOVEMENT AND THE COMPLEMENTIZER THAT
We discovered56
a subject-object asymmetry with respect to the ﬁller of the C position
in embedded clauses from which a constituent is moved. When we move an
interrogative subject out of an embedded clause, we cannot insert the conjunction
that in the C position adjacent to the non-pronounced copy of the subject. This was
illustrated by examples (54a, b, c) repeated here as (72).
(72) a *It bafﬂes me as to who Tony Blair imagines that will work in the universities
of the future.
b *Who do you think that is the more moderate politician?
c *What do you think that was the great appeal of the Tramp?
If relative pronouns move in a way similar to interrogative pronouns, then they
ought to display a similar subject/object asymmetry. In (69a), repeated here for
convenience as (73a) the object is extracted.
(73) a I know the candidates [CP who [IP they said [CP that [IP they will nominate
this year]]]].
The complementizer that is realized in the embedded clause, from which the
direct object who has been extracted. The complementizer may also remain
unexpressed:
(73) b I know the candidates [who they said [they will nominate this year]].
(69b), repeated here for convenience as (74a), is an illustration of long movement of
a subject relative pronoun. In the attested example, the complementizer that is not
realized in the C position of the clause from which the subject who has been moved.
Indeed, the complementizer that cannot be inserted in the lower clause (74b):
(74) a A Pakistani jeweler said today that his picture is among those of ﬁve men
[CP who the F.B.I. says [CP [IP may have entered the United States on doctored
passports]]]. (New York Times, 2.1.2003, p. A9, col. 2)
b *A Pakistani jeweler said today that his picture is among those of ﬁve men
[who the F.B.I. says [that may have entered the United States on doctored
passports]].
The contrast between object extraction in (73) and subject extraction in (74)
can be analyzed as another effect of the subject-object asymmetry in moving
wh-constituents.
56
Section 3.2.3.1.
TSC05 07/18/2005, 03:54PM348
The Periphery of the Sentence 349
5.3 A non-overt relative pronoun
5.3.1 THE DATA: RELATIVE CLAUSES WITHOUT A RELATIVE
PRONOUN?
Consider once again the derivation of the relative clause in (74a). What is the subject
of says? What is the subject of may have entered . . . passports? As you can see,
the relative pronoun who has undergone long movement. In representation (74c)
we indicate all the unpronounced copies of the relative pronoun by strikethrough
(who):57
(74) c ﬁve men [CP who [IP the F.B.I. says [CP who [IP who may [VP who have [VP
who entered the United States on doctored passports]]]]]].
Compare the underlined string in (75a) with the relative clause in (74a). What is
the subject of believe? What is the subject of may have entered the United States
illegally from Canada?
(75) a F.B.I. agents investigating falsiﬁed identity papers are expanding their dragnet
for a growing list of foreign-born men they believe may have entered
the Unites States illegally from Canada.
At ﬁrst sight it looks as if the subject of may have entered . . . is missing. Using the
constructed example (73a) as a model, we could insert a subject relative pronoun:
(75) b F.B.I. agents investigating falsiﬁed identity papers are expanding their
dragnet for a growing list of foreign-born men who they believe may have
entered the Unites States illegally from Canada.
In (75c) we represent all the copies of the relative pronoun who in (75b):
(75) c foreign-born men [CP who [IP they believe [CP who [IP who may [VP who have
[VP who entered the Unites States illegally from Canada]]]]]].
If the verb enter assigns a thematic role to who in (75b), then in (75a) enter will also
have to assign this thematic role. It is proposed that in (75a) a relative pronoun has
also moved to SpecCP but that this pronoun is itself not pronounced. In other
words (75a) would have the representation in (75d):
(75) d foreign-born men [CP who [IP they believe [CP who [IP who may [VP who have
[VP who entered the Unites States illegally from Canada]]]]]].
57
Recall that we assume that there is an unpronounced copy of the moved subject in the speciﬁer
position associated with the projection of the auxiliary have. See Chapter 4, section 4, for
discussion.
TSC05 07/18/2005, 03:55PM349
350 Chapter 5
When a relative pronoun is left unpronounced, we can refer to it as a non-overt or
null pronoun. In (75d) the non-overt pronoun undergoes movement in the same
way as its overt counterpart in (74c).58
5.3.2 EVIDENCE FOR MOVEMENT OF A NON-OVERT PRONOUN
We proposed that in (75a) the relative clause which modiﬁes foreign-born men is
introduced by a non-overt relative pronoun (who). As shown in (75d), we assume
that the non-overt pronoun starts out as the subject of entered and undergoes stepby-step
(long) movement to SpecCP. What kind of evidence could we provide to
support this analysis?
Recall that movement of wh-constituents is subject to a number of constraints. In
section 3.2.1 we introduced the shortest step constraint on movement: long movement
of an interrogative pronoun is blocked if a speciﬁer of an intervening CP is
already occupied by a wh-constituent. In section 3.2.3, we saw that long movement
of a subject wh-constituent is blocked if the adjacent C position is ﬁlled by that. If
(75a) is derived by long movement of a non-overt relative pronoun, then we predict
that both the effect of shortest step constraint on movement and that of the constraint
on the extraction of the subject should be manifested here. Let us examine
each of these points in turn.
5.3.2.1 Step-by-step movement
When we apply long movement in interrogative clauses, the moved wh-constituent
moves stepwise via the speciﬁer(s) of the intermediate CP(s). Coupled with the
assumption that each CP has only one speciﬁer, this allowed us to predict the
ungrammaticality of (71b) repeated here as (76):
(76) *I know the candidates [CP who [IP they wondered [CP when [IP they will
nominate who]]]].
(76) is not acceptable because the interrogative constituent when in the speciﬁer of
the lower CP blocks the transit of the relative pronoun who to the speciﬁer position
of the higher clause.
If relative clauses lacking a relative pronoun are derived by movement of a nonovert
pronoun, we predict similar intervention effects. Non-overt pronouns should
always move stepwise. In other words, there should not be any relative clauses in
58
In fact, there are two ways of looking at this. One option is to say that in examples without an
overt relative pronoun the pronoun is an abstract entity: it is merged as a non-overt element.
This is what is traditionally meant by a term such as “non-overt pronoun.” We might, however,
also say that in relative clauses without an overt pronoun, a genuine pronoun is merged
and then moved but that all copies, including the highest one, end up not being pronounced.
The difference between these two ways of thinking is subtle and relates to theoretical assumptions.
It is not clear that the two analyses would make different empirical predictions.
TSC05 07/18/2005, 03:55PM350
The Periphery of the Sentence 351
which a lower SpecCP inside the relative clause is ﬁlled by an interrogative constituent.
Our hypothesis predicts that certain patterns should not occur. As before,
the fact that we do not actually come across the relevant examples as such is not
conclusive.59
However, we can test our prediction by means of an experiment. By
inserting a wh-constituent in the speciﬁer position of an intermediate CP in (75a),
we will create a blockade for the movement of the non-overt pronoun who. As in
earlier similar experiments,60
we also must make sure that the embedded interrogative
CP can be the complement of the verb it is merged with. (77a) is a constructed
example of this type. We have replaced believe by wonder, and we have moved the
manner adjunct how to the speciﬁer of the CP embedded under wonder. If the
relative clause is indeed the result of movement of a non-overt pronoun, we predict
that there should be a degradation in grammaticality. Because we are moving a
subject we expect the degradation to be severe.61
The prediction is correct:
(77) a *F.B.I. agents investigating falsiﬁed identity papers are expanding their
dragnet for a growing list of foreign-born men they wonder how may have
entered the Unites States from Canada.
In (77b) we indicate the path of the illicit movement of the non-overt relative
pronoun by the strikethrough notation:62
(77) b *foreign-born men [CP who [IP they wonder [CP how [IP who may [VP who
have [VP who entered the Unites States from Canada how]]]]]].
5.3.2.2 Subject movement and the complementizer that
In sections 3.2.3 and 5.2.2 we discussed a subject-object asymmetry in relation to
the realization of the C position in embedded clauses from which a constituent is
extracted. When we move a subject interrogative pronoun or a subject relative
pronoun out of an embedded clause, we cannot insert that in the C position adjacent
to the non-pronounced copy of the moved subject. For interrogative pronouns this
was illustrated in examples (54a, b, c) repeated in (78). For relative pronouns this
was shown in (74b), repeated here as (78d).
(78) a *It bafﬂes me as to who Tony Blair imagines that will work in the universities
of the future.
b *Who do you think that is the more moderate politician?
c *What do you think that was the great appeal of the Tramp?
59
Again we encounter the “white swan problem.” Chapter 1, section 2.3, and this chapter,
section 3.2.1.2.
60
Sections 3.2.1.2 and 5.2.1.
61
See section 3.2.3.2.
62
Exercise 10 (A and B).
TSC05 07/18/2005, 03:55PM351
352 Chapter 5
d *A Pakistani jeweler said today that his picture is among those of ﬁve men
who the F.B.I. says that may have entered the United States on doctored
passports.
If, as we claim, the underlined relative clause in (75a), repeated here as (79a), is
derived by movement of a non-overt relative pronoun who, itself the subject of an
embedded clause, then we predict that insertion of the conjunction that should lead
to ungrammaticality. As shown by (79b) this prediction is correct:
(79) a F.B.I. agents investigating falsiﬁed identity papers are expanding their
dragnet for a growing list of foreign-born men they believe may have entered
the Unites States illegally from Canada. (New York Times, 2.1.2003, p. A9,
col. 2)
b *F.B.I. agents investigating falsiﬁed identity papers are expanding their
dragnet for a growing list of foreign-born men they believe that may have
entered the Unites States illegally from Canada.
(79b) is degraded in the same way that (78d) is degraded, conﬁrming the non-overt
relative pronoun hypothesis.63
6 Summary
This chapter completes the overview of the structure of the sentence. We returned
to the derivation of interrogative sentences discussed in Chapter 1, focusing on the
functional domain of the sentence in which illocutionary force is encoded. We also
examined how a fully formed sentence can be integrated into a larger structure and
become an embedded clause. In our discussion of the sentence we have postulated
another layer of functional structure, CP. We have also reﬁned the application of
Move.
The main body of the chapter focused on the derivation of interrogative clauses.
We assumed that a functional head C encodes illocutionary force. In embedded
clauses C hosts the conjunction (here called complementizer). In direct questions,
C hosts the fronted auxiliary of SAI. The speciﬁer of CP hosts the fronted interrogative
constituent which serves to deﬁne the scope of a constituent question.
The operation Move plays an important role in the derivation of questions. A
distinction is made between short movement, in which a constituent moves to the
speciﬁer position of the CP in whose VP it is merged, and long movement, in which
a constituent moves out of the CP in which it has ﬁrst been merged, and lands in the
speciﬁer of a higher CP. In the case of long movement, we have observed intervention
63
Exercise 14. For more extensive introductory discussion of the analysis of relative clauses see
also Haegeman and Guéron (1999, chapter 2, section 1.2).
TSC05 07/18/2005, 03:55PM352
The Periphery of the Sentence 353
effects. A moved wh-constituent cannot cross a wh-constituent in the speciﬁer of a
SpecCP located between the base position of the moved constituent and its landing
site. We have also discovered a subject-object asymmetry with respect to long movement.
When a subject undergoes long movement, the C-position that is left-adjacent
to its copy must not be ﬁlled by that. No such constraints hold for long movement
of an object.
Like interrogative clauses, relative clauses implicate the CP area, the area of the
sentence to the left of the subject. The mechanisms elaborated for the derivation of
interrogative clauses can be extended to derive relative clauses. The constraints on
movement of interrogative elements carry over to movement of relative elements.
We have also discovered that while the highest copy of a moved interrogative
constituent is always overt, all copies of the relative pronoun, including the highest
copy, may be non-pronounced. This can be interpreted to mean that a non-overt
pronoun is moved. Evidence for postulating such a non-overt pronoun is the fact
that relative clauses lacking an overt relative pronoun are subject to the same constraints
on movement as relative clauses introduced by an overt relative pronoun.
In the course of the discussion we have repeatedly had to rely on constructed
sentences to test our predictions. This was because the predictions that we wanted
to test concerned negative generalizations. If all movement proceeds stepwise, for
instance, then we predict that movement should never skip an intervening SpecCP.
In order to test such a “negative prediction” we cannot simply base ourselves on the
observation that the corresponding sentences do not actually occur. We have to
construct the relevant sentences ourselves to examine their status.
Needless to say, although this is the ﬁnal chapter of this introductory book, we
have not provided an exhaustive and deﬁnitive theory of syntax. The aim of the
book was to show how to think about syntax. The book tries to show how syntactic
research is done. It also presents a survey of some of the results that have been
formulated over the years using this methodology. Many points of syntax have not
been discussed at all, and for many others, the discussion is very partial and tentative;
this is also because there is still a lot of ongoing debate about the best way to
analyze the constructions. However, even if it had been feasible to provide a full
survey of current syntactic theory, we would still not have been able to claim that
this book is the deﬁnitive version of a theory of syntax. Research into the structure
of language continues and continuously brings with it novel discoveries and theoretical
innovations. After all, this is what science is all about:
In any branch of science there are only two possibilities. There is either nothing left to
discover, in which case why work on it, or there are big discoveries yet to be made, in
which case what the scientists say now is likely to be false. (Nigel Calder, author of
Magic Universe: The Oxford Guide to Modern Science. Cited in Guardian, 3.6.2004,
p. 6, col. 2)
TSC05 07/18/2005, 03:55PM353
354 Chapter 5
Exercises
Contents
Exercise 1 SpecCP or C (T) 354
Exercise 2 Fronting operations (T) 355
Exercise 3 The status of whether (T) 356
Exercise 4 Exclamatives (T) 357
Exercise 5 Complementizers and inversion (T, E) 357
Exercise 6 Complementizers and negative inversion (T, E) 358
Exercise 7 Non-adjacent inversion: A problem (T, E) 359
Exercise 8 Problems in the left periphery (T, E) 360
Exercise 9 Wh-movement (T) 360
Exercise 10 Constraints on extraction (T, E) 362
Exercise 11 Relative clauses and resumptive pronouns (T, E) 363
Exercise 12 Extraction from adjunct clauses and resumptive pronouns (E) 364
Exercise 13 Movement from adjunct clauses (E) 365
Exercise 14 That relatives (T, E) 366
Exercise 15 Postverbal subjects and inversion in French (E) 367
Exercise 1 SpecCP or C (T)
Consider the following sentences, focusing on the underlined embedded clause:
(1) a I wonder what they will buy.
b I wonder when they will come.
At ﬁrst glance, it might not be obvious whether the words what and when occupy
the speciﬁer position (c) or the head position (d) of the embedded CP.
c dCP
C′Spec
C IP
what
when
they
they
. . .
. . .
CP
C′Spec
C IP
what
when
they
they
. . .
. . .
TSC05 07/18/2005, 03:55PM354
The Periphery of the Sentence 355
What do you think is the appropriate analysis? How could the following examples
help us choose between representations (c) and (d)?
(2) a I wonder which book they will buy.
b I wonder on which day they will come.
(3) a What will they buy?
b When will they come?
(4) a %I wonder what will they buy. (OK in Northern Hiberno English)
b %I wonder when will they come. (OK in Northern Hiberno English)
Draw the complete tree diagram representation for the underlined string in (1a) and
describe the way the sentence is derived.
Exercise 2 Fronting operations (T)
In the following examples the underlined constituent has been fronted. Locate its
base position. What is its function? Has it undergone short movement or long
movement?
(1) While a probationary ofﬁcer, he said a more senior ofﬁcer had asked him: “Is
it in your religion to lie?” (Guardian, 29.10.2003, p. 8, col. 1)
(2) I’ve had plenty of advice over what I should say in this speech. Some of it I
have even asked for. (Guardian, 1.10.2003, p. 6, col. 5)
(3) A stunningly beautiful building, La Fenice certainly is. (Guardian, 6.12.2003,
p. 3, col. 1)
(4) “They must talk about it, and talk about it they must,” he said. Food for
thought, there! It’s a phrase that could add a measure of gravity to any press
conference. “We must do this, and do this we must.” (Guardian, 29.1.2003,
p. 2, col. 5)
(5) It’s unbelievable how unlucky he’s been, but he’s certainly proved he’s got
tenacity. Whether he’ll get out of it or not, only time will tell. (Guardian,
8.2.2003, p. 2, col. 8)
(6) Mr Blair’s point was that everything the British had asked for in Greece they
had got. (Guardian, 7.7.2003, p. 2, col. 5)
(7) How long he spent there, she couldn’t say. (Ian Rankin, The Falls, 2001: 328)
Exercises 355
TSC05 07/18/2005, 03:55PM355
356 Chapter 5
1
See Exercise 1 for evidence.
(8) If Tony Blair had listened to us, then I don’t think he would have been in the
mess he’s found himself in and if he starts to listen now, he could lead us
into a great third term. (Adapted from Guardian, 27.9.2003, p. 5, col. 6)
(9) By the next election, I intend that we will offer a really fresh alternative to
the other two parties. (Guardian, 24.9.2003, p. 16, col. 7)
(10) Kerry said he is running to “restore people’s trust that what we say we
mean.” (USA Today, 26.3.2004, p. A4, col. 5)
Exercise 3 The status of whether (T)
We have been assuming that the words that, if, whether, and for are conjunctions.
Discuss the type of clause these conjunctions introduce. Which position do conjunctions
occupy in our structures? Draw a tree diagram for the following sentences:
(1) I wonder whether I should marry this man.
(2) I wonder whom I should marry.
(3) I wonder whether to marry this man.
(4) I wonder whom to marry.
(5) I wonder how I can establish a family.
(6) I wonder how to establish a family.
COMMENT
In your representations, you will probably have inserted whether as a head under C
and you will have inserted the interrogative constituents whom and how under
SpecCP. Such a representation means that we assign whether to the class of heads,
while whom and how are maximal projections. What motivates treating whom and
how as maximal projections?1
Discuss the problems raised for this analysis by the
following attested example:
(7) Whether and whom to marry, how to express sexual intimacy, and whether
and how to establish a family – these are among the most basic of every
individual’s liberty and due process rights. (Guardian, 19.11.2003, p. 2, col. 3)
TSC05 07/18/2005, 03:55PM356
The Periphery of the Sentence 357
Exercise 4 Exclamatives (T)
Consider the following examples. What is the function of the underlined constituent?
What kind of clause does it introduce?
(1) What a picture of doom and gloom you paint. (Guardian, 26.4.2003, p. 10,
col. 7)
(2) What a player Heskey would be if he had Rooney’s conﬁdence. (Guardian,
13.3.2003, p. 15, col. 4)
(3) What a good memory you’ve got. (Muriel Spark, The Bachelors, 1963: 170)
(4) He’s shown what a genius he is again. Those ﬂashes show what a talent he is.
(based on Guardian, Sport, 7.4.2003, p. 3, col. 5)
(5) How badly money is wasted in education. (Guardian, 5.3.2002, p. 5, col. 6)
COMMENT
The fronted constituents in the above examples introduce exclamative clauses.2
Exercise 5 Complementizers and inversion (T, E)
Discuss the contrasts in grammaticality between the following sentences.
(1) a If your back-supporting muscles should tire, you will be at increased risk
of lower-back pain. (based on Independent on Sunday, Sports, 14.10.2001,
p. 29, col. 3)
b Should your back-supporting muscles tire, you will be at increased risk of
lower-back pain.
c *If should your back-supporting muscles tire, you will be at increased risk
of lower-back pain.
d *Should if your back-supporting muscles tire, you will be at increased risk
of lower-back pain.
In section 2.3 of the chapter we account for the complementary distribution of complementizers
and inverted auxiliaries by assuming that they target the same position,
2
For a detailed discussion of the syntax and semantics of exclamative sentences see Zanuttini
and Portner (2003).
Exercises 357
TSC05 07/18/2005, 03:56PM357
358 Chapter 5
C. In that section we use conditional clauses rather than embedded interrogative
clauses for the discussion. This is because on the one hand, we need embedded
clauses to test the distribution of conjunctions and on the other hand, SAI is not
possible in embedded interrogatives in Standard English.
Hiberno English does allow for embedded inversion. Consider the following data
from Henry (1995: 107, her (25)). Do they corroborate our analysis?3
(2) a They couldn’t work out whether we had left.
b They couldn’t work out if we had left.
c %They couldn’t work out had we left.
d %*They couldn’t work out whether had we left.
e %*They couldn’t work out if had we left.
Exercise 6 Complementizers and negative inversion
(T, E)
In Chapter 1, section 2.3.2, we discussed negative inversion. This is the phenomenon
whereby SAI is triggered by a negative constituent. The following examples,
correspond to (34) in Chapter 1. Identify the inverted auxiliary; identify the negative
constituent which triggers inversion.
(1) a Not one word of evidence have they brought to support that. (Guardian,
11.12.2001, p. 4, col. 7)
b Within a year of Hague becoming leader, the party had a ballot of its
membership to say that not within the lifetime of this parliament would
Britain enter the Euro. (Guardian, G2, 13.5.2002, p. 7, col. 2)
Consider the examples in (2): in (2a), the fronted constituent contains a negative
NP no account and it leads to inversion, in (2b), the fronted constituent also contains
a negative NP no time and yet there is no inversion. Can you see why there
should be this difference?
(2) a On no account should you talk to her.
b In no time she had ﬁnished her homework.
COMMENT
Though in both examples the fronted constituent is a PP containing a negative
element, the negative component serves to negate the clause only in (2a). Sentential
3
Henry (1995) offers an accessible introduction to some of the properties of Hiberno English.
See also Dufﬁeld (1993).
TSC05 07/18/2005, 03:56PM358
The Periphery of the Sentence 359
negation bears on the link subject–VP, hence on I. When the fronted negation has
sentential scope it attracts I.
In (2b) the negation expressed by no time does not affect the clause as a whole:
in no time does not negate the sentence. (2b) does not mean that she has not
ﬁnished her homework. What (2b) means is that ‘she had ﬁnished her homework’,
and that ‘ﬁnishing her work took very little time’. Since negation does not have
sentential scope it does not interact with I. A negation marker such as no in (2b)
whose scope is restricted to the containing constituent is sometimes said to express
constituent negation.
Compare the interpretation of the sentence-initial negative constituents in the
following pairs:4
(3) a With no job, Mary would be happy.
b With no job would Mary be happy.
(4) a With no clothes does Robin look attractive.
b With no clothes, Robin looks attractive.
Exercise 7 Non-adjacent inversion: A problem (T, E)
Discuss the problems that the following examples raise for the structure of the CP
that we have been postulating:
(1) Why for the ﬁrst two years of government did they do absolutely nothing?
(Guardian, G2, 23.1.2001, p. 13, col. 4)
(2) And why in Paris did the Americans modify the agreement at the last minute
with the purpose of gaining the signature of the KLA and avoiding that of
Yugoslavia? (Guardian, 13.4.1999, p. 4, col. 2)
(3) Why after the chaos on the railways and the near collapse of British Telecom
does he believe that private management will improve the efﬁciency of the
health service? (Guardian, 21.5.2001, p. 11, col. 7)
COMMENT
If we assume that a fronted interrogative constituent occupies a speciﬁer position
(i.e. SpecCP) and that an inverted auxiliary moves to the head associated with that
position (i.e. C), we predict that the fronted constituent and the inverted head will
be adjacent. Because there is no position in between the speciﬁer and the head,
there cannot be any intervening constituent. This prediction is contradicted by the
4
For discussion see Haegeman (2000a) and the references cited there.
Exercises 359
TSC05 07/18/2005, 03:56PM359
360 Chapter 5
examples above in which the fronted wh-constituent is separated from the inverted
auxiliary by an intervening constituent.5
Exercise 8 Problems in the left periphery (T, E)
Consider the following examples. Discuss which problems, if any, they pose for the
clause structure we have been elaborating so far.
You should discuss the examples individually and also try to discover more general
patterns that are shared by several examples. For your answer, you should classify
the examples according to the type of problem(s) they raise. Observe that the fact
that problems arise does not mean that the theory is to be rejected, simply that
improvements are needed.6
(1) Doctors’ leaders have opposed the proposal on the basis that in no other
profession are employees restricted from using their free time as they wish.
(Independent, 18.10.2000, p. 11, col. 1)
(2) They feel that it’s possible that not many months ago that anthrax – a small
quantity of it – was handed over in Prague, Czechoslovakia, to Mohammed
Atta, one of the pilots of one of the planes that ﬂew into the World Trade
Centre. (Guardian, 16.10.2001, p. 4, col. 2)
(3) Even now, a senior editor points out that, if she really is such a simple soul,
how did she wind up at a top literary agency, Peters, Fraser and Dunlop?
(Sunday Times, 18.2.2001, p. 5, col. 1)
(4) But I completely understand that once they found him that his daughter
wanted a funeral. (Guardian, G2, 7.2.2002, p. 9, col. 2)
(5) I feel very strongly that if women are experiencing domestic violence that they
should tell their GP. (Guardian, 22.12.2003, p. 7, col. 7)
Exercise 9 Wh-movement (T)
Identify all the instances of movement of a wh-constituent in the following examples.
Classify your examples depending on the type of clause that the moved constituent
5
For some discussion see Haegeman (2000b) and the references cited there.
Notice that in (1)–(3) above the interrogative constituent is why. For discussion of why see
Rizzi (2001).
6
For more detailed discussions of the structure of the CP domain see Rizzi (1997). For English
and Hiberno English see McCloskey (forthcoming).
TSC05 07/18/2005, 03:56PM360
The Periphery of the Sentence 361
introduces (relative/interrogative; ﬁnite/non-ﬁnite). Locate the base position of the
moved constituent and identify its function (subject, object, adjunct).
(1) Within 10 minutes you’ll hit upon a television program designed to monitor
the most intimate details of our lives in the hopes of ﬁnding something mildly
amusing with which to capture the attention of ﬁckle viewers. (Chicago
Tribune, 22.12.2003, Section 13, p. 2, col. 2)
(2) Thrilled, he accepted the change and brought along several stuffed animal
friends who he thought would like the ﬁre-truck bed too. (Washington Post,
10.12.2002, p. F4, col. 4)
(3) A lot of designers come from the perspective of being inventors, which I
think is so bogus. (New York Times, 28.11.2002, p. D5, col. 2)
(4) Amy and her mom are together in the kitchen one night. Her mom says she
should learn how to cook. Amy seizes the moment: “What kind of husband
do you see me with?” (Washington Post, 10.12.2002, p. A15, col. 1)
(5) The National Design Museum turned down the Boyms’ proposal in 1994 to
sell everyday items like light bulbs, graced with the museum’s logo, as gift
shop merchandise, which it told Mr Boym was uncommercial thinking. (New
York Times, 28.11.2002, p. D4, col. 1)
(6) [Her mother] doesn’t know that Amy has already fallen [for a boyfriend] . . .
Amy spends hours talking to him on her cell phone, which she sleeps with
under her pillow. (Washington Post, 10.12.2002, p. A15, col. 1)
(7) Inevitably, the changes have laid bare frictions – which all sides say were
inevitable, and perhaps healthy – within Trinity Parish, which operates St.
Paul’s, and New York’s Episcopal world. (New York Times, 28.11.2002,
p. A28, col. 1)
(8) The SEC has no chief accountant and can’t sensibly appoint one until it’s
clear whom this ofﬁcial would report to. (Washington Post, 10.12.2002,
p. A28, col. 1)
(9) There are dozens of great new TVs out there. The critical question, even
more than what to buy, is when to buy. (Chicago Tribune, 22.12.2002,
Section 15, p. 3, col. 1)
(10) [Nawid’s family] now have no land on which to build, even if they had the
money. (New York Times, 2.1.2003, p. A8, col. 3)
Exercises 361
TSC05 07/18/2005, 03:56PM361
362 Chapter 5
Exercise 10 Constraints on extraction (T, E)
10A Consider (1). Identify the antecedent of the relative pronoun who. Insert lefthand
brackets labeled “[IP” and “[CP” in the representation of the underlined relative
clause. Add the appropriate right-hand brackets. Represent the unpronounced copy
of the relative pronoun by strikethrough (who). Signal the implicit subject of an
inﬁnitive by means of the symbol [NP Ø] in the SpecIP position. Discuss any problems
that arise.
(1) For decades now, the post of Arts minister has been a dumping ground for
nice people who political leaders of the day don’t know where else to put.
(Independent on Sunday, 9.5.2004, p. 24 News, col. 1)
KEY AND COMMENTS
In (1) the relative pronoun who must have ﬁrst merged with the verb put. This verb
is found in a non-ﬁnite interrogative clause whose SpecCP is ﬁlled by a wh-constituent
where else. The extraction of who is unexpected because we have seen that moving
a wh-constituent across a ﬁlled speciﬁer of an intervening CP leads to a degradation.
(1′) nice people [CP who [IP political leaders of the day don’t know
[CP where else [IP [NP Ø] to [VP [NP Ø] put who]]]]].
10B In section 5.3 of the chapter we postulated that a relative pronoun may be
non-overt. Consider the attested examples in (2). For each example insert left-hand
brackets labeled “[IP” and “[CP” in the representation of the underlined relative
clauses. Add the matching right-hand brackets. Represent unpronounced copies of
the relative pronoun by strikethrough (which). Signal the implicit subject of inﬁnitives
by means of the symbol [NP Ø] in SpecIP. Discuss the problems raised by the examples.
Would the examples in (2) offer any arguments against the hypothesis that relative
clauses may be introduced by an unpronounced pronoun (which)?
(2) a These are things experienced inﬁelders know how to do. (USA Today,
26.3.2004, p. 15C, cols 2–3)
b There’s only one thing we don’t know how to do properly, and that’s sing
like the northern hemisphere sides. (Guardian, 13.11.2003, p. 21, col. 6)
c These are struggles the government decided how to conduct before it
came to power, and the case for its policy then remains as correct as ever.
(Guardian, 26.11.2002, p. 8, col. 6)
d Something I know how to do is close a deal. (New York Times, 1.8.2004,
p. 7 (ST), col. 4)
e We did everything we knew how to do to continue to pursue al Quaeda.
(Wall Street Journal, 29.3.2004, p. A14, col. 5)
TSC05 07/18/2005, 03:56PM362
The Periphery of the Sentence 363
KEY AND COMMENTS
Assuming that the examples in (2) are derived by movement of a non-pronounced
relative pronoun which, they display the same unexpected pattern illustrated in (1)
in section A:
(2)′ a things [CP which [IP experienced inﬁelders know [CP how [IP [NP Ø] to do
which]]]]
b one thing [CP which [IP we don’t know [CP how [IP [NP Ø] to do properly
which]]]]
c struggles [CP which [IP the government decided [CP how [IP [NP Ø] to conduct
which]]]]
d something [CP which [IP I know [CP how [IP [NP Ø] to do which]]]]
e everything [CP which [IP we knew [CP how [IP [NP Ø] to do which]]]]
Taken all by themselves the unexpected examples in (2) might at ﬁrst sight be
thought to constitute evidence against the non-overt relative pronoun hypothesis.
After all, a non-overt pronoun would be moved across an interrogative constituent
(how). However, example (1) in section A shows a similar unexpected pattern with
an overt pronoun. This means that even with respect to the non-expected patterns,
non-overt relative pronouns (represented as which in (2′)) behave in the same way
as pronounced pronouns (who in (1)).7
Exercise 11 Relative clauses and resumptive
pronouns (T, E)
While discussing the derivation of relative clauses and of interrogative clauses we
have been assuming that a wh-constituent is moved from its base position, in which
it leaves an unpronounced copy. For instance, for the embedded clauses for our
discussion examples (64b) and (65a), repeated here as (1a) and (1b), we proposed
representation (1c), in which who signals the unpronounced copy of the moved
relative pronoun who.
(1) a I wonder who he had met at the party.
b I interviewed the man who he had met at the party.
c [CP who [IP he [I had] [VP he have [VP he met who at the party]]]]8
7
Observe that several of the unexpected examples in (2) concern the sequence know how to. It
may well be that the pattern know how to has special properties. Cinque (2004b: 140) shows
that the Italian analogue of know how to also displays special properties.
8
Recall that we assume that there is an unpronounced copy of the moved subject in the speciﬁer
position associated with the projection of the auxiliary have. See Chapter 4, section 4, for
discussion.
Exercises 363
TSC05 07/18/2005, 03:56PM363
364 Chapter 5
Discuss the problems raised for this analysis by the following attested example:
(2) It was a background discussion which my understanding was that it would not
appear anywhere. (Guardian, 21.8.2003, p. 9, col. 5)
COMMENTS
In this example, instead of having an unpronounced copy in a lower position of the
relative pronoun which we ﬁnd the pronoun it. When a pronoun occupies a position
in which we would have expected an unpronounced copy of a wh-constituent we
refer to it as a resumptive pronoun.
(2′) It was a background discussion [CP which [IP my understanding was [CP that
[IP it would not appear anywhere]]]].
Could we remove the resumptive pronoun and replace it by an unpronounced
copy of the relative pronoun which? That is to say, would the spell-out of (3a)
be acceptable?
(3) a *It was a background discussion [CP which [IP my understanding was
[CP which that [IP which would not [VP which appear anywhere]]]]].
Forming relative clauses such as that in (2) without using the resumptive pronoun
strategy leads to an ungrammatical result, because such examples violate the
constraint on subject extraction (see section 3.2.3.1).
(3) b *It was a background discussion which my understanding was that would
not appear anywhere.
Exercise 12 Extraction from adjunct clauses and
resumptive pronouns (E)
In Exercise 11 we discovered that resumptive pronouns can be inserted to
overcome constraints on movement. Consider the examples below, which also
contain resumptive pronouns in a lower position of a relative pronoun. Locate
the relative pronoun and the related resumptive pronoun. Can you remove the
resumptive pronoun (that is, can you replace it by an unpronounced copy of the
relative pronoun)?
(1) Bernie is the type of man who when you shake hands with him, it’s a deal.
(Guardian, G2, 11.7.2001, p. 5, col. 3)
TSC05 07/18/2005, 03:56PM364
The Periphery of the Sentence 365
(2) They say he was a workaholic and that work was a drug which when he
couldn’t have it anymore he got depressed. (Jonathan Frantzen, The Corrections,
2001: 75).
KEY AND COMMENTS
The relative pronouns in examples (1) and (2) are related to a resumptive pronoun
inside an adjunct clause. The relative pronoun is outside the adjunct clause while
the resumptive pronoun is inside it.
(1′) Bernie is the type of man who [CP when you shake hands with him], it’s a deal.
(2′) They say he was a workaholic and that work was a drug which [CP when he
couldn’t have it anymore] he got depressed.
For (1) and in (2) replacing the resumptive pronouns by an unpronounced copy of
the relative pronoun would lead to a degradation. Representations (3a, b) would
correspond to sentences (4a, b):
(3) a Bernie is the type of man who [when you shake hands with who], it’s a deal.
b They say he was a workaholic and that work was a drug which [when he
couldn’t have which anymore] he got depressed.
(4) a *Bernie is the type of man who [when you shake hands with], it’s a deal.
b *They say he was a workaholic and that work was a drug which [when he
couldn’t have anymore] he got depressed.
These examples reveal a further constraint on movement of relative pronouns:
extracting a relative pronoun from an adjunct clause also leads to a degradation.
Exercise 13 Movement from adjunct clauses (E)
In Exercise 12 we discovered that it is not possible to extract a relative pronoun
from an adjunct clause. The relevant examples (4a, b) from Exercise 12 are repeated
here in (1).
(1) a *Bernie is the type of man who when you shake hands with, it’s a deal.
b *They say he was a workaholic and that work was a drug which when he
couldn’t have anymore he got depressed.
In the discussion we have assumed that movement of the relative wh-pronoun is
similar (or identical) to the movement of the interrogative pronoun. What predictions
do we make for the status of the following interrogative examples? Are these predictions
correct?
Exercises 365
TSC05 07/18/2005, 03:56PM365
366 Chapter 5
(2) a Who would it be a deal when you simply shake hands with?
b Which drug did he get depressed when he couldn’t have any more?
KEY AND COMMENTS
(2a) is unacceptable; (2b) is degraded and for many speakers it is also unacceptable.
The degradations are predicted by our hypothesis that movement of relative pronouns
and movement of interrogative pronouns are two instantiations of the same
operation Move.
Exercise 14 That relatives (T, E)
(1a) corresponds to the text example (75a). Compare the underlined string in (1b)
with the relative clause in (1a):
(1) a F.B.I. agents investigating falsiﬁed identity papers are expanding their dragnet
for a growing list of foreign-born men they believe may have entered
the Unites States illegally from Canada.
b F.B.I. agents investigating falsiﬁed identity papers are expanding their
dragnet for a growing list of foreign-born men that they believe may have
entered the Unites States illegally from Canada.
We have proposed that in (1a) a non-overt relative pronoun undergoes (long)
movement. In the literature it is often proposed that in (1b) a relative pronoun has
also moved to SpecCP, but again this pronoun is not pronounced. In other words
(1b) would have the partial representation in (1c).
(1) c foreign-born men [CP who that [IP they believe [CP who [IP who may [VP who
have [VP who entered the Unites States illegally from Canada]]]]]]9
Consider the examples below. Discuss how they provide support for the non-overt
pronoun hypothesis represented in (1c).
(2) a *F.B.I. agents investigating falsiﬁed identity papers are expanding their
dragnet for a growing list of foreign-born men that they wonder how may
have entered the Unites States from Canada.
b *F.B.I. agents investigating falsiﬁed identity papers are expanding their
dragnet for a growing list of foreign-born men that they believe that may
have entered the Unites States illegally from Canada.
9
Recall that we assume that there is an unpronounced copy of the moved subject in the speciﬁer
position associated with the projection of the auxiliary have. See Chapter 4, section 4, for
discussion.
TSC05 07/18/2005, 03:57PM366
The Periphery of the Sentence 367
KEY AND COMMENTS
In (2a) we have inserted the interrogative constituent how in the speciﬁer position
of an intermediate CP. This leads to a strong degradation in acceptability of the
example. Similarly, inserting that in the position adjacent to the subject position
from which the non-overt relative pronoun would have been moved leads to a
strong degradation in (2b). These two effects are identical to those discussed in
connection with movement of overt interrogative and relative pronouns in sections
3.2 and 5.2 of the chapter, and they also extend to movement of non-overt relative
pronouns as discussed in section 5.3.
One point needs to be added here: in Modern English the complementizer that
can only be used to introduce a relative clause if the constituent in SpecCP itself is
non-overt. If both the constituent in SpecCP and the complementizer that in the
adjacent C position are overt (representation (1d)), the resulting sentence is ungrammatical
(1e):10
(1) d *foreign-born men [CP who that [IP they believe [CP who [IP who may [VP who
have [VP who entered the Unites States illegally from Canada]]]]]].
e *F.B.I. agents investigating falsiﬁed identity papers are expanding their dragnet
for a growing list of foreign-born men who that they believe may have
entered the Unites States illegally from Canada.
Discuss how the that relative in (3) could be derived:
(3) These are precisely the kinds of things that students and faculty members will
ﬁnd at the newly renovated Milbank Memorial library. (New York Times,
Education, 1.8.2004, p. 19, col. 1)
Exercise 15 Postverbal subjects and inversion in
French (E)
Consider the French examples in (1). How could we represent the structure for (1a)?
And for (1b)? On the basis of their English glosses you might be tempted to assign
to (1b) the same structure as (1a). The only difference would be that while the subject
in (1b) is a lexical NP les étudiants (‘the students’), that in (1a) is a pronoun.
(1) a Combien d’argent dépensent-ils?
how much money spend-3PL they
‘How much money do they spend?’
10
There is a similar effect with interrogatives, cf. section 2.6.
Exercises 367
TSC05 07/18/2005, 03:57PM367
368 Chapter 5
b Combien d’argent dépensent les étudiants?
how much money spend-3PL the students
‘How much money do the students spend?’
On the basis of the data below, can we maintain that (1a) and (1b) have the same
derivation?
(2) a Combien d’argent ont- ils dépensé?
how much money have-3PL -they spent-PART
‘How much money have they spent?’
b *Combien d’argent ont les étudiants dépensé?
how much money have-3PL the studentsspent-PART
(3) a *Combien d’argent ont dépensé ils?
how much money have-3PL spent-PART they
b Combien d’argent ont dépensé les étudiants?
how much money have-3PL spent-PART the students
‘How much money have the students spent?’
(4) a *Je me demande combien d’argent dépensent-ils.
I wonder how much money spend-3PL they
b Je me demande combien d’argent dépensent les étudiants.
I wonder how much money spend-3PL the students
‘I wonder how much money the students spend.’
c Je me demande combien d’argent ont dépensé les étudiants.
I wonder how much money have-3PL spent-PART the students
‘I wonder how much money the students have spent.’
(5) a Dépensent-ils beaucoup d’argent?
spend-3PL they a lot of money
‘Do they spend a lot of money?’
b *Dépensent les étudiants beaucoup d’argent?
spend-3PL the students a lot of money
KEY AND COMMENTS
The derivation of (1a) is relatively straightforward: we assume that the inﬂected
verb dépensent (‘spend’) has undergone inversion (SVI) and has moved to C. The
fronted direct object combien d’argent (‘how much money’) is an interrogative
constituent in SpecCP:
TSC05 07/18/2005, 03:57PM368
The Periphery of the Sentence 369
(1) c [CP [NP Combien d’argent] [C dépensent] [IP ils [I dépensent] [VP ils dépens[NP
combien d’argent]]]].
One might think that (1b) has the same derivation, with a lexical NP les étudiants
(‘the students’) rather than a pronoun (ils, ‘they’) as the subject in SpecIP. But the
additional data show that the distribution of pronominal subjects is different from
that of full NP subjects.
In (2a) the inﬂected auxiliary ont (‘have’) has moved to C, the lexical VP is
headed by a past participle dépensé (‘spent’).
(2) c [CP [NP Combien d’argent] [C ont] [IP ils [I ont] [VP ils av-11
[VP ils dépensé [NP
combien d’argent]]]]]?
(2b/d) shows that the analogue of derivation (2a/c) with an NP subject is
ungrammatical:
(2) d *[CP [NP Combien d’argent] [C ont] [IP les étudiants [I ont] [VP les étudiants av[VP
les étudiants dépensé [NP combien d’argent]]]]]?
Rather, as shown by (3b), a lexical subject must occupy a position to the right of
the participle. This is a position in which the pronominal subject cannot occur (3a).
As shown by (4a), with pronominal subjects, SVI is restricted to main clauses. In
contrast, the inverted position occupied by lexical subjects is also available in
embedded interrogatives (4b), in which the lexical subject can also follow a participle
(4c).
(5a) shows that with a pronominal subject, SVI is also found in yes/no questions.
On the other hand, in yes/no questions lexical subjects cannot occupy the postverbal
position.
We conclude that though both (1a) and (1b) contain postverbal subjects, we
have to distinguish their structures.12
11
In (2c) we represent the root of the verb avoir (‘have’) as av-. This is an approximation.
12
For introductory discussions of these data see Battye, Hintze, and Rowlett (2000: 202–5) and
Rowlett (2005). For advanced discussion see Kayne (1972), and Kayne and Pollock (1978,
2001).
For examples in English in which a postverbal NP subject is not derived by SAI see Chapter
1, Exercises 11 and 12, and Chapter 4, Exercise 23.
Exercises 369
TSC05 07/18/2005, 03:57PM369
370 Bibliography
Bibliography
Aarts, Bas and Liliane Haegeman (forthcoming) English word classes and phrases. In Bas
Aarts and April McMahon (eds), The Handbook of English Linguistics, Oxford: Blackwell.
Abney, Steven (1987) The English Noun Phrase in its Sentential Aspect. PhD dissertation,
MIT.
Abraham, Werner, Sam Epstein, Hoskuldur Thráinsson, and Jan Wouter Zwart (eds) (1996)
Minimal Ideas, New York and Amsterdam: John Benjamins.
Adger, David (2003) Core Syntax, Oxford and New York: Oxford University Press.
Alexiadou, Artemis (1997) Adverb Placement, New York and Amsterdam: John Benjamins.
Alexiadou, Artemis (2001) Functional Structure in Nominals, New York and Amsterdam:
John Benjamins.
Alexiadou, Artemis (2004a) On the development of possessive determiners: Consequences
for DP structure. In Eric Fuss and Carola Trips (eds), Diachronic Clues to Synchronic
Grammar, New York and Amsterdam: John Benjamins, pp. 31–58.
Alexiadou, Artemis (ed.) (2004b) Adverbs across Frameworks. Special edition of Lingua,
114.
Alexiadou, Artemis and Elena Anagnostopoulou (1998) Parametrizing AGR: Word order,
verb movement and EPP checking. Natural Language and Linguistic Theory, 16, 491–539.
Alexiadou, Artemis and Elena Anagnostopoulou (2001) The subject in situ generalisation
and the role of case in driving computations. Linguistic Inquiry, 32, 193–213.
Authier, J.-M. (1992) Iterated CPs and embedded topicalisation. Linguistic Inquiry, 23,
329–36.
Baker, Carl Lee (1991) The syntax of English not: The limits of core grammar. Linguistic
Inquiry, 22, 387–429.
Baker, Mark (1997) Thematic roles and syntactic structure. In Liliane Haegeman (ed.), Elements
of Grammar: A Handbook of Generative Syntax, Dordrecht: Kluwer, pp. 73–138.
Baker, Mark (2002) Building and merging, not checking: the non existence of (Aux)-S-V-O
languages. Linguistic Inquiry, 33, 321–8.
Baker, Mark, Kyle Johnson, and Ian Roberts (1989) Passive arguments raised. Linguistic
Inquiry, 20, 219–51.
Baltin, Mark R. and Chris Collins (eds) (2001) The Handbook of Contemporary Syntactic
Theory, Oxford: Blackwell.
Basilico, David (1998) Object position and predication forms. Natural Language and Linguistic
Theory, 16, 541–95.
Battistella, Ed (1995) The syntax of the double modal construction. Linguistica Atlantica,
21, 49–65.
TSD01 07/18/2005, 03:57PM370
Bibliography 371
Battye, Adrian, Marie-Anne Hintze, and Paul Rowlett (2000) The French Language Today:
A Linguistic Introduction, 2nd edition, London: Routledge.
Belletti, Adriana (1988) The case of unaccusatives. Linguistic Inquiry, 19, 1–35.
Belletti, Adriana (1990) Generalised Verb Movement: Aspects of Verb Syntax, Turin:
Rosenberg and Tellier.
Belletti, Adriana and Luigi Rizzi (1988) Psych verbs and theta theory. Natural Language and
Linguistic Theory, 6, 291–352.
Bernstein, Judy B. (2001) The DP hypothesis: Identifying clausal properties in the nominal
domain. In Mark Baltin and Chris Collins (eds), The Handbook of Contemporary Syntactic
Theory, Oxford: Blackwell, pp. 536–61.
Birner, Betty (1992) The Discourse Function of Inversion in English. PhD dissertation, Northwestern
University.
Birner, Betty (1994) Information status and word order: An analysis of English inversion.
Language, 70, 233–59.
Birner, Betty and Gary L. Ward (1992) On the interpretation of VP inversion in American
English. Journal of Linguistics, 28, 1–12.
Blakemore, Diane and Robyn Carston (eds) (2005) Coordination: Syntax, Semantics and
Pragmatics. Special issue of Lingua, 115.
Bobaljik, Jonathan (1995) Morpho Syntax: The Syntax of Verbal Inﬂection. PhD dissertation,
MIT.
Bobaljik, Jonathan (2002) Realizing Germanic inﬂection: Why morphology does not drive
syntax. The Journal of Comparative Germanic Linguistics, 6, 129–67.
Bobaljik, Jonathan (2003) Floating quantiﬁers: Handle with care (revised version). In Lisa
Cheng and Rint Sybesma (eds), The Second Glot International State-of-The-Article Book,
Berlin: Mouton de Gruyter, pp. 107–48.
Bobaljik, Jonathan and Diane Jonas (1996) Subject positions and the role of TP. Linguistic
Inquiry, 27, 195–236.
Borsley, Robert and Richard Ingham (2002) Grow your own linguistics? On some applied
linguists’ views of the subject. Lingua, 112 (1), 1–6.
Borsley, Robert and Richard Ingham (2003) More on “some applied linguistics”: A response
to Stubbs. Lingua, 113, 193–6.
Boskovic, Zeljko (2002) A-movement and the EPP. Syntax, 5, 167–218.
Boskovic, Zeljko (2004) Be careful where you ﬂoat your quantiﬁers. Natural Language and
Linguistic Theory, 22, 681–742.
Bowers, John (1993) The syntax of predication. Linguistic Inquiry, 24, 591–665.
Bowers, John ( 2001) Predication. In Mark Baltin and Chris Collins (eds), The Handbook of
Contemporary Syntactic Theory, Oxford: Blackwell, pp. 299–334.
Bowers, John (2002) Transitivity. Linguistic Inquiry, 33, 183–225.
Bresnan, Joan (1994) Locative inversion and the architecture of universal grammar. Language,
70, 72–131.
Brown, Keith (1992) Double modals in Hawick Scots. In Peter Trudgill and J. K. Chambers,
(eds), Dialects of English: Studies in Grammatical Variation, London: Longmans, pp. 74–
103.
Camacho, José (2003) The Structure of Coordination, Dordrecht: Kluwer.
Caponigro, Ivano and Carson T. Schütze (2003) Parametrizing passive participle movement.
Linguistic Inquiry, 34, 293–308.
Carnie, Andrew (2002) Syntax: A Generative Introduction, Oxford: Blackwell.
TSD01 07/18/2005, 03:57PM371
372 Bibliography
Chomsky, Noam (1970) Remarks on nominalizations. In Roderick Jacobs and Peter
Rosenbaum (eds), Readings in English Transformational Grammar, Waltham, MA: Ginn.
Also in Noam Chomsky (1972) Studies on Semantics in Generative Grammar, The Hague:
Mouton.
Chomsky, Noam (1972) Studies on Semantics in Generative Grammar, The Hague: Mouton.
Chomsky, Noam (1975) The Logical Structure of Linguistic Theory, New York: Plenum.
Chomsky, Noam (1995) The Minimalist Program, Cambridge, MA: MIT Press.
Cinque, Guglielmo (1999) Adverbs and Functional Heads: A Cross-linguistic Perspective,
New York/Oxford: Oxford University Press.
Cinque, Guglielmo (2004a) Issues in adverbial syntax. In Artemis Alexiadou (ed.), Adverbs
across Frameworks, Special edition of Lingua, 114, 683–710.
Cinque, Guglielmo (2004b) Restructuring and functional structure. In Adriana Belletti (ed.),
Structures and Beyond: The Cartography of Syntactic Structures, vol. 3, Oxford and New
York: Oxford University Press, pp. 132–91.
Clahsen, Harald (ed.) (1996) Generative Perspectives on Language Acquisition, Amsterdam:
John Benjamins.
Collins, Chris and Höskuldur Thráinsson (1993) Object shift in double object constructions
and the theory of case. MIT Working Papers in Linguistics, 19, 131–74.
Coopmans, Peter (1989) Where stylistic and syntactic processes meet: Locative inversion in
English. Language, 65, 728–51.
Cormack, Annabel and Neil Smith (2002) Modals and negation in English. In Sjef Barbiers,
Frits Beukema, and Wim van der Wurff (eds), Modality and Its Interaction with the Verbal
System, New York and Amsterdam: John Benjamins, pp. 133–64.
Cottell, Siobhan (2002) The Comparative Syntax of Cleft Constructions. PhD dissertation,
University of Wales, Bangor.
Cottell, Siobhan and Alison Henry (2004) Transitive expletives, quantiﬁer stranding and
subject positions in an English dialect. Paper presented at the Comparative Germanic
Syntax Workshop 19, CUNY.
Crystal, David (1971) Linguistics, London: Penguin.
Culicover, Peter (1991) Topicalization, inversion and complementizers in English. In Denis
Delﬁtto, Martin Everaert, Arnold Evers, and Frits Stuurman (eds), OTS Working Papers:
Going Romance and Beyond, Utrecht: University of Utrecht, pp. 1–45.
Culicover, Peter (1993) Evidence against ECP accounts of the that-t effect. Linguistic Inquiry
24, 557–61.
Culicover, Peter and Robert Levine (2001) Stylistic inversion in English: A reconsideration.
Natural Language and Linguistic Theory, 19, 283–310.
Davies, William D. and Stanley Dubinsky (1999) Sentential subjects as complex NPs: New
reasons for an old account of subjacency. Chicago Linguistics Society, 34, 83–94.
Davies, William D. and Stanley Dubinsky (2001a) Functional architecture and the distribution
of subject properties. In William D. Davies and Stanley Dubinsky (eds), Objects and
other Subjects, Dordrecht: Kluwer, pp. 247–80.
Davies, William D. and Stanley Dubinsky (eds) (2001b) Objects and other Subjects, Dordrecht:
Kluwer.
DeGraff, Michel (1997) Verb syntax in creolization (and beyond). In Liliane Haegeman (ed.),
The New Comparative Syntax, London: Longman, Addison and Wesley, pp. 64–94.
DeGraff, Michel (1999) Language Creation and Language Change: Creolization, Diachrony
and Development, Boston, MA: MIT Press.
TSD01 07/18/2005, 03:57PM372
Bibliography 373
Diesing, Molly (1997) Yiddish VP order and the typology of object movement in Germanic.
Natural Language and Linguistic Theory, 15, 369–427.
Dik, Simon (1989) Functional grammar and its relevance to grammar writing. In Gottfried
Graustein and Gerhard Leitner (eds), Reference Grammars and Modern Linguistic Theory,
Tübingen: Niemeyer, pp. 33–55.
Dikken, Marcel den and Alma Næss (1993) Case dependencies: The case of predicate inversion.
The Linguistic Review, 10, 303–36.
Doetjes, Jenny (1992) Rightward ﬂoating quantiﬁers ﬂoat to the left. The Linguistic Review,
9, 313–32.
Dufﬁeld, Nigel (1993) On case checking and NPI licensing in Hiberno English. Rivista di
Linguistica, 5, 215–44.
Einstein, Albert (1954) Ideas and Opinions, New York: Crown.
Emonds, Joseph (1976) A Transformational Approach to English Syntax, New York: Academic
Press.
Emonds, Joseph (1978) The verbal complex V′-V in French. Linguistic Inquiry, 9, 151–75.
Ernst, Thomas (1992) The phrase structure of English negation. The Linguistic Review, 9,
109–44.
Ernst, Thomas (2002) The Syntax of Adjuncts, Cambridge: Cambridge University Press.
Ernst, Thomas (2004) Principles of adverbial distribution in the lower clause. In Artemis
Alexiadou (ed.), Adverbs across Frameworks, special edition of Lingua, 114, 755–78.
Fischer, Olga, Ans van Kemenade, Willem Koopman, and Wim van der Wurff (2000) The
Syntax of Early English, Cambridge: Cambridge University Press.
Frazier, Lynn and C. Clifton, Jr. (1989) Successive cyclicity in the grammar and parsing.
Language and Cognitive Processes, 4, 93–126.
Friedemann, MarcAriel and Luigi Rizzi (eds) (1999) The Acquisition of Syntax, London:
Longman, Addison, Wesley.
Friedemann, MarcAriel and Tal Siloni (1993) AgrO is not AgrPart. GenGenP, 1, 41–53.
Fu, Jingqi, Thomas Roeper, and Hagit Borer (2001) The VP within process nominals: Evidence
from adverbs and the VP anaphor do so. Natural Language and Linguistic Theory,
19, 549–82.
Gavruseva, Elena (2000) On the syntax of possessor extraction. Lingua, 110, 743–72.
Gibson, Edward and Tessa Warren (2004) Reading–time evidence for intermediate linguistic
structure in long-distance dependencies. Syntax, 7, 55–78.
Ginzburg, Johnathan (1999) Interrogatives: Questions, facts and dialogue. In Shalom Lappin
(ed.), The Handbook of Contemporary Semantic Theory, Oxford: Blackwell, pp. 385–
422.
Giorgi, Alessandra and Giuseppe Longobardi (1991) The Syntax of Noun Phrases: Conﬁguration,
Parameters and Empty Categories, Cambridge: Cambridge University Press.
Goodall, Grant (1987) Parallel Structures in Syntax, Dordrecht: Kluwer.
Goodall, Grant (2001) The EPP in Spanish. In William D. Davies and Stanley Dubinsky,
Objects and other Subjects, Dordrecht: Kluwer, pp. 193–223.
Graustein, Gottfried and Gerhard Leitner (1989) Reference Grammars and Modern Linguistic
Theory, Tübingen: Niemeyer.
Grimshaw, Jane (1990) Argument Structure, Cambridge, MA: MIT Press.
Haeberli, Erik (2000) Adjuncts and the syntax of subjects in Old and Middle English. In
Susan Pintzuk, George Tsoulas, and Anthony Warner (eds), Diachronic Syntax: Models
and Mechanisms, Oxford: Oxford University Press, pp. 109–31.
TSD01 07/18/2005, 03:57PM373
374 Bibliography
Haegeman, Liliane (1994) Introduction to Government and Binding Theory. Oxford:
Blackwell.
Haegeman, Liliane (1995) The Syntax of Negation, Cambridge: Cambridge University Press.
Haegeman, Liliane (ed.) (1997a) Elements of Grammar: A Handbook of Generative Syntax,
Dordrecht: Kluwer.
Haegeman, Liliane (ed.) (1997b) The New Comparative Syntax, London: Longman, Addison
and Wesley.
Haegeman, Liliane (2000a) Negative inversion, the NEG criterion and the structure of CP. In
Larry Horn and Yasuhiko Kato (eds), Negation and Polarity, Oxford and New York:
Oxford University Press, pp. 29–69.
Haegeman, Liliane (2000b) Inversion, non-adjacent inversion and adjuncts in CP. In Paul
Rowlett (ed.), Transactions of the Philological Society, Special Issue: Papers from the
Salford Negation conference, 98 (1), 121–60.
Haegeman, Liliane (2003) Notes on long adverbial fronting in English and the left periphery.
Linguistic Inquiry, 34, 640–9.
Haegeman, Liliane and Jacqueline Guéron (1999) English Grammar: A Generative Perspective.
Oxford: Blackwell.
Harlow, Stephen (1981) Government and relativisation in Celtic. In Frank Heny (ed.), Binding
and Filtering, London: Croom Helm, pp. 213–54.
Hegarty, Michael (2003) Semantic types of abstract entities. Lingua, 113, 891–927.
Henry, Alison (1995) Belfast English and Standard English: Dialect Variation and Parameter
Setting, Oxford: Oxford University Press.
Henry, Alison (2001) The direction and motivation of movement: Transitive expletives
and ﬂoating associates in Belfast English. Paper presented at the Motivating Movement
conference, Jordanstown, January.
Heycock, Caroline and Anthony Kroch (1997) Inversion and equation in copular sentences.
ZAS Papers in Linguistics, 10, 71–87.
Hoekstra, Teun and René Mulder (1990) Unergatives as copular verbs: Locational and
existential predication. The Linguistic Review, 7, 1–79.
Holmberg, Anders (1986) Word Order and Syntactic Features in the Scandinavian Languages
and English. PhD dissertation, University of Stockholm.
Holmberg, Anders (1999) Remarks on Holmberg’s generalization. Studia Linguistica, 53, 1–
39.
Holmberg, Anders and Christer Platzack (1995) The Role of Inﬂection in Scandinavian
Syntax, Oxford and New York: Oxford University Press.
Hoop, Helen de, Marco Haverkot, and Maurits van den Noort (2004) Variation in Form
Versus Variation in Meaning. Special issue of Lingua, 114, 1071–89.
Hooper, John and Sandra Thompson (1973) On the applicability of Root Transformations.
Linguistic Inquiry, 4, 465–97.
Hornstein, Norbert, Jairo Nunes, and Kleanthes K. Grohmann (forthcoming) Understanding
Minimalism: An Introduction to Minimalist Syntax. Cambridge: Cambridge University
Press.
Horrocks, Geoff and Melita Stavrou (1987) Bounding theory and Greek syntax: Evidence for
wh-movement in NP. Journal of Linguistics, 23, 79–108.
Huddleston, Rodney (1994) The contrast between interrogatives and questions. Journal of
Linguistics, 30, 411–40.
Jackendoff, Ray (1977) X′ Syntax: A Study of Phrase Structure, Cambridge, MA: MIT Press.
TSD01 07/18/2005, 03:57PM374
Bibliography 375
Jackendoff, Ray (1987) The status of thematic relations in linguistic theory. Linguistic
Inquiry, 30, 69–96.
Jaworska, Ewa (1986) Prepositional phrases as subjects and objects. Journal of Linguistics,
22, 355–75.
Jonas, Diane and Jonathan Bobaljik (1993) Specs for subjects: The role of TP in Icelandic. In
Jonathan Bobaljik and Collin Phillips (eds), MIT Working Papers in Linguistics: Papers on
Case and Agreement, I, 59–98.
Kayne, Richard (1972) Subject inversion in French interrogatives. In J. Casagrande and
B. Saciuk (eds), Generative Studies in Romance Languages, Rowley, MA: Newbury
House, pp. 70–126.
Kayne, Richard (1984) Connectedness and Binary Branching, Dordrecht: Foris.
Kayne, Richard (1994) The Antisymmetry of Syntax, Cambridge, MA: MIT Press.
Kayne, Richard and Jean-Yves Pollock (1978) Stylistic inversion, successive cyclicity, and
move NP in French. Linguistic Inquiry, 9, 595–621.
Kayne, Richard and Jean-Yves Pollock (2001) New thoughts on stylistic inversion. In Aafke
Hulk and Jean-Yves Pollock (eds), Subject Inversion in Romance and the Theory of Universal
Grammar, Oxford: Oxford University Press, pp. 107–61.
Kemenade, Ans van (1987) Syntactic Case and Morphological Case in the History of English.
PhD dissertation, University of Utrecht.
Kiparsky, Paul and Carol Kiparsky (1971) Fact. In Danny Steinberg and Lee Jakobovitz
(eds), Semantics: An Interdisciplinary Reader, Cambridge: Cambridge University Press,
pp. 345–69.
Kitagawa, Yoshihisha (1994) Subjects in Japanese and English, New York: Garland.
Klima, Joseph (1964) Negation in English. In Jerry Fodor and Jerold Katz (eds), The Structure
of Language, New Jersey: Prentice-Hall, pp. 246–323.
Koopman, Hilda and Dominique Sportiche (1991) The position of subjects. Lingua, 85,
211–58.
Kornai, Andras and Geoffrey K. Pullum (1990) The X-bar theory of phrase structure. Language,
66, 24–50.
Koster, Jan (1978) Why subject sentences don’t exist. In Jay Keyser (ed.), Recent Transformational
Studies in European Languages, Cambridge, MA: MIT Press, pp. 53–64.
Kroch, Anthony and Ann Taylor (2000) Verb-order in Early Middle English. In Susan Pintzuk,
Georges Tsoulas, and Anthony Warner (eds), Diachronic Syntax: Models and Mechanisms,
Oxford: Oxford University Press, pp. 132–63.
Kuroda, Sige-Yuki (1986) Whether we agree or not: A comparative syntax of English and
Japanese. Linguisticae Investigationes, 12, 1–47.
Law, Paul (1999) On the passive existential construction. Studia Linguistica, 53, 183–208.
Lehmann, Christian (2004) Data in linguistics. The Linguistic Review, 21, 175–210.
Levin, Beth and Malka Rappaport-Hovav (1995) Unaccusativity: At the Syntax-Lexical
Semantics Interface, Cambridge, MA: MIT Press.
Lightfoot, David (1979) Principles of Diachronic Syntax, Cambridge: Cambridge University
Press.
Longobardi, Giuseppe (1996) The syntax of N-raising: A minimalist theory. OTS
publications, Utrecht University, Research Institute for Language and Speech,
OTS-WP-TL-96-105.
Lumsden, Michael (1988) Existential Sentences: Their Structure and Meaning, London: Croom
Helm.
TSD01 07/18/2005, 03:57PM375
376 Bibliography
McCawley, James D. (1988) The Syntactic Phenomena of English, Chicago: University of
Chicago Press.
McCloskey, Jim (1996) Subjects and subject positions in Irish. In Robert D. Borsley and
Ian Roberts (eds), The Syntax of Celtic Languages, Cambridge: Cambridge University
Press, pp. 241–83.
McCloskey, Jim (1997) Subjects and subject positions. In Liliane Haegeman (ed.), Elements
of Grammar: A Handbook of Generative Syntax, Dordrecht: Kluwer, pp. 197–235.
McCloskey, Jim (2000) Quantiﬁer ﬂoat and wh-movement in an Irish English. Linguistic
Inquiry, 31, 57–84.
McCloskey, Jim (2001) The distribution of subject properties in Irish. In William D. Davies
and Stanley Dubinsky (eds), Objects and other Subjects, Dordrecht: Kluwer, pp. 157–
92.
McCloskey, Jim (forthcoming) Questions and questioning in a local English. In Raffaella
Zanuttini, Hector Campos, Elena Herburger, and Paul Portner (eds), Negation, Tense and
Clausal Architecture: Cross-linguistic Investigations, Georgetown: Georgetown University
Press.
Miller, Jim (1993) The Grammar of Scottish English. In James Milroy and Lesley Milroy
(eds), Real English: The Grammar of English Dialects of the British Isles, London: Longman,
pp. 99–138.
Miller, Philip (2001) Discourse constraints on (non) extraposition from subject in English.
Linguistics, 39, 683–701.
Milsark, Gary Lee (1974) Existential Sentences in English. PhD dissertation, MIT.
Milsark, Gary Lee (1979) Toward an explanation of certain peculiarities of the existential
construction in English. Linguistic Analysis, 3, 1–30.
Moro, Andrea (1997) The Raising of Predicates: Predicative Noun Phrases and the Theory
of Phrase Structure, Cambridge: Cambridge University Press.
Newmeyer, Frederick (1983) Grammatical Theory: Its Limits and its Possibilities, Chicago
and London: University of Chicago Press.
Newmeyer, Frederick (2001) Grammatical functions, thematic roles, and phrase structure:
Their underlying disunity. In William D. Davies and Stanley Dubinsky, Objects and other
Subjects, Dordrecht: Kluwer, pp. 53–76.
Nilsen, Øystein (2004) Domains for adverbs. In Artemis Alexiadou (ed.), Adverbs across
Frameworks, special edition of Lingua, 114, 779–808.
Oirsouw, Robert van (1987) The Syntax of Coordination, London: Croom Helm.
Ouhalla, Jamal and Ur Shlonsky (2002) Themes in Arabic and Hebrew Syntax, Dordrecht:
Kluwer.
Panagiotidis, Phaevos (2002) Pronouns, Clitics and Empty Nouns, New York and Amsterdam:
John Benjamins.
Panagiotidis, Phaevos (2003) One, empty nouns, and θ-assignment. Linguistic Inquiry, 34,
281–92.
Penke, Martina and Anette Rosenbach (eds) (2004) What counts as evidence in linguistics?
The case of innateness. Studies in Language, 28, 4.
Perlmutter, Paul (1971) Deep and Surface Structure Constraints in Syntax, New York: Holt,
Rinehart and Winston.
Pesetsky, David and Esther Torrego (2000) T-to-C movement: Causes and consequences.
In Michael Kenstowicz (ed.), Ken Hale: A Life in Language, Cambridge, MA: MIT Press,
pp. 355–426.
TSD01 07/18/2005, 03:57PM376
Bibliography 377
Pintzuk, Suzan (1991) Phrase Structure in Competition: Variation and Change in Old English
Word Order. PhD dissertation, University of Pennsylvania.
Pollock, Jean-Yves (1989) Verb movement, UG and the structure of IP. Linguistic Inquiry,
20, 365–425.
Pollock, Jean-Yves (1994) Checking theory and bare verbs. In Guglielmo Cinque, Jan Koster,
Jean-Yves Pollock, Luigi Rizzi, and Raffaella Zanuttini (eds), Paths toward Universal Grammar,
Georgetown: Georgetown University Press, pp. 293–310.
Pollock, Jean-Yves (1997) Notes on clause structure. In Liliane Haegeman (ed.), Elements of
Grammar: A Handbook of Generative Syntax, Dordrecht: Kluwer, pp. 237–80.
Prince, Ellen (1981) Toward a taxonomy of given/new information. In Peter Cole (ed.),
Radical Pragmatics, New York: Academic Press, pp. 223–54.
Radford, Andrew (2004) Syntactic Theory and English Syntax. Oxford: Blackwell.
Reinhart, Tanya (1981) Pragmatics and linguistics: An analysis of sentence topics. Philosophica,
27, 53–94.
Reinhart, Tanya (2000) The theta system: Syntactic realisation of verbal concepts. OTS
Working Papers in Linguistics (00,01/TL), February. (To appear (extended) in the LI
Monographs Series, MIT Press.)
Richards, Norvin (2001) Movement in Language: Interactions and Architectures, Oxford
and New York: Oxford University Press.
Rizzi, Luigi (1990) Relativized Minimality, Cambridge, MA: MIT Press.
Rizzi, Luigi (1996) Residual V-Second and the Wh-Criterion. In Adriana Belletti and Luigi
Rizzi (eds), Parameters and Functional Heads: Essays in Comparative Syntax, New York
and Oxford: Oxford University Press, pp. 63–90.
Rizzi, Luigi (1997) The ﬁne structure of the left periphery. In Liliane Haegeman (ed.), Elements
of Grammar: A Handbook of Generative Syntax, Dordrecht: Kluwer, pp. 281–337.
Rizzi, Luigi (2001) On the position Int(errogative) in the left periphery of the clause. In
Guglielmo Cinque and Giorgio P. Salvi (eds), Current Studies in Italian Syntax: Essays
Offered to Lorenzo Renzi, Oxford: Elsevier, North Holland.
Roberts, Ian (1985) Agreement parameters and the development of English modal auxiliaries.
Natural Language and Linguistic Theory, 3, 21–58.
Roberts, Ian (1990) Some notes on VP-fronting and head government. In Joan Mascaró and
Marina Nespor (eds), Grammar in Progress, Dordrecht: Foris, pp. 386–96.
Roberts, Ian (1993) Verbs and Diachronic Syntax: A Comparative study of English and
French, Dordrecht: Kluwer.
Roberts, Ian (1998) Have/be raising, move F, and procrastinate. Linguistic Inquiry, 29,
113–25.
Roberts, Ian (2005) Principles and Parameters in VSO Languages, New York and Oxford:
Oxford University Press.
Rohrbacher, Bernhard (1999) Morphology-Driven Syntax: A Theory of V to I Raising and
Pro-drop, New York and Amsterdam: John Benjamins.
Rowlett, Paul (1998) Sentential Negation in French, Oxford: Oxford University Press.
Rowlett, Paul (2005) The Syntax of French, Cambridge: Cambridge University Press.
Saﬁr, Ken (1985) Syntactic Chains, Cambridge: Cambridge University Press.
Schütze, Carson (1999) English expletive constructions are not infected. Linguistic Inquiry,
30, 467–84.
Shlonsky, Ur (1991) Quantiﬁers as functional heads: A study of quantiﬁer ﬂoat in Hebrew.
Lingua, 84, 159–80.
TSD01 07/18/2005, 03:57PM377
378 Bibliography
Shlonsky, Ur (1997) Clause Structure and Word Order in Hebrew and Arabic, Oxford:
Oxford University Press.
Sobin, Nicholas (1997) Agreement, default rules and grammatical viruses. Linguistic Inquiry,
28, 318–43.
Sperber, Dan and Deirdre Wilson (1986) Relevance, Oxford: Blackwell.
Sportiche, Dominique (1988) A theory of ﬂoating quantiﬁers and its corollaries for phrase
structure. Linguistic Inquiry, 19, 425–40.
Stubbs, Michael (2002) On text and corpus analysis: A reply to Borsley and Ingham. Lingua,
112 (1), 7–11.
Stuurman, Frits (1989) Generative grammar and descriptive grammar: Beyond juxtaposition.
In Gottfried Graustein and Gerhard Leitner (eds), Reference Grammars and Modern
Linguistic Theory, Tübingen: Niemeyer, pp. 229–54.
Svenonius, Peter (2002) Subject positions and the placement of adverbials. In Peter Svenonius
(ed.), Subjects, Expletives, and the EPP, Oxford: Oxford University Press, pp. 201–42.
Szabolcsi, Anna (1983) The possessor that ran away from home. Linguistic Review, 1, 89–
102.
Szabolcsi, Anna (1994) The noun phrase. Syntax and Semantics, 27: The Syntactic Structure
of Hungarian, New York: Academic Press, pp. 179–274.
Vikner, Sten (1995) V° to I° movement and inﬂection for person in all tenses. Working
Papers in Scandinavian Syntax, 55, 1–27.
Vikner, Sten (1997) V to I movement and inﬂection for person in all tenses. In Liliane
Haegeman (ed.), The New Comparative Syntax, London: Longman, Addison and Wesley,
pp. 189–213.
Ward, Gregory L. (1988) The Semantics and Pragmatics of Preposing, New York and London:
Garland Publishing.
Wilkins, Wendy (1988) Syntax and Semantics 21: Thematic Relations. New York: Academic
Press.
Williams, Edwin (1981) Argument structure and morphology. Linguistic Review, 1, 81–114.
Zanuttini, Raffaella (1997a) Negation and Clausal Structure: A Comparative Study of
Romance Languages, New York and Oxford: Oxford University Press.
Zanuttini, Raffaella (1997b) Negation and V-movement. In Liliane Haegeman (ed.), The
New Comparative Syntax, London: Longman, Addison and Wesley, pp. 214–45.
Zanuttini, Raffaella and Paul Portner (2003) Exclamative clauses: At the syntax–semantics
interface. Language, 79, 29–81.
Zwart, Jan-Wouter (1997) The Germanic SOV languages and the universal base. In Liliane
Haegeman (ed.), The New Comparative Syntax, London: Longman, Addison and Wesley,
pp. 246–67.
Sources of examples
Newspapers
The Atlanta Journal-Constitution
The Chicago Tribune
The Guardian
TSD01 07/18/2005, 03:58PM378
Bibliography 379
The Independent
The Independent on Sunday
The Los Angeles Times
The New York Times
The Observer
San Francisco Chronicle
The Sunday Times
The Times
USA Today
The Washington Post
The Wall Street Journal
Fiction
Dexter, Colin (1992) The Jewel that was Ours, London: Macmillan. First published 1991,
Pan Books.
Dexter, Colin (1995) The Daughters of Cain, London: Pan Books. First published 1994,
Macmillan.
Dexter, Colin (2000) The Remorseful Day, London: Pan Books.
Francis Fyﬁeld (2001) Undercurrents, Warner Books 2001. First published 2001, Little Brown
and Company.
Frantzen, Jonathan (2001) The Corrections, New York: Fourth Estate.
George, Elizabeth (1993/1996) Missing Joseph, London: Bantam Books.
Lloyd, Josie and Emlyn Rees (1999) Come Together, London: Arrow Books.
McCarthy, Mary (1989) The Company She Keeps, Harmondsworth: Penguin. First
published 1942, Weidenfeld and Nicolson.
Muller, Marcia (1993) Edwin of the Iron Shoes, London: The Women’s Press. First
published in the US in 1977, McKay Washburn.
Rankin, Ian (2001) The Falls, London: Orion.
Spark, Muriel (1963) The Bachelors, Harmondsworth: Penguin. First published 1960,
Macmillan.
Spark, Muriel (1977) Memento Mori, Harmondsworth: Penguin. First published 1959,
Macmillan.
Townsend, Sue (1993) Adrian Mole: The Wilderness Years, London: Mandarin.
Wharton, Edith (1998) The House of Mirth, New York: Berkley. First published 1905.
Non-ﬁction
Bryson, Bill (1990) The Lost Continent, London: Abacus. First published 1989, Martin
Secker and Warburg.
Bryson, Bill (1996) Notes from a Small Island, London: Swan. First published 1995, Doubleday.
Glendinning,Victoria (1984) Vita, Harmondsworth: Penguin.
MacCarthy, Fiona (1995) William Morris, London: Faber and Faber.
Shields, Carol (2001) Jane Austin, London: Phoenix. First published in the UK 1991,
Weidenfeld and Nicolson.
TSD01 07/18/2005, 03:58PM379
380 Bibliography
Dictionaries
Concise Oxford Dictionary of Current English (6th edition) (1976), edited by J. B. Sykes.
Oxford: Clarendon Press. First edition 1911.
Longman Dictionary of English Language and Culture (1998). London: Longman, Addison,
Wesley.
Webster’s Third New International Dictionary of the English Language (1981). Springﬁeld,
MA: Merriam-Webster.
TSD01 07/18/2005, 03:58PM380
Index 381
Index
Italicized page numbers refer to exercises.
a 108
abbreviations viii
actor see agent
adjoined clause 364–5
adjunct
complement vs. 95–7
movement 365–6
adjunction 212–15
attachment of 138–41, 187–8
and core constituent 241–3
direction 97–9, 187–8
ﬂoating quantiﬁer and 278–81, 280–3
French 280–1
manner 279
and noun phrases 117–18
and sentence 141–4
afﬁx 164
agent 192, 222–7
agreement relation 111–12, 114
all 337
ambiguity 7–11, 49–50, 68
structural 70, 130, 132–4
anaphora 74
antecedent 74, 345
Arabic 271–5
articles 108
articulated structure 92–3
associating V and I 166–9
asymmetry
subject-object 74, 77, 93, 339
auxiliaries 22, 51
clustering (hypothesis) 197–9
distributional properties 40–1
ﬁnite 202–6
as heads of VP (hypothesis) 199–202
modal 205–6
multiple 197–209, 262
strings 137–8
structure of sequences 206–9
subject positions and 262–6
and verb phrases 93–5, 116
vs. lexical verb 195–6
see also subject-auxiliary inversion
backgrounding 85–6
base form 39
base position 172, 190, 240, 248
and movement 100–1
be 40, 52, 203–5, 207–8, 262
+ passive 227–9
Belfast English 260, 270
beneﬁciary 192
binary branching 94–5, 102–5, 198
noun phrase 107–10
word structure 105–7
blocking movement 334–7, 340
bound morpheme 38, 164
bracketing 9, 72, 79, 126
branches 77
binary structure 102–5, 198
ternary structure 198
canonical position 79–80, 101
categories 75
Chomsky, Noam vi, 186
clauses 30, 317
cleft sentences 85–7
closed class words 41, 205, 317
clustering auxiliaries 197–9
co-ordination 89–92, 134–5
representing 131–2
structural ambiguity and 130
TSD02-Index 07/18/2005, 03:58PM381
382 Index
diachronic linguistics 45
diachronic variation 44–5
diagnostics for structure 68–92, 74
dialectal variation 33
Dik, Simon 15
direct object NP 192
direct question 30, 34, 313–14
direct subject interrogative 328–30
discontinuous constituent 250
displaced constituent 247–9
distributional information 39
do/does/did 25, 26, 27–8
-insertion 164–6, 171, 173, 178, 180,
329, 330
emphatic 52–4
dominance 103
dummy element 256
Dutch 43–4
economy 16–18, 26, 27–8, 81
-ed 37–8, 160, 161, 178
ellipsis 54–5, 84–5, 183, 275–7
embedded clause 30, 140, 317
question 30, 33, 327–8
emphatic do 52–4
empirical argument 15, 32, 34, 79, 238
exclamatives 357
existential pattern 253–62, 257, 269–70
Belfast English 270
verb choices 258–62
experiencer 192
expletive subject 256
expletive there 256, 258
extraction 362–3, 364–5
features 111
ﬁnite verb form 25, 38
ﬁnite auxiliaries 202–6
ﬁnite clauses 183–4
ﬁnite inﬂection 163
ﬂat structure 92–3
ﬂoating quantiﬁer 250–3, 269, 271
and adjunct 278–81
Hebrew 281–3, 295–6
and passivization 288–96
to inﬁnitives 270
verb positions 277–8
for 341
foregrounding 85–6
form-interpretation relation 19, 28–36,
241–6, 246
free morphemes 38, 164
comparative syntax 41–4, 45–6, 217–20
ﬂoating quantiﬁer 277–8
OV orders 148–51
verb position 171–81, 277–8
competence 12
complement 67, 115–16
vs. adjunct 95–7
complementary distribution 113, 318
complementizer 317, 338–40, 348, 351–2,
357–9
complex sentences 140
compositionality 11
conceptual arguments 238
conditional clause 31, 32, 319
conjunctions 316–20, 317, 341
constituency tests 123–4, 152–3
constituent negation 359
constituent question 83, 314–15
constituent structure 123–6, 210–11
evidence for 127–8
noun phrases 123–6, 137
constituents 72, 75, 82, 239–40
binary branching structure 108
discontinuous 250
displaced 247–9
focalizing 85–7, 128–30
immediate 74, 115
moved 220–2
ultimate 74
wh- 83
constraints
on movement 337, 346–8, 362–3
constructed examples 12
copies 247, 248–9, 320
evidence for 320–2
copula 52, 203–5, 207–8
core constituents 241–3
counter-evidence 27–8, 30, 33
crash 327
Creole 218–20
cross-linguistic variation 44
Crystal, David 4, 16
Danish 217–18
declaratives 22
deﬁniteness effect 258
deletion 84–5
demonstratives 108, 110–11
derivation 23, 105, 334
determiner 108, 109, 112, 124–6,
257–8
deverbal noun 145–8, 224
TSD02-Index 07/18/2005, 03:58PM382
Index 383
French
adjunction 280–1
ﬂoating quantiﬁer 278, 280–1
inﬂectional patterns 176–7, 178,
196–7
negation 179
past tense paradigm 176
postverbal subject 367–9
present tense paradigm 176
question formation 171–4
speciﬁers in NP 111, 113–14
subject-verb inversion 42–3, 172,
177–8
verb position 171–7
fronting 81–2, 170, 355–6
interrogative constituents 343
full verbs 41
functional head 196–7, 296–301
functional projection 155–236, 196,
229–30
genitive see prenominal genitive
German 43–4
grammatical subject 253–7
Haitian Creole 218–20
have 203–5, 207–8, 262
head 67, 75, 158–9
agreement with speciﬁer 110–12
auxiliary as, in VP 199–202
functional 196–7, 296–301
INFL as 161–4
lexical 196–7, 229–30
and projection 101–2
Hebrew 281–3, 295–6
hierarchical structure 92–3
hypotheses 12, 14, 15, 18, 79
binary composition 105
I
associating with V 166–9
do-insertion 164–6
I-to-C movement 329
Icelandic 217–18, 232–3, 261–2, 284–5
if 317, 318, 319, 324
immediate constituent 74
independent question see direct question
indirect constituent question 322–5
indirect question see embedded question
inﬁnitival clause 183
inﬁnitive 39
with to 270
inﬂection
ﬁnite 163
French vs. English 176–7, 178,
196–7
non-ﬁnite 230–1
question formation 178
relation with VP 161–71
strong 177
inﬂectional morpheme 38, 157
-ing 38
instrument 192
intermediate projection 119
interrogative 22, 343
intervention effect 334–40, 346–7
inversion 40
complementizer 357–9
direct question 177–8
direct subject interrogative 328–30
French 177–8
locative 57–61
negative 31, 32
non-adjacent 359–60
postverbal subject (French) 367–9
predicate 61–4, 301–3
verbs 35–6
see also subject-auxiliary inversion
Irish 275–7
labeled bracketing 79
landing site 101, 240
language typology 303–4
Latin 160
layered structure 95–101
left-adjunction 98, 187–8
lexical head 196–7, 229–30
lexical projection 155–236, 196
lexical verb 41
vs. auxiliary 195–6
locative inversion 57–61
logical subject 256
long movement 331–3
manner adjunct 279
marked word order 81
maximal projection 119
meaning relations 48–9
and structure 68–72, 191–7
Merge 157, 186–7, 189–90, 320
constituents 105, 107
multiple 249
methodology vi, ix, 3, 13–19
minimalism x, 186
TSD02-Index 07/18/2005, 03:58PM383
384 Index
modal auxiliaries 205–6
and to 215–16
Modern Greek 285–6
modiﬁer 117
morpheme 38
morphology 38–9
Move 157, 186–7, 190–1, 220–2, 320,
331–40
movement 79–82, 232–3
adjunct clause 365–6
base position and 100–1
blocking 334–7, 340
constraints on 337, 346–8
displaced constituent 247–9
I-to-C 329
long 331–3
object 338–40
relative pronoun 341–2: non-overt 350–2
rightward focus 128–30
subject 262–5, 340, 348, 351–2
verbs 232–3
wh- 336, 360–1
see also step-by-step movement
multiple auxiliaries 197–209, 262
multiple meaning 48–9
multiple merge 249
multiple questions 325–7
must 240
negation 179–81, 216–17, 359
inversion 31, 32
Newmeyer, F. 17
node 77, 115
non-adjacent inversion 359–60
non-ﬁnite clause
for as conjunction 341
interrogatives 341–4
with to 181–6
vs. ﬁnite clause 183–4
non-ﬁnite inﬂection 230–1
notional subject 256
noun phrases (NP) 74–5
adjunction 117–18
binary branching structure 107–10
constituent structure 123–6, 137
prenominal genitive 112–14
speciﬁer–head agreement 110–12
speciﬁer 110–17, 153–4
object 68–9, 316
movement 338–40
realization 135–7
object clause 317
object shift 232–3
Ockham’s Razor 17
Old English 44–5, 173, 178, 218
open class words 41
ordering restrictions 213
OV orders 99–100, 148–51
overgeneration 258
parameter 42
participle 203, 208
passivization 223, 225, 227–9, 288–96
past tense 37–8
patient 192
possessive pronoun 120
possessor 192
postverbal subject 367–9
predicate inversion 61–4, 301–3
predication 95
prenominal genitive 112–14, 120
prepositional phrase (PP) 75
present participle 38
present tense paradigm
French and English 176
Old English 178
presentation 86, 87
projection
functional 155–236, 196, 229–30
head and 101–2
intermediate 119
lexical 155–236
maximal 119
nominal x
provisional subject 256
pseudo-cleft sentence 87–9
psychological verb 193
quantiﬁer see ﬂoating quantiﬁer
question formation 19–41, 82–4
direct yes/no 313–14
French 171–4
indirect 322–5, 327–8
subject-auxiliary inversion 21–4,
28–36
relative clause 344–52, 345, 349–50,
363–4
relative pronoun 345, 366–7
movement 341–2, 346–7, 350–2
non-overt 349–52
zero 349–50
reported question see embedded question
TSD02-Index 07/18/2005, 03:58PM384
Index 385
representations
bracketing 126
co-ordination 131–2
structure 126–7
resumptive pronoun 363–5
right-adjunction 98, 187–8
rightward focus 128–30
scope relations 139, 214, 242
selection 191
sentence
cleft 85–7
complex 140
negation 179–81, 216–17
pseudo-cleft 87–9
structure 158–86, 211–12
verb phrase and 92–107
see also negative sentence
so 57
SpecCP 319, 334–7, 354–5
speciﬁer 67, 107–21, 110, 124–6
agreement with head 110–12
and multiple questions 325–7
in NP 110–17, 153–4: French 111,
113–14
and VPs 243–6
zero VP 119–20
speciﬁer–head agreement 110–12
SpecIP 249, 250
SpecNP 243
SpecVP 244, 245, 250
SpecXP 314–16
spell-out 240, 246
split inﬁnitive 185–6
Split Inﬂ Hypothesis 301
square brackets 9, 72
labeled 79
step-by-step movement 334–40
empirical support for 265–6, 337–8
non-overt pronoun 350–1
strikethrough 240, 248, 249
strong inﬂection 177, 178
structural ambiguity 70, 130, 132–4
structure 9–10, 14, 46
diagnostics for 68–92, 74
meaning relations and 68–72,
191–7
representation 126–7
see also sentence structure
subject 22, 120
absence 159
expletive 256
grammatical 253–7
link with VP 160–1, 194–5
logical 256
notional 256
postverbal 367–9
provisional 256
realization 135–7
requirement 160–1
thematic 253–7
and verbs 194–5
subject interrogative 327–30
subject position 234–5, 237–304
and auxiliaries 262–6: see also
subject-auxiliary inversion
Arabic 271–5
ﬂoating quantiﬁer 292–3
Irish 275–7
Modern Greek 287–8
movement 262–5, 340, 348, 351–2
passive sentence 292–3
transitive expletive construction
(Icelandic) 284–5
VSO order 285–8
Welsh 283
subject requirement 160–1
subject-auxiliary inversion 21–4, 28–36
conjunction and 316–20
subject-object asymmetry 74, 93
subject-verb inversion (SVI) 43–4, 45,
172
subordinate clause 30, 317
subordinating conjunction 140, 317
substitution 74–9
by so 57
constituents containing verbs 55–7
synonym 19
syntax 20–1, 70
ternary branching 198
that 317, 338–40, 348, 351–2, 366–7
-omission 330–1
the 108
thematic role 191–3, 193–4, 222–7
thematic subject 253–7
theme 192
there 256, 258, 259–60
to 181–6
to inﬁnitive 270
transitive expletive construction (Icelandic)
284–5
transitive verb 96
tree diagram 73–4, 76–7, 78, 79
TSD02-Index 07/18/2005, 03:58PM385
386 Index
ultimate constituent 74
ungrammaticality 100
unmarked word order 81
verb phrase (VP) 75–9
auxiliary and 93–5
auxiliary as head 199–202
complement vs. adjunct 95–7
ellipsis 85
and inﬂection 161–71
layered 78, 88
link with subject 160–1, 194–5
question about 118–21
and sentence 92–107
verbs
deﬁnition 37–41
full 41
inversion 35–6
lexical 41, 195–6
movement 232–3
position 171–81, 277–8
subject and 194–5
VP ellipsis 275–7
VP-fronting 170, 171
VP-Internal Subject Hypothesis 247–62,
249
empirical support 350–62
VSO languages 271–5
Arabic 271–5
Irish 275–7
Modern Greek 287–8
subject positions 285–8
VP ellipsis 275–7
Welsh 283
Welsh 283
West Ulster English 322, 337, 338
wh-constituent 83, 314, 315, 324, 326
wh-movement 336, 360–1
wh-question 83, 315
whether 317, 318, 319, 324, 341–3,
356
word order 212–15
marked 81–2
unmarked 81
see also question formation
word structure 105–7
words
closed class 41, 205, 317
open class 41
X-bar theory 102
yes/no question 313–14
TSD02-Index 07/18/2005, 03:58PM386
Index 387
TSD02-Index 07/18/2005, 03:58PM387
TSD02-Index 07/18/2005, 03:58PM388